feat: add filter presets and update version to v0.9.15

- Added filter presets feature allowing users to save and quickly switch between filter combinations
- Fixed various bugs to improve overall stability
- Updated project version from 0.9.14 to 0.9.15 in pyproject.toml

.agents/skills/lora-manager-e2e/SKILL.md

@@ -0,0 +1,201 @@
+---
+name: lora-manager-e2e
+description: End-to-end testing and validation for LoRa Manager features. Use when performing automated E2E validation of LoRa Manager standalone mode, including starting/restarting the server, using Chrome DevTools MCP to interact with the web UI at http://127.0.0.1:8188/loras, and verifying frontend-to-backend functionality. Covers workflow validation, UI interaction testing, and integration testing between the standalone Python backend and the browser frontend.
+---
+
+# LoRa Manager E2E Testing
+
+This skill provides workflows and utilities for end-to-end testing of LoRa Manager using Chrome DevTools MCP.
+
+## Prerequisites
+
+- LoRa Manager project cloned and dependencies installed (`pip install -r requirements.txt`)
+- Chrome browser available for debugging
+- Chrome DevTools MCP connected
+
+## Quick Start Workflow
+
+### 1. Start LoRa Manager Standalone
+
+```python
+# Use the provided script to start the server
+python .agents/skills/lora-manager-e2e/scripts/start_server.py --port 8188
+```
+
+Or manually:
+```bash
+cd /home/miao/workspace/ComfyUI/custom_nodes/ComfyUI-Lora-Manager
+python standalone.py --port 8188
+```
+
+Wait for server ready message before proceeding.
+
+### 2. Open Chrome Debug Mode
+
+```bash
+# Chrome with remote debugging on port 9222
+google-chrome --remote-debugging-port=9222 --user-data-dir=/tmp/chrome-lora-manager http://127.0.0.1:8188/loras
+```
+
+### 3. Connect Chrome DevTools MCP
+
+Ensure the MCP server is connected to Chrome at `http://localhost:9222`.
+
+### 4. Navigate and Interact
+
+Use Chrome DevTools MCP tools to:
+- Take snapshots: `take_snapshot`
+- Click elements: `click`
+- Fill forms: `fill` or `fill_form`
+- Evaluate scripts: `evaluate_script`
+- Wait for elements: `wait_for`
+
+## Common E2E Test Patterns
+
+### Pattern: Full Page Load Verification
+
+```python
+# Navigate to LoRA list page
+navigate_page(type="url", url="http://127.0.0.1:8188/loras")
+
+# Wait for page to load
+wait_for(text="LoRAs", timeout=10000)
+
+# Take snapshot to verify UI state
+snapshot = take_snapshot()
+```
+
+### Pattern: Restart Server for Configuration Changes
+
+```python
+# Stop current server (if running)
+# Start with new configuration
+python .agents/skills/lora-manager-e2e/scripts/start_server.py --port 8188 --restart
+
+# Wait and refresh browser
+navigate_page(type="reload", ignoreCache=True)
+wait_for(text="LoRAs", timeout=15000)
+```
+
+### Pattern: Verify Backend API via Frontend
+
+```python
+# Execute script in browser to call backend API
+result = evaluate_script(function="""
+async () => {
+  const response = await fetch('/loras/api/list');
+  const data = await response.json();
+  return { count: data.length, firstItem: data[0]?.name };
+}
+""")
+```
+
+### Pattern: Form Submission Flow
+
+```python
+# Fill a form (e.g., search or filter)
+fill_form(elements=[
+    {"uid": "search-input", "value": "character"},
+])
+
+# Click submit button
+click(uid="search-button")
+
+# Wait for results
+wait_for(text="Results", timeout=5000)
+
+# Verify results via snapshot
+snapshot = take_snapshot()
+```
+
+### Pattern: Modal Dialog Interaction
+
+```python
+# Open modal (e.g., add LoRA)
+click(uid="add-lora-button")
+
+# Wait for modal to appear
+wait_for(text="Add LoRA", timeout=3000)
+
+# Fill modal form
+fill_form(elements=[
+    {"uid": "lora-name", "value": "Test LoRA"},
+    {"uid": "lora-path", "value": "/path/to/lora.safetensors"},
+])
+
+# Submit
+click(uid="modal-submit-button")
+
+# Wait for success message or close
+wait_for(text="Success", timeout=5000)
+```
+
+## Available Scripts
+
+### scripts/start_server.py
+
+Starts or restarts the LoRa Manager standalone server.
+
+```bash
+python scripts/start_server.py [--port PORT] [--restart] [--wait]
+```
+
+Options:
+- `--port`: Server port (default: 8188)
+- `--restart`: Kill existing server before starting
+- `--wait`: Wait for server to be ready before exiting
+
+### scripts/wait_for_server.py
+
+Polls server until ready or timeout.
+
+```bash
+python scripts/wait_for_server.py [--port PORT] [--timeout SECONDS]
+```
+
+## Test Scenarios Reference
+
+See [references/test-scenarios.md](references/test-scenarios.md) for detailed test scenarios including:
+- LoRA list display and filtering
+- Model metadata editing
+- Recipe creation and management
+- Settings configuration
+- Import/export functionality
+
+## Network Request Verification
+
+Use `list_network_requests` and `get_network_request` to verify API calls:
+
+```python
+# List recent XHR/fetch requests
+requests = list_network_requests(resourceTypes=["xhr", "fetch"])
+
+# Get details of specific request
+details = get_network_request(reqid=123)
+```
+
+## Console Message Monitoring
+
+```python
+# Check for errors or warnings
+messages = list_console_messages(types=["error", "warn"])
+```
+
+## Performance Testing
+
+```python
+# Start performance trace
+performance_start_trace(reload=True, autoStop=False)
+
+# Perform actions...
+
+# Stop and analyze
+results = performance_stop_trace()
+```
+
+## Cleanup
+
+Always ensure proper cleanup after tests:
+1. Stop the standalone server
+2. Close browser pages (keep at least one open)
+3. Clear temporary data if needed

.agents/skills/lora-manager-e2e/references/mcp-cheatsheet.md

@@ -0,0 +1,324 @@
+# Chrome DevTools MCP Cheatsheet for LoRa Manager
+
+Quick reference for common MCP commands used in LoRa Manager E2E testing.
+
+## Navigation
+
+```python
+# Navigate to LoRA list page
+navigate_page(type="url", url="http://127.0.0.1:8188/loras")
+
+# Reload page with cache clear
+navigate_page(type="reload", ignoreCache=True)
+
+# Go back/forward
+navigate_page(type="back")
+navigate_page(type="forward")
+```
+
+## Waiting
+
+```python
+# Wait for text to appear
+wait_for(text="LoRAs", timeout=10000)
+
+# Wait for specific element (via evaluate_script)
+evaluate_script(function="""
+() => {
+    return new Promise((resolve) => {
+        const check = () => {
+            if (document.querySelector('.lora-card')) {
+                resolve(true);
+            } else {
+                setTimeout(check, 100);
+            }
+        };
+        check();
+    });
+}
+""")
+```
+
+## Taking Snapshots
+
+```python
+# Full page snapshot
+snapshot = take_snapshot()
+
+# Verbose snapshot (more details)
+snapshot = take_snapshot(verbose=True)
+
+# Save to file
+take_snapshot(filePath="test-snapshots/page-load.json")
+```
+
+## Element Interaction
+
+```python
+# Click element
+click(uid="element-uid-from-snapshot")
+
+# Double click
+click(uid="element-uid", dblClick=True)
+
+# Fill input
+fill(uid="search-input", value="test query")
+
+# Fill multiple inputs
+fill_form(elements=[
+    {"uid": "input-1", "value": "value 1"},
+    {"uid": "input-2", "value": "value 2"},
+])
+
+# Hover
+hover(uid="lora-card-1")
+
+# Upload file
+upload_file(uid="file-input", filePath="/path/to/file.safetensors")
+```
+
+## Keyboard Input
+
+```python
+# Press key
+press_key(key="Enter")
+press_key(key="Escape")
+press_key(key="Tab")
+
+# Keyboard shortcuts
+press_key(key="Control+A")  # Select all
+press_key(key="Control+F")  # Find
+```
+
+## JavaScript Evaluation
+
+```python
+# Simple evaluation
+result = evaluate_script(function="() => document.title")
+
+# Async evaluation
+result = evaluate_script(function="""
+async () => {
+    const response = await fetch('/loras/api/list');
+    return await response.json();
+}
+""")
+
+# Check element existence
+exists = evaluate_script(function="""
+() => document.querySelector('.lora-card') !== null
+""")
+
+# Get element count
+count = evaluate_script(function="""
+() => document.querySelectorAll('.lora-card').length
+""")
+```
+
+## Network Monitoring
+
+```python
+# List all network requests
+requests = list_network_requests()
+
+# Filter by resource type
+xhr_requests = list_network_requests(resourceTypes=["xhr", "fetch"])
+
+# Get specific request details
+details = get_network_request(reqid=123)
+
+# Include preserved requests from previous navigations
+all_requests = list_network_requests(includePreservedRequests=True)
+```
+
+## Console Monitoring
+
+```python
+# List all console messages
+messages = list_console_messages()
+
+# Filter by type
+errors = list_console_messages(types=["error", "warn"])
+
+# Include preserved messages
+all_messages = list_console_messages(includePreservedMessages=True)
+
+# Get specific message
+details = get_console_message(msgid=1)
+```
+
+## Performance Testing
+
+```python
+# Start trace with page reload
+performance_start_trace(reload=True, autoStop=False)
+
+# Start trace without reload
+performance_start_trace(reload=False, autoStop=True, filePath="trace.json.gz")
+
+# Stop trace
+results = performance_stop_trace()
+
+# Stop and save
+performance_stop_trace(filePath="trace-results.json.gz")
+
+# Analyze specific insight
+insight = performance_analyze_insight(
+    insightSetId="results.insightSets[0].id",
+    insightName="LCPBreakdown"
+)
+```
+
+## Page Management
+
+```python
+# List open pages
+pages = list_pages()
+
+# Select a page
+select_page(pageId=0, bringToFront=True)
+
+# Create new page
+new_page(url="http://127.0.0.1:8188/loras")
+
+# Close page (keep at least one open!)
+close_page(pageId=1)
+
+# Resize page
+resize_page(width=1920, height=1080)
+```
+
+## Screenshots
+
+```python
+# Full page screenshot
+take_screenshot(fullPage=True)
+
+# Viewport screenshot
+take_screenshot()
+
+# Element screenshot
+take_screenshot(uid="lora-card-1")
+
+# Save to file
+take_screenshot(filePath="screenshots/page.png", format="png")
+
+# JPEG with quality
+take_screenshot(filePath="screenshots/page.jpg", format="jpeg", quality=90)
+```
+
+## Dialog Handling
+
+```python
+# Accept dialog
+handle_dialog(action="accept")
+
+# Accept with text input
+handle_dialog(action="accept", promptText="user input")
+
+# Dismiss dialog
+handle_dialog(action="dismiss")
+```
+
+## Device Emulation
+
+```python
+# Mobile viewport
+emulate(viewport={"width": 375, "height": 667, "isMobile": True, "hasTouch": True})
+
+# Tablet viewport
+emulate(viewport={"width": 768, "height": 1024, "isMobile": True, "hasTouch": True})
+
+# Desktop viewport
+emulate(viewport={"width": 1920, "height": 1080})
+
+# Network throttling
+emulate(networkConditions="Slow 3G")
+emulate(networkConditions="Fast 4G")
+
+# CPU throttling
+emulate(cpuThrottlingRate=4)  # 4x slowdown
+
+# Geolocation
+emulate(geolocation={"latitude": 37.7749, "longitude": -122.4194})
+
+# User agent
+emulate(userAgent="Mozilla/5.0 (Custom)")
+
+# Reset emulation
+emulate(viewport=None, networkConditions="No emulation", userAgent=None)
+```
+
+## Drag and Drop
+
+```python
+# Drag element to another
+drag(from_uid="draggable-item", to_uid="drop-zone")
+```
+
+## Common LoRa Manager Test Patterns
+
+### Verify LoRA Cards Loaded
+
+```python
+navigate_page(type="url", url="http://127.0.0.1:8188/loras")
+wait_for(text="LoRAs", timeout=10000)
+
+# Check if cards loaded
+result = evaluate_script(function="""
+() => {
+    const cards = document.querySelectorAll('.lora-card');
+    return {
+        count: cards.length,
+        hasData: cards.length > 0
+    };
+}
+""")
+```
+
+### Search and Verify Results
+
+```python
+fill(uid="search-input", value="character")
+press_key(key="Enter")
+wait_for(timeout=2000)  # Wait for debounce
+
+# Check results
+result = evaluate_script(function="""
+() => {
+    const cards = document.querySelectorAll('.lora-card');
+    const names = Array.from(cards).map(c => c.dataset.name || c.textContent);
+    return { count: cards.length, names };
+}
+""")
+```
+
+### Check API Response
+
+```python
+# Trigger API call
+evaluate_script(function="""
+() => window.loraApiCallPromise = fetch('/loras/api/list').then(r => r.json())
+""")
+
+# Wait and get result
+import time
+time.sleep(1)
+
+result = evaluate_script(function="""
+async () => await window.loraApiCallPromise
+""")
+```
+
+### Monitor Console for Errors
+
+```python
+# Before test: clear console (navigate reloads)
+navigate_page(type="reload")
+
+# ... perform actions ...
+
+# Check for errors
+errors = list_console_messages(types=["error"])
+assert len(errors) == 0, f"Console errors: {errors}"
+```

.agents/skills/lora-manager-e2e/references/test-scenarios.md

@@ -0,0 +1,272 @@
+# LoRa Manager E2E Test Scenarios
+
+This document provides detailed test scenarios for end-to-end validation of LoRa Manager features.
+
+## Table of Contents
+
+1. [LoRA List Page](#lora-list-page)
+2. [Model Details](#model-details)
+3. [Recipes](#recipes)
+4. [Settings](#settings)
+5. [Import/Export](#importexport)
+
+---
+
+## LoRA List Page
+
+### Scenario: Page Load and Display
+
+**Objective**: Verify the LoRA list page loads correctly and displays models.
+
+**Steps**:
+1. Navigate to `http://127.0.0.1:8188/loras`
+2. Wait for page title "LoRAs" to appear
+3. Take snapshot to verify:
+   - Header with "LoRAs" title is visible
+   - Search/filter controls are present
+   - Grid/list view toggle exists
+   - LoRA cards are displayed (if models exist)
+   - Pagination controls (if applicable)
+
+**Expected Result**: Page loads without errors, UI elements are present.
+
+### Scenario: Search Functionality
+
+**Objective**: Verify search filters LoRA models correctly.
+
+**Steps**:
+1. Ensure at least one LoRA exists with known name (e.g., "test-character")
+2. Navigate to LoRA list page
+3. Enter search term in search box: "test"
+4. Press Enter or click search button
+5. Wait for results to update
+
+**Expected Result**: Only LoRAs matching search term are displayed.
+
+**Verification Script**:
+```python
+# After search, verify filtered results
+evaluate_script(function="""
+() => {
+  const cards = document.querySelectorAll('.lora-card');
+  const names = Array.from(cards).map(c => c.dataset.name);
+  return { count: cards.length, names };
+}
+""")
+```
+
+### Scenario: Filter by Tags
+
+**Objective**: Verify tag filtering works correctly.
+
+**Steps**:
+1. Navigate to LoRA list page
+2. Click on a tag (e.g., "character", "style")
+3. Wait for filtered results
+
+**Expected Result**: Only LoRAs with selected tag are displayed.
+
+### Scenario: View Mode Toggle
+
+**Objective**: Verify grid/list view toggle works.
+
+**Steps**:
+1. Navigate to LoRA list page
+2. Click list view button
+3. Verify list layout
+4. Click grid view button
+5. Verify grid layout
+
+**Expected Result**: View mode changes correctly, layout updates.
+
+---
+
+## Model Details
+
+### Scenario: Open Model Details
+
+**Objective**: Verify clicking a LoRA opens its details.
+
+**Steps**:
+1. Navigate to LoRA list page
+2. Click on a LoRA card
+3. Wait for details panel/modal to open
+
+**Expected Result**: Details panel shows:
+- Model name
+- Preview image
+- Metadata (trigger words, tags, etc.)
+- Action buttons (edit, delete, etc.)
+
+### Scenario: Edit Model Metadata
+
+**Objective**: Verify metadata editing works end-to-end.
+
+**Steps**:
+1. Open a LoRA's details
+2. Click "Edit" button
+3. Modify trigger words field
+4. Add/remove tags
+5. Save changes
+6. Refresh page
+7. Reopen the same LoRA
+
+**Expected Result**: Changes persist after refresh.
+
+### Scenario: Delete Model
+
+**Objective**: Verify model deletion works.
+
+**Steps**:
+1. Open a LoRA's details
+2. Click "Delete" button
+3. Confirm deletion in dialog
+4. Wait for removal
+
+**Expected Result**: Model removed from list, success message shown.
+
+---
+
+## Recipes
+
+### Scenario: Recipe List Display
+
+**Objective**: Verify recipes page loads and displays recipes.
+
+**Steps**:
+1. Navigate to `http://127.0.0.1:8188/recipes`
+2. Wait for "Recipes" title
+3. Take snapshot
+
+**Expected Result**: Recipe list displayed with cards/items.
+
+### Scenario: Create New Recipe
+
+**Objective**: Verify recipe creation workflow.
+
+**Steps**:
+1. Navigate to recipes page
+2. Click "New Recipe" button
+3. Fill recipe form:
+   - Name: "Test Recipe"
+   - Description: "E2E test recipe"
+   - Add LoRA models
+4. Save recipe
+5. Verify recipe appears in list
+
+**Expected Result**: New recipe created and displayed.
+
+### Scenario: Apply Recipe
+
+**Objective**: Verify applying a recipe to ComfyUI.
+
+**Steps**:
+1. Open a recipe
+2. Click "Apply" or "Load in ComfyUI"
+3. Verify action completes
+
+**Expected Result**: Recipe applied successfully.
+
+---
+
+## Settings
+
+### Scenario: Settings Page Load
+
+**Objective**: Verify settings page displays correctly.
+
+**Steps**:
+1. Navigate to `http://127.0.0.1:8188/settings`
+2. Wait for "Settings" title
+3. Take snapshot
+
+**Expected Result**: Settings form with various options displayed.
+
+### Scenario: Change Setting and Restart
+
+**Objective**: Verify settings persist after restart.
+
+**Steps**:
+1. Navigate to settings page
+2. Change a setting (e.g., default view mode)
+3. Save settings
+4. Restart server: `python scripts/start_server.py --restart --wait`
+5. Refresh browser page
+6. Navigate to settings
+
+**Expected Result**: Changed setting value persists.
+
+---
+
+## Import/Export
+
+### Scenario: Export Models List
+
+**Objective**: Verify export functionality.
+
+**Steps**:
+1. Navigate to LoRA list
+2. Click "Export" button
+3. Select format (JSON/CSV)
+4. Download file
+
+**Expected Result**: File downloaded with correct data.
+
+### Scenario: Import Models
+
+**Objective**: Verify import functionality.
+
+**Steps**:
+1. Prepare import file
+2. Navigate to import page
+3. Upload file
+4. Verify import results
+
+**Expected Result**: Models imported successfully, confirmation shown.
+
+---
+
+## API Integration Tests
+
+### Scenario: Verify API Endpoints
+
+**Objective**: Verify backend API responds correctly.
+
+**Test via browser console**:
+```javascript
+// List LoRAs
+fetch('/loras/api/list').then(r => r.json()).then(console.log)
+
+// Get LoRA details
+fetch('/loras/api/detail/<id>').then(r => r.json()).then(console.log)
+
+// Search LoRAs
+fetch('/loras/api/search?q=test').then(r => r.json()).then(console.log)
+```
+
+**Expected Result**: APIs return valid JSON with expected structure.
+
+---
+
+## Console Error Monitoring
+
+During all tests, monitor browser console for errors:
+
+```python
+# Check for JavaScript errors
+messages = list_console_messages(types=["error"])
+assert len(messages) == 0, f"Console errors found: {messages}"
+```
+
+## Network Request Verification
+
+Verify key API calls are made:
+
+```python
+# List XHR requests
+requests = list_network_requests(resourceTypes=["xhr", "fetch"])
+
+# Look for specific endpoints
+lora_list_requests = [r for r in requests if "/api/list" in r.get("url", "")]
+assert len(lora_list_requests) > 0, "LoRA list API not called"
+```

.agents/skills/lora-manager-e2e/scripts/example_e2e_test.py

@@ -0,0 +1,193 @@
+#!/usr/bin/env python3
+"""
+Example E2E test demonstrating LoRa Manager testing workflow.
+
+This script shows how to:
+1. Start the standalone server
+2. Use Chrome DevTools MCP to interact with the UI
+3. Verify functionality end-to-end
+
+Note: This is a template. Actual execution requires Chrome DevTools MCP.
+"""
+
+import subprocess
+import sys
+import time
+
+
+def run_test():
+    """Run example E2E test flow."""
+    
+    print("=" * 60)
+    print("LoRa Manager E2E Test Example")
+    print("=" * 60)
+    
+    # Step 1: Start server
+    print("\n[1/5] Starting LoRa Manager standalone server...")
+    result = subprocess.run(
+        [sys.executable, "start_server.py", "--port", "8188", "--wait", "--timeout", "30"],
+        capture_output=True,
+        text=True
+    )
+    if result.returncode != 0:
+        print(f"Failed to start server: {result.stderr}")
+        return 1
+    print("Server ready!")
+    
+    # Step 2: Open Chrome (manual step - show command)
+    print("\n[2/5] Open Chrome with debug mode:")
+    print("google-chrome --remote-debugging-port=9222 --user-data-dir=/tmp/chrome-lora-manager http://127.0.0.1:8188/loras")
+    print("(In actual test, this would be automated via MCP)")
+    
+    # Step 3: Navigate and verify page load
+    print("\n[3/5] Page Load Verification:")
+    print("""
+    MCP Commands to execute:
+    1. navigate_page(type="url", url="http://127.0.0.1:8188/loras")
+    2. wait_for(text="LoRAs", timeout=10000)
+    3. snapshot = take_snapshot()
+    """)
+    
+    # Step 4: Test search functionality
+    print("\n[4/5] Search Functionality Test:")
+    print("""
+    MCP Commands to execute:
+    1. fill(uid="search-input", value="test")
+    2. press_key(key="Enter")
+    3. wait_for(text="Results", timeout=5000)
+    4. result = evaluate_script(function="""
+        () => {
+            const cards = document.querySelectorAll('.lora-card');
+            return { count: cards.length };
+        }
+        """)
+    """)
+    
+    # Step 5: Verify API
+    print("\n[5/5] API Verification:")
+    print("""
+    MCP Commands to execute:
+    1. api_result = evaluate_script(function="""
+        async () => {
+            const response = await fetch('/loras/api/list');
+            const data = await response.json();
+            return { count: data.length, status: response.status };
+        }
+        """)
+    2. Verify api_result['status'] == 200
+    """)
+    
+    print("\n" + "=" * 60)
+    print("Test flow completed!")
+    print("=" * 60)
+    
+    return 0
+
+
+def example_restart_flow():
+    """Example: Testing configuration change that requires restart."""
+    
+    print("\n" + "=" * 60)
+    print("Example: Server Restart Flow")
+    print("=" * 60)
+    
+    print("""
+    Scenario: Change setting and verify after restart
+    
+    Steps:
+    1. Navigate to settings page
+       - navigate_page(type="url", url="http://127.0.0.1:8188/settings")
+    
+    2. Change a setting (e.g., theme)
+       - fill(uid="theme-select", value="dark")
+       - click(uid="save-settings-button")
+    
+    3. Restart server
+       - subprocess.run([python, "start_server.py", "--restart", "--wait"])
+    
+    4. Refresh browser
+       - navigate_page(type="reload", ignoreCache=True)
+       - wait_for(text="LoRAs", timeout=15000)
+    
+    5. Verify setting persisted
+       - navigate_page(type="url", url="http://127.0.0.1:8188/settings")
+       - theme = evaluate_script(function="() => document.querySelector('#theme-select').value")
+       - assert theme == "dark"
+    """)
+
+
+def example_modal_interaction():
+    """Example: Testing modal dialog interaction."""
+    
+    print("\n" + "=" * 60)
+    print("Example: Modal Dialog Interaction")
+    print("=" * 60)
+    
+    print("""
+    Scenario: Add new LoRA via modal
+    
+    Steps:
+    1. Open modal
+       - click(uid="add-lora-button")
+       - wait_for(text="Add LoRA", timeout=3000)
+    
+    2. Fill form
+       - fill_form(elements=[
+           {"uid": "lora-name", "value": "Test Character"},
+           {"uid": "lora-path", "value": "/models/test.safetensors"},
+       ])
+    
+    3. Submit
+       - click(uid="modal-submit-button")
+    
+    4. Verify success
+       - wait_for(text="Successfully added", timeout=5000)
+       - snapshot = take_snapshot()
+    """)
+
+
+def example_network_monitoring():
+    """Example: Network request monitoring."""
+    
+    print("\n" + "=" * 60)
+    print("Example: Network Request Monitoring")
+    print("=" * 60)
+    
+    print("""
+    Scenario: Verify API calls during user interaction
+    
+    Steps:
+    1. Clear network log (implicit on navigation)
+       - navigate_page(type="url", url="http://127.0.0.1:8188/loras")
+    
+    2. Perform action that triggers API call
+       - fill(uid="search-input", value="character")
+       - press_key(key="Enter")
+    
+    3. List network requests
+       - requests = list_network_requests(resourceTypes=["xhr", "fetch"])
+    
+    4. Find search API call
+       - search_requests = [r for r in requests if "/api/search" in r.get("url", "")]
+       - assert len(search_requests) > 0, "Search API was not called"
+    
+    5. Get request details
+       - if search_requests:
+           details = get_network_request(reqid=search_requests[0]["reqid"])
+           - Verify request method, response status, etc.
+    """)
+
+
+if __name__ == "__main__":
+    print("LoRa Manager E2E Test Examples\n")
+    print("This script demonstrates E2E testing patterns.\n")
+    print("Note: Actual execution requires Chrome DevTools MCP connection.\n")
+    
+    run_test()
+    example_restart_flow()
+    example_modal_interaction()
+    example_network_monitoring()
+    
+    print("\n" + "=" * 60)
+    print("All examples shown!")
+    print("=" * 60)

.agents/skills/lora-manager-e2e/scripts/start_server.py

@@ -0,0 +1,169 @@
+#!/usr/bin/env python3
+"""
+Start or restart LoRa Manager standalone server for E2E testing.
+"""
+
+import argparse
+import subprocess
+import sys
+import time
+import socket
+import signal
+import os
+
+
+def find_server_process(port: int) -> list[int]:
+    """Find PIDs of processes listening on the given port."""
+    try:
+        result = subprocess.run(
+            ["lsof", "-ti", f":{port}"],
+            capture_output=True,
+            text=True,
+            check=False
+        )
+        if result.returncode == 0 and result.stdout.strip():
+            return [int(pid) for pid in result.stdout.strip().split("\n") if pid]
+    except FileNotFoundError:
+        # lsof not available, try netstat
+        try:
+            result = subprocess.run(
+                ["netstat", "-tlnp"],
+                capture_output=True,
+                text=True,
+                check=False
+            )
+            pids = []
+            for line in result.stdout.split("\n"):
+                if f":{port}" in line:
+                    parts = line.split()
+                    for part in parts:
+                        if "/" in part:
+                            try:
+                                pid = int(part.split("/")[0])
+                                pids.append(pid)
+                            except ValueError:
+                                pass
+            return pids
+        except FileNotFoundError:
+            pass
+    return []
+
+
+def kill_server(port: int) -> None:
+    """Kill processes using the specified port."""
+    pids = find_server_process(port)
+    for pid in pids:
+        try:
+            os.kill(pid, signal.SIGTERM)
+            print(f"Sent SIGTERM to process {pid}")
+        except ProcessLookupError:
+            pass
+    
+    # Wait for processes to terminate
+    time.sleep(1)
+    
+    # Force kill if still running
+    pids = find_server_process(port)
+    for pid in pids:
+        try:
+            os.kill(pid, signal.SIGKILL)
+            print(f"Sent SIGKILL to process {pid}")
+        except ProcessLookupError:
+            pass
+
+
+def is_server_ready(port: int, timeout: float = 0.5) -> bool:
+    """Check if server is accepting connections."""
+    try:
+        with socket.create_connection(("127.0.0.1", port), timeout=timeout):
+            return True
+    except (socket.timeout, ConnectionRefusedError, OSError):
+        return False
+
+
+def wait_for_server(port: int, timeout: int = 30) -> bool:
+    """Wait for server to become ready."""
+    start = time.time()
+    while time.time() - start < timeout:
+        if is_server_ready(port):
+            return True
+        time.sleep(0.5)
+    return False
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="Start LoRa Manager standalone server for E2E testing"
+    )
+    parser.add_argument(
+        "--port",
+        type=int,
+        default=8188,
+        help="Server port (default: 8188)"
+    )
+    parser.add_argument(
+        "--restart",
+        action="store_true",
+        help="Kill existing server before starting"
+    )
+    parser.add_argument(
+        "--wait",
+        action="store_true",
+        help="Wait for server to be ready before exiting"
+    )
+    parser.add_argument(
+        "--timeout",
+        type=int,
+        default=30,
+        help="Timeout for waiting (default: 30)"
+    )
+    
+    args = parser.parse_args()
+    
+    # Get project root (parent of .agents directory)
+    script_dir = os.path.dirname(os.path.abspath(__file__))
+    skill_dir = os.path.dirname(script_dir)
+    project_root = os.path.dirname(os.path.dirname(os.path.dirname(skill_dir)))
+    
+    # Restart if requested
+    if args.restart:
+        print(f"Killing existing server on port {args.port}...")
+        kill_server(args.port)
+        time.sleep(1)
+    
+    # Check if already running
+    if is_server_ready(args.port):
+        print(f"Server already running on port {args.port}")
+        return 0
+    
+    # Start server
+    print(f"Starting LoRa Manager standalone server on port {args.port}...")
+    cmd = [sys.executable, "standalone.py", "--port", str(args.port)]
+    
+    # Start in background
+    process = subprocess.Popen(
+        cmd,
+        cwd=project_root,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.PIPE,
+        start_new_session=True
+    )
+    
+    print(f"Server process started with PID {process.pid}")
+    
+    # Wait for ready if requested
+    if args.wait:
+        print(f"Waiting for server to be ready (timeout: {args.timeout}s)...")
+        if wait_for_server(args.port, args.timeout):
+            print(f"Server ready at http://127.0.0.1:{args.port}/loras")
+            return 0
+        else:
+            print(f"Timeout waiting for server")
+            return 1
+    
+    print(f"Server starting at http://127.0.0.1:{args.port}/loras")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

.agents/skills/lora-manager-e2e/scripts/wait_for_server.py

@@ -0,0 +1,61 @@
+#!/usr/bin/env python3
+"""
+Wait for LoRa Manager server to become ready.
+"""
+
+import argparse
+import socket
+import sys
+import time
+
+
+def is_server_ready(port: int, timeout: float = 0.5) -> bool:
+    """Check if server is accepting connections."""
+    try:
+        with socket.create_connection(("127.0.0.1", port), timeout=timeout):
+            return True
+    except (socket.timeout, ConnectionRefusedError, OSError):
+        return False
+
+
+def wait_for_server(port: int, timeout: int = 30) -> bool:
+    """Wait for server to become ready."""
+    start = time.time()
+    while time.time() - start < timeout:
+        if is_server_ready(port):
+            return True
+        time.sleep(0.5)
+    return False
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="Wait for LoRa Manager server to become ready"
+    )
+    parser.add_argument(
+        "--port",
+        type=int,
+        default=8188,
+        help="Server port (default: 8188)"
+    )
+    parser.add_argument(
+        "--timeout",
+        type=int,
+        default=30,
+        help="Timeout in seconds (default: 30)"
+    )
+    
+    args = parser.parse_args()
+    
+    print(f"Waiting for server on port {args.port} (timeout: {args.timeout}s)...")
+    
+    if wait_for_server(args.port, args.timeout):
+        print(f"Server ready at http://127.0.0.1:{args.port}/loras")
+        return 0
+    else:
+        print(f"Timeout: Server not ready after {args.timeout}s")
+        return 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())

.github/FUNDING.yml

@@ -0,0 +1,5 @@
+# These are supported funding model platforms
+
+ko_fi: pixelpawsai
+patreon: PixelPawsAI
+custom: ['paypal.me/pixelpawsai', 'https://afdian.com/a/pixelpawsai']

.github/copilot-instructions.md

@@ -0,0 +1 @@
+Always use English for comments.

.github/workflows/backend-tests.yml

@@ -0,0 +1,93 @@
+name: Backend Tests
+
+on:
+  push:
+    branches:
+      - main
+      - master
+    paths:
+      - 'py/**'
+      - 'standalone.py'
+      - 'tests/**'
+      - 'requirements.txt'
+      - 'requirements-dev.txt'
+      - 'pyproject.toml'
+      - 'pytest.ini'
+      - '.github/workflows/backend-tests.yml'
+  pull_request:
+    paths:
+      - 'py/**'
+      - 'standalone.py'
+      - 'tests/**'
+      - 'requirements.txt'
+      - 'requirements-dev.txt'
+      - 'pyproject.toml'
+      - 'pytest.ini'
+      - '.github/workflows/backend-tests.yml'
+
+jobs:
+  pytest:
+    name: Run pytest with coverage
+    runs-on: ubuntu-latest
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@v4
+
+      - name: Set up Python 3.11
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+          cache: 'pip'
+          cache-dependency-path: |
+            requirements.txt
+            requirements-dev.txt
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements-dev.txt
+
+      - name: Verify symlink support
+        run: |
+          python - <<'PY'
+          import os
+          import pathlib
+          import tempfile
+
+          root = pathlib.Path(tempfile.mkdtemp(prefix="lm-symlink-check-"))
+          target = root / "target"
+          target.mkdir()
+          link = root / "link"
+          try:
+              link.symlink_to(target, target_is_directory=True)
+          except OSError as exc:
+              raise SystemExit(f"Failed to create directory symlink in CI: {exc}")
+
+          is_link = os.path.islink(link)
+          is_dir = os.path.isdir(link)
+          realpath = os.path.realpath(link)
+          print(f"islink={is_link} isdir={is_dir} realpath={realpath}")
+          if not (is_link and is_dir and realpath == str(target)):
+              raise SystemExit("Directory symlink is not functioning correctly in CI; aborting.")
+          PY
+
+      - name: Run pytest with coverage
+        env:
+          COVERAGE_FILE: coverage/backend/.coverage
+        run: |
+          mkdir -p coverage/backend
+          python -m pytest \
+            --cov=py \
+            --cov=standalone \
+            --cov-report=term-missing \
+            --cov-report=xml:coverage/backend/coverage.xml \
+            --cov-report=html:coverage/backend/html \
+            --cov-report=json:coverage/backend/coverage.json
+
+      - name: Upload coverage artifact
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: backend-coverage
+          path: coverage/backend
+          if-no-files-found: warn

.github/workflows/frontend-tests.yml

@@ -0,0 +1,52 @@
+name: Frontend Tests
+
+on:
+  push:
+    branches:
+      - main
+      - master
+    paths:
+      - 'package.json'
+      - 'package-lock.json'
+      - 'vitest.config.js'
+      - 'tests/frontend/**'
+      - 'static/js/**'
+      - 'scripts/run_frontend_coverage.js'
+      - '.github/workflows/frontend-tests.yml'
+  pull_request:
+    paths:
+      - 'package.json'
+      - 'package-lock.json'
+      - 'vitest.config.js'
+      - 'tests/frontend/**'
+      - 'static/js/**'
+      - 'scripts/run_frontend_coverage.js'
+      - '.github/workflows/frontend-tests.yml'
+
+jobs:
+  vitest:
+    name: Run Vitest with coverage
+    runs-on: ubuntu-latest
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@v4
+
+      - name: Use Node.js 20
+        uses: actions/setup-node@v4
+        with:
+          node-version: 20
+          cache: 'npm'
+
+      - name: Install dependencies
+        run: npm ci
+
+      - name: Run frontend tests with coverage
+        run: npm run test:coverage
+
+      - name: Upload coverage artifact
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: frontend-coverage
+          path: coverage/frontend
+          if-no-files-found: warn

.gitignore

@@ -1,2 +1,21 @@
 __pycache__/
-settings.json
+.pytest_cache/
+settings.json
+path_mappings.yaml
+output/*
+py/run_test.py
+.vscode/
+cache/
+civitai/
+node_modules/
+coverage/
+.coverage
+model_cache/
+
+# agent
+.opencode/
+
+# Vue widgets development cache (but keep build output)
+vue-widgets/node_modules/
+vue-widgets/.vite/
+vue-widgets/dist/

AGENTS.md

@@ -0,0 +1,192 @@
+# AGENTS.md
+
+This file provides guidance for agentic coding assistants working in this repository.
+
+## Development Commands
+
+### Backend Development
+
+```bash
+# Install dependencies
+pip install -r requirements.txt
+pip install -r requirements-dev.txt
+
+# Run standalone server (port 8188 by default)
+python standalone.py --port 8188
+
+# Run all backend tests
+pytest
+
+# Run specific test file
+pytest tests/test_recipes.py
+
+# Run specific test function
+pytest tests/test_recipes.py::test_function_name
+
+# Run backend tests with coverage
+COVERAGE_FILE=coverage/backend/.coverage pytest \
+  --cov=py \
+  --cov=standalone \
+  --cov-report=term-missing \
+  --cov-report=html:coverage/backend/html \
+  --cov-report=xml:coverage/backend/coverage.xml \
+  --cov-report=json:coverage/backend/coverage.json
+```
+
+### Frontend Development
+
+```bash
+# Install frontend dependencies
+npm install
+
+# Run frontend tests
+npm test
+
+# Run frontend tests in watch mode
+npm run test:watch
+
+# Run frontend tests with coverage
+npm run test:coverage
+```
+
+## Python Code Style
+
+### Imports
+
+- Use `from __future__ import annotations` for forward references in type hints
+- Group imports: standard library, third-party, local (separated by blank lines)
+- Use absolute imports within `py/` package: `from ..services import X`
+- Mock ComfyUI dependencies in tests using `tests/conftest.py` patterns
+
+### Formatting & Types
+
+- PEP 8 with 4-space indentation
+- Type hints required for function signatures and class attributes
+- Use `TYPE_CHECKING` guard for type-checking-only imports
+- Prefer dataclasses for simple data containers
+- Use `Optional[T]` for nullable types, `Union[T, None]` only when necessary
+
+### Naming Conventions
+
+- Files: `snake_case.py` (e.g., `model_scanner.py`, `lora_service.py`)
+- Classes: `PascalCase` (e.g., `ModelScanner`, `LoraService`)
+- Functions/variables: `snake_case` (e.g., `get_instance`, `model_type`)
+- Constants: `UPPER_SNAKE_CASE` (e.g., `VALID_LORA_TYPES`)
+- Private members: `_single_underscore` (protected), `__double_underscore` (name-mangled)
+
+### Error Handling
+
+- Use `logging.getLogger(__name__)` for module-level loggers
+- Define custom exceptions in `py/services/errors.py`
+- Use `asyncio.Lock` for thread-safe singleton patterns
+- Raise specific exceptions with descriptive messages
+- Log errors at appropriate levels (DEBUG, INFO, WARNING, ERROR, CRITICAL)
+
+### Async Patterns
+
+- Use `async def` for I/O-bound operations
+- Mark async tests with `@pytest.mark.asyncio`
+- Use `async with` for context managers
+- Singleton pattern with class-level locks: see `ModelScanner.get_instance()`
+- Use `aiohttp.web.Response` for HTTP responses
+
+### Testing Patterns
+
+- Use `pytest` with `--import-mode=importlib`
+- Fixtures in `tests/conftest.py` handle ComfyUI mocking
+- Use `@pytest.mark.no_settings_dir_isolation` for tests needing real paths
+- Test files: `tests/test_*.py`
+- Use `tmp_path_factory` for temporary directory isolation
+
+## JavaScript Code Style
+
+### Imports & Modules
+
+- ES modules with `import`/`export`
+- Use `import { app } from "../../scripts/app.js"` for ComfyUI integration
+- Export named functions/classes: `export function foo() {}`
+- Widget files use `*_widget.js` suffix
+
+### Naming & Formatting
+
+- camelCase for functions, variables, object properties
+- PascalCase for classes/constructors
+- Constants: `UPPER_SNAKE_CASE` (e.g., `CONVERTED_TYPE`)
+- Files: `snake_case.js` or `kebab-case.js`
+- 2-space indentation preferred (follow existing file conventions)
+
+### Widget Development
+
+- Use `app.registerExtension()` to register ComfyUI extensions
+- Use `node.addDOMWidget(name, type, element, options)` for custom widgets
+- Event handlers attached via `addEventListener` or widget callbacks
+- See `web/comfyui/utils.js` for shared utilities
+
+## Architecture Patterns
+
+### Service Layer
+
+- Use `ServiceRegistry` singleton for dependency injection
+- Services follow singleton pattern via `get_instance()` class method
+- Separate scanners (discovery) from services (business logic)
+- Handlers in `py/routes/handlers/` implement route logic
+
+### Model Types
+
+- BaseModelService is abstract base for LoRA, Checkpoint, Embedding services
+- ModelScanner provides file discovery and hash-based deduplication
+- Persistent cache in SQLite via `PersistentModelCache`
+- Metadata sync from CivitAI/CivArchive via `MetadataSyncService`
+
+### Routes & Handlers
+
+- Route registrars organize endpoints by domain: `ModelRouteRegistrar`, etc.
+- Handlers are pure functions taking dependencies as parameters
+- Use `WebSocketManager` for real-time progress updates
+- Return `aiohttp.web.json_response` or `web.Response`
+
+### Recipe System
+
+- Base metadata in `py/recipes/base.py`
+- Enrichment adds model metadata: `RecipeEnrichmentService`
+- Parsers for different formats in `py/recipes/parsers/`
+
+## Important Notes
+
+- Always use English for comments (per copilot-instructions.md)
+- Dual mode: ComfyUI plugin (uses folder_paths) vs standalone (reads settings.json)
+- Detection: `os.environ.get("LORA_MANAGER_STANDALONE", "0") == "1"`
+- Settings auto-saved in user directory or portable mode
+- WebSocket broadcasts for real-time updates (downloads, scans)
+- Symlink handling requires normalized paths
+- API endpoints follow `/loras/*`, `/checkpoints/*`, `/embeddings/*` patterns
+- Run `python scripts/sync_translation_keys.py` after UI string updates
+
+## Frontend UI Architecture
+
+This project has two distinct UI systems:
+
+### 1. Standalone Lora Manager Web UI
+- Location: `./static/` and `./templates/`
+- Purpose: Full-featured web application for managing LoRA models
+- Tech stack: Vanilla JS + CSS, served by the standalone server
+- Development: Uses npm for frontend testing (`npm test`, `npm run test:watch`, etc.)
+
+### 2. ComfyUI Custom Node Widgets
+- Location: `./web/comfyui/`
+- Purpose: Widgets and UI logic that ComfyUI loads as custom node extensions
+- Tech stack: Vanilla JS + Vue.js widgets (in `./vue-widgets/` and built to `./web/comfyui/vue-widgets/`)
+- Widget styling: Primary styles in `./web/comfyui/lm_styles.css` (NOT `./static/css/`)
+- Development: No npm build step for these widgets (Vue widgets use build system)
+
+### Widget Development Guidelines
+- Use `app.registerExtension()` to register ComfyUI extensions (ComfyUI integration layer)
+- Use `node.addDOMWidget()` for custom DOM widgets
+- Widget styles should follow the patterns in `./web/comfyui/lm_styles.css`
+  - Selected state: `rgba(66, 153, 225, 0.3)` background, `rgba(66, 153, 225, 0.6)` border
+  - Hover state: `rgba(66, 153, 225, 0.2)` background
+  - Color palette matches the Lora Manager accent color (blue #4299e1)
+  - Use oklch() for color values when possible (defined in `./static/css/base.css`)
+- Vue widget components are in `./vue-widgets/src/components/` and built to `./web/comfyui/vue-widgets/`
+- When modifying widget styles, check `./web/comfyui/lm_styles.css` for consistency with other ComfyUI widgets
+

CLAUDE.md

@@ -0,0 +1,211 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Overview
+
+ComfyUI LoRA Manager is a comprehensive LoRA management system for ComfyUI that combines a Python backend with browser-based widgets. It provides model organization, downloading from CivitAI/CivArchive, recipe management, and one-click workflow integration.
+
+## Development Commands
+
+### Backend Development
+```bash
+# Install dependencies
+pip install -r requirements.txt
+
+# Install development dependencies (for testing)
+pip install -r requirements-dev.txt
+
+# Run standalone server (port 8188 by default)
+python standalone.py --port 8188
+
+# Run backend tests with coverage
+COVERAGE_FILE=coverage/backend/.coverage pytest \
+  --cov=py \
+  --cov=standalone \
+  --cov-report=term-missing \
+  --cov-report=html:coverage/backend/html \
+  --cov-report=xml:coverage/backend/coverage.xml \
+  --cov-report=json:coverage/backend/coverage.json
+
+# Run specific test file
+pytest tests/test_recipes.py
+```
+
+### Frontend Development
+```bash
+# Install frontend dependencies
+npm install
+
+# Run frontend tests
+npm test
+
+# Run frontend tests in watch mode
+npm run test:watch
+
+# Run frontend tests with coverage
+npm run test:coverage
+```
+
+### Localization
+```bash
+# Sync translation keys after UI string updates
+python scripts/sync_translation_keys.py
+```
+
+## Architecture
+
+### Backend Structure (Python)
+
+**Core Entry Points:**
+- `__init__.py` - ComfyUI plugin entry point, registers nodes and routes
+- `standalone.py` - Standalone server that mocks ComfyUI dependencies
+- `py/lora_manager.py` - Main LoraManager class that registers HTTP routes
+
+**Service Layer** (`py/services/`):
+- `ServiceRegistry` - Singleton service registry for dependency management
+- `ModelServiceFactory` - Factory for creating model services (LoRA, Checkpoint, Embedding)
+- Scanner services (`lora_scanner.py`, `checkpoint_scanner.py`, `embedding_scanner.py`) - Model file discovery and indexing
+- `model_scanner.py` - Base scanner with hash-based deduplication and metadata extraction
+- `persistent_model_cache.py` - SQLite-based cache for model metadata
+- `metadata_sync_service.py` - Syncs metadata from CivitAI/CivArchive APIs
+- `civitai_client.py` / `civarchive_client.py` - API clients for external services
+- `downloader.py` / `download_manager.py` - Model download orchestration
+- `recipe_scanner.py` - Recipe file management and image association
+- `settings_manager.py` - Application settings with migration support
+- `websocket_manager.py` - WebSocket broadcasting for real-time updates
+- `use_cases/` - Business logic orchestration (auto-organize, bulk refresh, downloads)
+
+**Routes Layer** (`py/routes/`):
+- Route registrars organize endpoints by domain (models, recipes, previews, example images, updates)
+- `handlers/` - Request handlers implementing business logic
+- Routes use aiohttp and integrate with ComfyUI's PromptServer
+
+**Recipe System** (`py/recipes/`):
+- `base.py` - Base recipe metadata structure
+- `enrichment.py` - Enriches recipes with model metadata
+- `merger.py` - Merges recipe data from multiple sources
+- `parsers/` - Parsers for different recipe formats (PNG, JSON, workflow)
+
+**Custom Nodes** (`py/nodes/`):
+- `lora_loader.py` - LoRA loader nodes with preset support
+- `save_image.py` - Enhanced save image with pattern-based filenames
+- `trigger_word_toggle.py` - Toggle trigger words in prompts
+- `lora_stacker.py` - Stack multiple LoRAs
+- `prompt.py` - Prompt node with autocomplete
+- `wanvideo_lora_select.py` - WanVideo-specific LoRA selection
+
+**Configuration** (`py/config.py`):
+- Manages folder paths for models, checkpoints, embeddings
+- Handles symlink mappings for complex directory structures
+- Auto-saves paths to settings.json in ComfyUI mode
+
+### Frontend Structure (JavaScript)
+
+**ComfyUI Widgets** (`web/comfyui/`):
+- Vanilla JavaScript ES modules extending ComfyUI's LiteGraph-based UI
+- `loras_widget.js` - Main LoRA selection widget with preview
+- `loras_widget_events.js` - Event handling for widget interactions
+- `autocomplete.js` - Autocomplete for trigger words and embeddings
+- `preview_tooltip.js` - Preview tooltip for model cards
+- `top_menu_extension.js` - Adds "Launch LoRA Manager" menu item
+- `trigger_word_highlight.js` - Syntax highlighting for trigger words
+- `utils.js` - Shared utilities and API helpers
+
+**Widget Development:**
+- Widgets use `app.registerExtension` and `getCustomWidgets` hooks
+- `node.addDOMWidget(name, type, element, options)` embeds HTML in nodes
+- See `docs/dom_widget_dev_guide.md` for complete DOMWidget development guide
+
+**Web Source** (`web-src/`):
+- Modern frontend components (if migrating from static)
+- `components/` - Reusable UI components
+- `styles/` - CSS styling
+
+### Key Patterns
+
+**Dual Mode Operation:**
+- ComfyUI plugin mode: Integrates with ComfyUI's PromptServer, uses folder_paths
+- Standalone mode: Mocks ComfyUI dependencies via `standalone.py`, reads paths from settings.json
+- Detection: `os.environ.get("LORA_MANAGER_STANDALONE", "0") == "1"`
+
+**Settings Management:**
+- Settings stored in user directory (via `platformdirs`) or portable mode (in repo)
+- Migration system tracks settings schema version
+- Template in `settings.json.example` defines defaults
+
+**Model Scanning Flow:**
+1. Scanner walks folder paths, computes file hashes
+2. Hash-based deduplication prevents duplicate processing
+3. Metadata extracted from safetensors headers
+4. Persistent cache stores results in SQLite
+5. Background sync fetches CivitAI/CivArchive metadata
+6. WebSocket broadcasts updates to connected clients
+
+**Recipe System:**
+- Recipes store LoRA combinations with parameters
+- Supports import from workflow JSON, PNG metadata
+- Images associated with recipes via sibling file detection
+- Enrichment adds model metadata for display
+
+**Frontend-Backend Communication:**
+- REST API for CRUD operations
+- WebSocket for real-time progress updates (downloads, scans)
+- API endpoints follow `/loras/*` pattern
+
+## Code Style
+
+**Python:**
+- PEP 8 with 4-space indentation
+- snake_case for files, functions, variables
+- PascalCase for classes
+- Type hints preferred
+- English comments only (per copilot-instructions.md)
+- Loggers via `logging.getLogger(__name__)`
+
+**JavaScript:**
+- ES modules with camelCase
+- Files use `*_widget.js` suffix for ComfyUI widgets
+- Prefer vanilla JS, avoid framework dependencies
+
+## Testing
+
+**Backend Tests:**
+- pytest with `--import-mode=importlib`
+- Test files: `tests/test_*.py`
+- Fixtures in `tests/conftest.py`
+- Mock ComfyUI dependencies using standalone.py patterns
+- Markers: `@pytest.mark.asyncio` for async tests, `@pytest.mark.no_settings_dir_isolation` for real paths
+
+**Frontend Tests:**
+- Vitest with jsdom environment
+- Test files: `tests/frontend/**/*.test.js`
+- Setup in `tests/frontend/setup.js`
+- Coverage via `npm run test:coverage`
+
+## Important Notes
+
+**Settings Location:**
+- ComfyUI mode: Auto-saves folder paths to user settings directory
+- Standalone mode: Use `settings.json` (copy from `settings.json.example`)
+- Portable mode: Set `"use_portable_settings": true` in settings.json
+
+**API Integration:**
+- CivitAI API key required for downloads (add to settings)
+- CivArchive API used as fallback for deleted models
+- Metadata archive database available for offline metadata
+
+**Symlink Handling:**
+- Config scans symlinks to map virtual paths to physical locations
+- Preview validation uses normalized preview root paths
+- Fingerprinting prevents redundant symlink rescans
+
+**ComfyUI Node Development:**
+- Nodes defined in `py/nodes/`, registered in `__init__.py`
+- Frontend widgets in `web/comfyui/`, matched by node type
+- Use `WEB_DIRECTORY = "./web/comfyui"` convention
+
+**Recipe Image Association:**
+- Recipes scan for sibling images in same directory
+- Supports repair/migration of recipe image paths
+- See `py/services/recipe_scanner.py` for implementation details

LICENSE

@@ -1,21 +1,674 @@
-MIT License
+GNU GENERAL PUBLIC LICENSE
+                       Version 3, 29 June 2007
 
-Copyright (c) 2023 Will Miao
+ Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
 
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
+                            Preamble
 
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
+  The GNU General Public License is a free, copyleft license for
+software and other kinds of works.
 
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE. 
+  The licenses for most software and other practical works are designed
+to take away your freedom to share and change the works.  By contrast,
+the GNU General Public License is intended to guarantee your freedom to
+share and change all versions of a program--to make sure it remains free
+software for all its users.  We, the Free Software Foundation, use the
+GNU General Public License for most of our software; it applies also to
+any other work released this way by its authors.  You can apply it to
+your programs, too.
+
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+them if you wish), that you receive source code or can get it if you
+want it, that you can change the software or use pieces of it in new
+free programs, and that you know you can do these things.
+
+  To protect your rights, we need to prevent others from denying you
+these rights or asking you to surrender the rights.  Therefore, you have
+certain responsibilities if you distribute copies of the software, or if
+you modify it: responsibilities to respect the freedom of others.
+
+  For example, if you distribute copies of such a program, whether
+gratis or for a fee, you must pass on to the recipients the same
+freedoms that you received.  You must make sure that they, too, receive
+or can get the source code.  And you must show them these terms so they
+know their rights.
+
+  Developers that use the GNU GPL protect your rights with two steps:
+(1) assert copyright on the software, and (2) offer you this License
+giving you legal permission to copy, distribute and/or modify it.
+
+  For the developers' and authors' protection, the GPL clearly explains
+that there is no warranty for this free software.  For both users' and
+authors' sake, the GPL requires that modified versions be marked as
+changed, so that their problems will not be attributed erroneously to
+authors of previous versions.
+
+  Some devices are designed to deny users access to install or run
+modified versions of the software inside them, although the manufacturer
+can do so.  This is fundamentally incompatible with the aim of
+protecting users' freedom to change the software.  The systematic
+pattern of such abuse occurs in the area of products for individuals to
+use, which is precisely where it is most unacceptable.  Therefore, we
+have designed this version of the GPL to prohibit the practice for those
+products.  If such problems arise substantially in other domains, we
+stand ready to extend this provision to those domains in future versions
+of the GPL, as needed to protect the freedom of users.
+
+  Finally, every program is threatened constantly by software patents.
+States should not allow patents to restrict development and use of
+software on general-purpose computers, but in those that do, we wish to
+avoid the special danger that patents applied to a free program could
+make it effectively proprietary.  To prevent this, the GPL assures that
+patents cannot be used to render the program non-free.
+
+  The precise terms and conditions for copying, distribution and
+modification follow.
+
+                       TERMS AND CONDITIONS
+
+  0. Definitions.
+
+  "This License" refers to version 3 of the GNU General Public License.
+
+  "Copyright" also means copyright-like laws that apply to other kinds of
+works, such as semiconductor masks.
+
+  "The Program" refers to any copyrightable work licensed under this
+License.  Each licensee is addressed as "you".  "Licensees" and
+"recipients" may be individuals or organizations.
+
+  To "modify" a work means to copy from or adapt all or part of the work
+in a fashion requiring copyright permission, other than the making of an
+exact copy.  The resulting work is called a "modified version" of the
+earlier work or a work "based on" the earlier work.
+
+  A "covered work" means either the unmodified Program or a work based
+on the Program.
+
+  To "propagate" a work means to do anything with it that, without
+permission, would make you directly or secondarily liable for
+infringement under applicable copyright law, except executing it on a
+computer or modifying a private copy.  Propagation includes copying,
+distribution (with or without modification), making available to the
+public, and in some countries other activities as well.
+
+  To "convey" a work means any kind of propagation that enables other
+parties to make or receive copies.  Mere interaction with a user through
+a computer network, with no transfer of a copy, is not conveying.
+
+  An interactive user interface displays "Appropriate Legal Notices"
+to the extent that it includes a convenient and prominently visible
+feature that (1) displays an appropriate copyright notice, and (2)
+tells the user that there is no warranty for the work (except to the
+extent that warranties are provided), that licensees may convey the
+work under this License, and how to view a copy of this License.  If
+the interface presents a list of user commands or options, such as a
+menu, a prominent item in the list meets this criterion.
+
+  1. Source Code.
+
+  The "source code" for a work means the preferred form of the work
+for making modifications to it.  "Object code" means any non-source
+form of a work.
+
+  A "Standard Interface" means an interface that either is an official
+standard defined by a recognized standards body, or, in the case of
+interfaces specified for a particular programming language, one that
+is widely used among developers working in that language.
+
+  The "System Libraries" of an executable work include anything, other
+than the work as a whole, that (a) is included in the normal form of
+packaging a Major Component, but which is not part of that Major
+Component, and (b) serves only to enable use of the work with that
+Major Component, or to implement a Standard Interface for which an
+implementation is available to the public in source code form.  A
+"Major Component", in this context, means a major essential component
+(kernel, window system, and so on) of the specific operating system
+(if any) on which the executable work runs, or a compiler used to
+produce the work, or an object code interpreter used to run it.
+
+  The "Corresponding Source" for a work in object code form means all
+the source code needed to generate, install, and (for an executable
+work) run the object code and to modify the work, including scripts to
+control those activities.  However, it does not include the work's
+System Libraries, or general-purpose tools or generally available free
+programs which are used unmodified in performing those activities but
+which are not part of the work.  For example, Corresponding Source
+includes interface definition files associated with source files for
+the work, and the source code for shared libraries and dynamically
+linked subprograms that the work is specifically designed to require,
+such as by intimate data communication or control flow between those
+subprograms and other parts of the work.
+
+  The Corresponding Source need not include anything that users
+can regenerate automatically from other parts of the Corresponding
+Source.
+
+  The Corresponding Source for a work in source code form is that
+same work.
+
+  2. Basic Permissions.
+
+  All rights granted under this License are granted for the term of
+copyright on the Program, and are irrevocable provided the stated
+conditions are met.  This License explicitly affirms your unlimited
+permission to run the unmodified Program.  The output from running a
+covered work is covered by this License only if the output, given its
+content, constitutes a covered work.  This License acknowledges your
+rights of fair use or other equivalent, as provided by copyright law.
+
+  You may make, run and propagate covered works that you do not
+convey, without conditions so long as your license otherwise remains
+in force.  You may convey covered works to others for the sole purpose
+of having them make modifications exclusively for you, or provide you
+with facilities for running those works, provided that you comply with
+the terms of this License in conveying all material for which you do
+not control copyright.  Those thus making or running the covered works
+for you must do so exclusively on your behalf, under your direction
+and control, on terms that prohibit them from making any copies of
+your copyrighted material outside their relationship with you.
+
+  Conveying under any other circumstances is permitted solely under
+the conditions stated below.  Sublicensing is not allowed; section 10
+makes it unnecessary.
+
+  3. Protecting Users' Legal Rights From Anti-Circumvention Law.
+
+  No covered work shall be deemed part of an effective technological
+measure under any applicable law fulfilling obligations under article
+11 of the WIPO copyright treaty adopted on 20 December 1996, or
+similar laws prohibiting or restricting circumvention of such
+measures.
+
+  When you convey a covered work, you waive any legal power to forbid
+circumvention of technological measures to the extent such circumvention
+is effected by exercising rights under this License with respect to
+the covered work, and you disclaim any intention to limit operation or
+modification of the work as a means of enforcing, against the work's
+users, your or third parties' legal rights to forbid circumvention of
+technological measures.
+
+  4. Conveying Verbatim Copies.
+
+  You may convey verbatim copies of the Program's source code as you
+receive it, in any medium, provided that you conspicuously and
+appropriately publish on each copy an appropriate copyright notice;
+keep intact all notices stating that this License and any
+non-permissive terms added in accord with section 7 apply to the code;
+keep intact all notices of the absence of any warranty; and give all
+recipients a copy of this License along with the Program.
+
+  You may charge any price or no price for each copy that you convey,
+and you may offer support or warranty protection for a fee.
+
+  5. Conveying Modified Source Versions.
+
+  You may convey a work based on the Program, or the modifications to
+produce it from the Program, in the form of source code under the
+terms of section 4, provided that you also meet all of these conditions:
+
+    a) The work must carry prominent notices stating that you modified
+    it, and giving a relevant date.
+
+    b) The work must carry prominent notices stating that it is
+    released under this License and any conditions added under section
+    7.  This requirement modifies the requirement in section 4 to
+    "keep intact all notices".
+
+    c) You must license the entire work, as a whole, under this
+    License to anyone who comes into possession of a copy.  This
+    License will therefore apply, along with any applicable section 7
+    additional terms, to the whole of the work, and all its parts,
+    regardless of how they are packaged.  This License gives no
+    permission to license the work in any other way, but it does not
+    invalidate such permission if you have separately received it.
+
+    d) If the work has interactive user interfaces, each must display
+    Appropriate Legal Notices; however, if the Program has interactive
+    interfaces that do not display Appropriate Legal Notices, your
+    work need not make them do so.
+
+  A compilation of a covered work with other separate and independent
+works, which are not by their nature extensions of the covered work,
+and which are not combined with it such as to form a larger program,
+in or on a volume of a storage or distribution medium, is called an
+"aggregate" if the compilation and its resulting copyright are not
+used to limit the access or legal rights of the compilation's users
+beyond what the individual works permit.  Inclusion of a covered work
+in an aggregate does not cause this License to apply to the other
+parts of the aggregate.
+
+  6. Conveying Non-Source Forms.
+
+  You may convey a covered work in object code form under the terms
+of sections 4 and 5, provided that you also convey the
+machine-readable Corresponding Source under the terms of this License,
+in one of these ways:
+
+    a) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by the
+    Corresponding Source fixed on a durable physical medium
+    customarily used for software interchange.
+
+    b) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by a
+    written offer, valid for at least three years and valid for as
+    long as you offer spare parts or customer support for that product
+    model, to give anyone who possesses the object code either (1) a
+    copy of the Corresponding Source for all the software in the
+    product that is covered by this License, on a durable physical
+    medium customarily used for software interchange, for a price no
+    more than your reasonable cost of physically performing this
+    conveying of source, or (2) access to copy the
+    Corresponding Source from a network server at no charge.
+
+    c) Convey individual copies of the object code with a copy of the
+    written offer to provide the Corresponding Source.  This
+    alternative is allowed only occasionally and noncommercially, and
+    only if you received the object code with such an offer, in accord
+    with subsection 6b.
+
+    d) Convey the object code by offering access from a designated
+    place (gratis or for a charge), and offer equivalent access to the
+    Corresponding Source in the same way through the same place at no
+    further charge.  You need not require recipients to copy the
+    Corresponding Source along with the object code.  If the place to
+    copy the object code is a network server, the Corresponding Source
+    may be on a different server (operated by you or a third party)
+    that supports equivalent copying facilities, provided you maintain
+    clear directions next to the object code saying where to find the
+    Corresponding Source.  Regardless of what server hosts the
+    Corresponding Source, you remain obligated to ensure that it is
+    available for as long as needed to satisfy these requirements.
+
+    e) Convey the object code using peer-to-peer transmission, provided
+    you inform other peers where the object code and Corresponding
+    Source of the work are being offered to the general public at no
+    charge under subsection 6d.
+
+  A separable portion of the object code, whose source code is excluded
+from the Corresponding Source as a System Library, need not be
+included in conveying the object code work.
+
+  A "User Product" is either (1) a "consumer product", which means any
+tangible personal property which is normally used for personal, family,
+or household purposes, or (2) anything designed or sold for incorporation
+into a dwelling.  In determining whether a product is a consumer product,
+doubtful cases shall be resolved in favor of coverage.  For a particular
+product received by a particular user, "normally used" refers to a
+typical or common use of that class of product, regardless of the status
+of the particular user or of the way in which the particular user
+actually uses, or expects or is expected to use, the product.  A product
+is a consumer product regardless of whether the product has substantial
+commercial, industrial or non-consumer uses, unless such uses represent
+the only significant mode of use of the product.
+
+  "Installation Information" for a User Product means any methods,
+procedures, authorization keys, or other information required to install
+and execute modified versions of a covered work in that User Product from
+a modified version of its Corresponding Source.  The information must
+suffice to ensure that the continued functioning of the modified object
+code is in no case prevented or interfered with solely because
+modification has been made.
+
+  If you convey an object code work under this section in, or with, or
+specifically for use in, a User Product, and the conveying occurs as
+part of a transaction in which the right of possession and use of the
+User Product is transferred to the recipient in perpetuity or for a
+fixed term (regardless of how the transaction is characterized), the
+Corresponding Source conveyed under this section must be accompanied
+by the Installation Information.  But this requirement does not apply
+if neither you nor any third party retains the ability to install
+modified object code on the User Product (for example, the work has
+been installed in ROM).
+
+  The requirement to provide Installation Information does not include a
+requirement to continue to provide support service, warranty, or updates
+for a work that has been modified or installed by the recipient, or for
+the User Product in which it has been modified or installed.  Access to a
+network may be denied when the modification itself materially and
+adversely affects the operation of the network or violates the rules and
+protocols for communication across the network.
+
+  Corresponding Source conveyed, and Installation Information provided,
+in accord with this section must be in a format that is publicly
+documented (and with an implementation available to the public in
+source code form), and must require no special password or key for
+unpacking, reading or copying.
+
+  7. Additional Terms.
+
+  "Additional permissions" are terms that supplement the terms of this
+License by making exceptions from one or more of its conditions.
+Additional permissions that are applicable to the entire Program shall
+be treated as though they were included in this License, to the extent
+that they are valid under applicable law.  If additional permissions
+apply only to part of the Program, that part may be used separately
+under those permissions, but the entire Program remains governed by
+this License without regard to the additional permissions.
+
+  When you convey a copy of a covered work, you may at your option
+remove any additional permissions from that copy, or from any part of
+it.  (Additional permissions may be written to require their own
+removal in certain cases when you modify the work.)  You may place
+additional permissions on material, added by you to a covered work,
+for which you have or can give appropriate copyright permission.
+
+  Notwithstanding any other provision of this License, for material you
+add to a covered work, you may (if authorized by the copyright holders of
+that material) supplement the terms of this License with terms:
+
+    a) Disclaiming warranty or limiting liability differently from the
+    terms of sections 15 and 16 of this License; or
+
+    b) Requiring preservation of specified reasonable legal notices or
+    author attributions in that material or in the Appropriate Legal
+    Notices displayed by works containing it; or
+
+    c) Prohibiting misrepresentation of the origin of that material, or
+    requiring that modified versions of such material be marked in
+    reasonable ways as different from the original version; or
+
+    d) Limiting the use for publicity purposes of names of licensors or
+    authors of the material; or
+
+    e) Declining to grant rights under trademark law for use of some
+    trade names, trademarks, or service marks; or
+
+    f) Requiring indemnification of licensors and authors of that
+    material by anyone who conveys the material (or modified versions of
+    it) with contractual assumptions of liability to the recipient, for
+    any liability that these contractual assumptions directly impose on
+    those licensors and authors.
+
+  All other non-permissive additional terms are considered "further
+restrictions" within the meaning of section 10.  If the Program as you
+received it, or any part of it, contains a notice stating that it is
+governed by this License along with a term that is a further
+restriction, you may remove that term.  If a license document contains
+a further restriction but permits relicensing or conveying under this
+License, you may add to a covered work material governed by the terms
+of that license document, provided that the further restriction does
+not survive such relicensing or conveying.
+
+  If you add terms to a covered work in accord with this section, you
+must place, in the relevant source files, a statement of the
+additional terms that apply to those files, or a notice indicating
+where to find the applicable terms.
+
+  Additional terms, permissive or non-permissive, may be stated in the
+form of a separately written license, or stated as exceptions;
+the above requirements apply either way.
+
+  8. Termination.
+
+  You may not propagate or modify a covered work except as expressly
+provided under this License.  Any attempt otherwise to propagate or
+modify it is void, and will automatically terminate your rights under
+this License (including any patent licenses granted under the third
+paragraph of section 11).
+
+  However, if you cease all violation of this License, then your
+license from a particular copyright holder is reinstated (a)
+provisionally, unless and until the copyright holder explicitly and
+finally terminates your license, and (b) permanently, if the copyright
+holder fails to notify you of the violation by some reasonable means
+prior to 60 days after the cessation.
+
+  Moreover, your license from a particular copyright holder is
+reinstated permanently if the copyright holder notifies you of the
+violation by some reasonable means, this is the first time you have
+received notice of violation of this License (for any work) from that
+copyright holder, and you cure the violation prior to 30 days after
+your receipt of the notice.
+
+  Termination of your rights under this section does not terminate the
+licenses of parties who have received copies or rights from you under
+this License.  If your rights have been terminated and not permanently
+reinstated, you do not qualify to receive new licenses for the same
+material under section 10.
+
+  9. Acceptance Not Required for Having Copies.
+
+  You are not required to accept this License in order to receive or
+run a copy of the Program.  Ancillary propagation of a covered work
+occurring solely as a consequence of using peer-to-peer transmission
+to receive a copy likewise does not require acceptance.  However,
+nothing other than this License grants you permission to propagate or
+modify any covered work.  These actions infringe copyright if you do
+not accept this License.  Therefore, by modifying or propagating a
+covered work, you indicate your acceptance of this License to do so.
+
+  10. Automatic Licensing of Downstream Recipients.
+
+  Each time you convey a covered work, the recipient automatically
+receives a license from the original licensors, to run, modify and
+propagate that work, subject to this License.  You are not responsible
+for enforcing compliance by third parties with this License.
+
+  An "entity transaction" is a transaction transferring control of an
+organization, or substantially all assets of one, or subdividing an
+organization, or merging organizations.  If propagation of a covered
+work results from an entity transaction, each party to that
+transaction who receives a copy of the work also receives whatever
+licenses to the work the party's predecessor in interest had or could
+give under the previous paragraph, plus a right to possession of the
+Corresponding Source of the work from the predecessor in interest, if
+the predecessor has it or can get it with reasonable efforts.
+
+  You may not impose any further restrictions on the exercise of the
+rights granted or affirmed under this License.  For example, you may
+not impose a license fee, royalty, or other charge for exercise of
+rights granted under this License, and you may not initiate litigation
+(including a cross-claim or counterclaim in a lawsuit) alleging that
+any patent claim is infringed by making, using, selling, offering for
+sale, or importing the Program or any portion of it.
+
+  11. Patents.
+
+  A "contributor" is a copyright holder who authorizes use under this
+License of the Program or a work on which the Program is based.  The
+work thus licensed is called the contributor's "contributor version".
+
+  A contributor's "essential patent claims" are all patent claims
+owned or controlled by the contributor, whether already acquired or
+hereafter acquired, that would be infringed by some manner, permitted
+by this License, of making, using, or selling its contributor version,
+but do not include claims that would be infringed only as a
+consequence of further modification of the contributor version.  For
+purposes of this definition, "control" includes the right to grant
+patent sublicenses in a manner consistent with the requirements of
+this License.
+
+  Each contributor grants you a non-exclusive, worldwide, royalty-free
+patent license under the contributor's essential patent claims, to
+make, use, sell, offer for sale, import and otherwise run, modify and
+propagate the contents of its contributor version.
+
+  In the following three paragraphs, a "patent license" is any express
+agreement or commitment, however denominated, not to enforce a patent
+(such as an express permission to practice a patent or covenant not to
+sue for patent infringement).  To "grant" such a patent license to a
+party means to make such an agreement or commitment not to enforce a
+patent against the party.
+
+  If you convey a covered work, knowingly relying on a patent license,
+and the Corresponding Source of the work is not available for anyone
+to copy, free of charge and under the terms of this License, through a
+publicly available network server or other readily accessible means,
+then you must either (1) cause the Corresponding Source to be so
+available, or (2) arrange to deprive yourself of the benefit of the
+patent license for this particular work, or (3) arrange, in a manner
+consistent with the requirements of this License, to extend the patent
+license to downstream recipients.  "Knowingly relying" means you have
+actual knowledge that, but for the patent license, your conveying the
+covered work in a country, or your recipient's use of the covered work
+in a country, would infringe one or more identifiable patents in that
+country that you have reason to believe are valid.
+
+  If, pursuant to or in connection with a single transaction or
+arrangement, you convey, or propagate by procuring conveyance of, a
+covered work, and grant a patent license to some of the parties
+receiving the covered work authorizing them to use, propagate, modify
+or convey a specific copy of the covered work, then the patent license
+you grant is automatically extended to all recipients of the covered
+work and works based on it.
+
+  A patent license is "discriminatory" if it does not include within
+the scope of its coverage, prohibits the exercise of, or is
+conditioned on the non-exercise of one or more of the rights that are
+specifically granted under this License.  You may not convey a covered
+work if you are a party to an arrangement with a third party that is
+in the business of distributing software, under which you make payment
+to the third party based on the extent of your activity of conveying
+the work, and under which the third party grants, to any of the
+parties who would receive the covered work from you, a discriminatory
+patent license (a) in connection with copies of the covered work
+conveyed by you (or copies made from those copies), or (b) primarily
+for and in connection with specific products or compilations that
+contain the covered work, unless you entered into that arrangement,
+or that patent license was granted, prior to 28 March 2007.
+
+  Nothing in this License shall be construed as excluding or limiting
+any implied license or other defenses to infringement that may
+otherwise be available to you under applicable patent law.
+
+  12. No Surrender of Others' Freedom.
+
+  If conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot convey a
+covered work so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you may
+not convey it at all.  For example, if you agree to terms that obligate you
+to collect a royalty for further conveying from those to whom you convey
+the Program, the only way you could satisfy both those terms and this
+License would be to refrain entirely from conveying the Program.
+
+  13. Use with the GNU Affero General Public License.
+
+  Notwithstanding any other provision of this License, you have
+permission to link or combine any covered work with a work licensed
+under version 3 of the GNU Affero General Public License into a single
+combined work, and to convey the resulting work.  The terms of this
+License will continue to apply to the part which is the covered work,
+but the special requirements of the GNU Affero General Public License,
+section 13, concerning interaction through a network will apply to the
+combination as such.
+
+  14. Revised Versions of this License.
+
+  The Free Software Foundation may publish revised and/or new versions of
+the GNU General Public License from time to time.  Such new versions will
+be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+  Each version is given a distinguishing version number.  If the
+Program specifies that a certain numbered version of the GNU General
+Public License "or any later version" applies to it, you have the
+option of following the terms and conditions either of that numbered
+version or of any later version published by the Free Software
+Foundation.  If the Program does not specify a version number of the
+GNU General Public License, you may choose any version ever published
+by the Free Software Foundation.
+
+  If the Program specifies that a proxy can decide which future
+versions of the GNU General Public License can be used, that proxy's
+public statement of acceptance of a version permanently authorizes you
+to choose that version for the Program.
+
+  Later license versions may give you additional or different
+permissions.  However, no additional obligations are imposed on any
+author or copyright holder as a result of your choosing to follow a
+later version.
+
+  15. Disclaimer of Warranty.
+
+  THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
+APPLICABLE LAW.  EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
+HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
+OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
+THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+PURPOSE.  THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
+IS WITH YOU.  SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
+ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
+
+  16. Limitation of Liability.
+
+  IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
+THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
+GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
+USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
+DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
+PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
+EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
+SUCH DAMAGES.
+
+  17. Interpretation of Sections 15 and 16.
+
+  If the disclaimer of warranty and limitation of liability provided
+above cannot be given local legal effect according to their terms,
+reviewing courts shall apply local law that most closely approximates
+an absolute waiver of all civil liability in connection with the
+Program, unless a warranty or assumption of liability accompanies a
+copy of the Program in return for a fee.
+
+                     END OF TERMS AND CONDITIONS
+
+            How to Apply These Terms to Your New Programs
+
+  If you develop a new program, and you want it to be of the greatest
+possible use to the public, the best way to achieve this is to make it
+free software which everyone can redistribute and change under these terms.
+
+  To do so, attach the following notices to the program.  It is safest
+to attach them to the start of each source file to most effectively
+state the exclusion of warranty; and each file should have at least
+the "copyright" line and a pointer to where the full notice is found.
+
+    ComfyUI Lora Manager - A ComfyUI custom node for managing models
+    Copyright (C) 2025  Will Miao
+
+    This program is free software: you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation, either version 3 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License
+    along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+Also add information on how to contact you by electronic and paper mail.
+
+  If the program does terminal interaction, make it output a short
+notice like this when it starts in an interactive mode:
+
+    ComfyUI Lora Manager  Copyright (C) 2025  Will Miao
+    This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
+    This is free software, and you are welcome to redistribute it
+    under certain conditions; type `show c' for details.
+
+The hypothetical commands `show w' and `show c' should show the appropriate
+parts of the General Public License.  Of course, your program's commands
+might be different; for a GUI interface, you would use an "about box".
+
+  You should also get your employer (if you work as a programmer) or school,
+if any, to sign a "copyright disclaimer" for the program, if necessary.
+For more information on this, and how to apply and follow the GNU GPL, see
+<https://www.gnu.org/licenses/>.
+
+  The GNU General Public License does not permit incorporating your program
+into proprietary programs.  If your program is a subroutine library, you
+may consider it more useful to permit linking proprietary applications with
+the library.  If this is what you want to do, use the GNU Lesser General
+Public License instead of this License.  But first, please read
+<https://www.gnu.org/licenses/why-not-lgpl.html>.

README.md

@@ -1,56 +1,104 @@
 # ComfyUI LoRA Manager
 
-A web-based management interface designed to help you organize and manage your local LoRA models in ComfyUI. Access the interface at: `http://localhost:8188/loras`
+> **Revolutionize your workflow with the ultimate LoRA companion for ComfyUI!**
 
-![Interface Preview](https://github.com/willmiao/ComfyUI-Lora-Manager/blob/main/static/images/Screenshot%202025-01-27%20172349.png)
+[![Discord](https://img.shields.io/discord/1346296675538571315?color=7289DA&label=Discord&logo=discord&logoColor=white)](https://discord.gg/vcqNrWVFvM)
+[![Release](https://img.shields.io/github/v/release/willmiao/ComfyUI-Lora-Manager?include_prereleases&color=blue&logo=github)](https://github.com/willmiao/ComfyUI-Lora-Manager/releases)
+[![Release Date](https://img.shields.io/github/release-date/willmiao/ComfyUI-Lora-Manager?color=green&logo=github)](https://github.com/willmiao/ComfyUI-Lora-Manager/releases)
+
+A comprehensive toolset that streamlines organizing, downloading, and applying LoRA models in ComfyUI. With powerful features like recipe management, checkpoint organization, and one-click workflow integration, working with models becomes faster, smoother, and significantly easier. Access the interface at: `http://localhost:8188/loras`
+
+![Interface Preview](https://github.com/willmiao/ComfyUI-Lora-Manager/blob/main/static/images/screenshot.png)
 
 ## 📺 Tutorial: One-Click LoRA Integration
 Watch this quick tutorial to learn how to use the new one-click LoRA integration feature:
 
-[![One-Click LoRA Integration Tutorial](https://img.youtube.com/vi/qS95OjX3e70/0.jpg)](https://youtu.be/qS95OjX3e70)
+[![One-Click LoRA Integration Tutorial](https://github.com/willmiao/ComfyUI-Lora-Manager/blob/main/static/images/video-thumbnails/getting-started.jpg)](https://youtu.be/hvKw31YpE-U)
+
+## 🌐 Browser Extension
+Enhance your Civitai browsing experience with our companion browser extension! See which models you already have, download new ones with a single click, and manage your downloads efficiently.
+
+![LM Civitai Extension Preview](https://github.com/willmiao/ComfyUI-Lora-Manager/blob/main/wiki-images/civitai-models-page.png)
+
+<div>
+  <a href="https://chromewebstore.google.com/detail/lm-civitai-extension/capigligggeijgmocnaflanlbghnamgm?utm_source=item-share-cb" style="display: inline-block; background-color: #4285F4; color: white; padding: 8px 16px; text-decoration: none; border-radius: 4px; font-weight: bold; margin: 10px 0;">
+    <img src="https://www.google.com/chrome/static/images/chrome-logo.svg" width="20" style="vertical-align: middle; margin-right: 8px;"> Get Extension from Chrome Web Store
+  </a>
+</div>
+
+<div id="firefox-install" class="install-ok"><a href="https://github.com/willmiao/lm-civitai-extension-firefox/releases/latest/download/extension.xpi">📦 Install Firefox Extension (reviewed and verified by Mozilla)</a></div>
+
+📚 [Learn More: Complete Tutorial](https://github.com/willmiao/ComfyUI-Lora-Manager/wiki/LoRA-Manager-Civitai-Extension-(Chrome-Extension))
 
 ---
 
 ## Release Notes
 
-### v0.7.36
-* Enhanced LoRA details view with model descriptions and tags display
-* Added tag filtering system for improved model discovery
-* Implemented editable trigger words functionality
-* Improved TriggerWord Toggle node with new group mode option for granular control
-* Added new Lora Stacker node with cross-compatibility support (works with efficiency nodes, ComfyRoll, easy-use, etc.)
-* Fixed several bugs
+### v0.9.15
+* **Filter Presets** - Save filter combinations as presets for quick switching and reapplication.
+* **Bug Fixes** - Fixed various bugs for improved stability.
 
-### v0.7.35-beta
-* Added base model filtering
-* Implemented bulk operations (copy syntax, move multiple LoRAs)
-* Added ability to edit LoRA model names in details view
-* Added update checker with notification system
-* Added support modal for user feedback and community links
+### v0.9.14
+* **LoRA Cycler Node** - Introduced a new LoRA Cycler node that enables iteration through specified LoRAs with support for repeat count and pause iteration functionality. Refer to the new "Lora Cycler" template workflow for concrete example.
+* **Enhanced Prompt Node with Tag Autocomplete** - Enhanced the Prompt node with comprehensive tag autocomplete based on merged Danbooru + e621 tags. Supports tag search and autocomplete functionality. Implemented a command system with shortcuts like `/char` or `/artist` for category-specific tag searching. Added `/ac` or `/noac` commands to quickly enable or disable autocomplete. Refer to the "Lora Manager Basic" template workflow in ComfyUI -> Templates -> ComfyUI-Lora-Manager for detailed tips.
+* **Bug Fixes & Stability** - Addressed multiple bugs and improved overall stability.
 
-### v0.7.33
-* Enhanced LoRA Loader node with visual strength adjustment widgets
-* Added toggle switches for LoRA enable/disable
-* Implemented image tooltips for LoRA preview
-* Added TriggerWord Toggle node with visual word selection
-* Fixed various bugs and improved stability
+### v0.9.12
+* **LoRA Randomizer System** - Introduced a comprehensive LoRA randomization system featuring LoRA Pool and LoRA Randomizer nodes for flexible and dynamic generation workflows.
+* **LoRA Randomizer Template** - Refer to the new "LoRA Randomizer" template workflow for detailed examples of flexible randomization modes, lock & reuse options, and other features.
+* **Recipe Folders** - Introduced a folder system for the Recipes page, allowing users to freely organize recipes just like they do with models.
+* **Recipe Bulk Operations** - Added bulk mode support for batch moving, deleting, and setting base models for selected recipes with intuitive controls like click-and-drag selection, drag-to-folder, and Ctrl+A (Select All).
+* **Prompt Search & Sorting** - Search recipes by prompt content and sort by Recipe Name, Imported Date, or LoRA Count for better browsing.
+* **Recipe Favorites** - Mark specific recipes as favorites for quick access.
+* **Video Recipe Support** - Enabled support for video recipes (import via LM extension or URL; video file import not supported).
+* **Performance Improvements** - Fixed performance issues for dramatically improved startup and loading speed. After first scan, subsequent loads are instant regardless of collection size.
+* **ComfyUI Nodes 2.0 Support** - Basic support for ComfyUI Nodes 2.0.
 
-### v0.7.3
-* Added "Lora Loader (LoraManager)" custom node for workflows
-* Implemented one-click LoRA integration
-* Added direct copying of LoRA syntax from manager interface
-* Added automatic preset strength value application
-* Added automatic trigger word loading
+### v0.9.10
+* **Smarter Update Matching** - Users can now choose to check and group updates by matching base model only or with no base-model constraint; version lists also support toggling between same-base versions or all versions.
+* **Flexible Tag Filtering** - The filter panel now supports tag exclusion: click a tag to include, click again to exclude, and click a third time to clear, enabling stronger and more flexible tag filters.
+* **License Visibility & Controls** - Model detail headers and ComfyUI preview popups now show Civitai license icons. The filter panel gains license include/exclude options, and a new global context menu action, "Refresh license metadata," fetches missing license data.
+* **Recipe Improvements** - Recipes now allow importing with zero LoRAs, and recipe detail pages show the related checkpoint for easier reference.
+* **Better ZIP Downloads** - When downloading models packaged in ZIPs, model files are extracted into the target model folder; ZIPs containing multiple model files (e.g., WanVideo high/low LoRA pairs) are added as separate models.
+* **Template Workflow Update** - Refreshed the "Illustrious Pony Example" template workflow with usage guidance for each LoRA Manager node.
+* **Bug Fixes & Stability** - General fixes and stability improvements.
 
-### v0.7.0
-* Added direct CivitAI integration for downloading LoRAs
-* Implemented version selection for model downloads
-* Added target folder selection for downloads
-* Added context menu with quick actions
-* Added force refresh for CivitAI data
-* Implemented LoRA movement between folders
-* Added personal usage tips and notes for LoRAs
-* Improved performance for details window
+### v0.9.9
+* **Check for Updates Feature** - Users can now check for updates for all models or selected models in bulk mode. Models with available updates will display an "update available" badge on their model card, and users can filter to show only models with updates.
+* **Model Versions Management** - Added a new Versions tab in the model modal that centralizes all versions of a model, providing download, delete, and ignore update functions.
+* **Send Checkpoint to ComfyUI** - Users can now click the send button on a checkpoint card to send the checkpoint directly to the current workflow's checkpoint or diffusion model loader node in ComfyUI.
+* **Customizable Model Card Display** - Added a new setting that allows users to choose whether to display the model name or filename on model cards.
+* **New Path Template Placeholders** - Added new path template placeholders: `{model_name}` and `{version_name}` for more flexible organization.
+* **ComfyUI Auto Path Correction Setting** - Added a new setting within ComfyUI to enable or disable the auto path correction feature.
+
+### v0.9.8
+* **Full CivArchive API Support** - Added complete support for the CivArchive API as a fallback metadata source beyond Civitai API. Models deleted from Civitai can now still retrieve metadata through the CivArchive API.
+* **Download Models from CivArchive** - Added support for downloading models directly from CivArchive, similar to downloading from Civitai. Simply click the Download button and paste the model URL to download the corresponding model.
+* **Custom Priority Tags** - Introduced Custom Priority Tags feature, allowing users to define custom priority tags. These tags will appear as suggestions when editing tags or during auto organization/download using default paths, providing more precise and controlled folder organization. [Guide](https://github.com/willmiao/ComfyUI-Lora-Manager/wiki/Priority-Tags-Configuration-Guide)
+* **Drag and Drop Tag Reordering** - Added drag and drop functionality to reorder tags in the tags edit mode for improved usability.
+* **Download Control in Example Images Panel** - Added stop control in the Download Example Images Panel for better download management.
+* **Prompt (LoraManager) Node with Autocomplete** - Added new Prompt (LoraManager) node with autocomplete feature for adding embeddings.
+* **Lora Manager Nodes in Subgraphs** - Lora Manager nodes now support being placed within subgraphs for more flexible workflow organization.
+
+### v0.9.6
+* **Metadata Archive Database Support** - Added the ability to download and utilize a metadata archive database, enabling access to metadata for models that have been deleted from CivitAI.
+* **App-Level Proxy Settings** - Introduced support for configuring a global proxy within the application, making it easier to use the manager behind network restrictions.
+* **Bug Fixes** - Various bug fixes for improved stability and reliability.
+
+### v0.9.2
+* **Bulk Auto-Organization Action** - Added a new bulk auto-organization feature. You can now select multiple models and automatically organize them according to your current path template settings for streamlined management.
+* **Bug Fixes** - Addressed several bugs to improve stability and reliability.
+
+### v0.9.1
+* **Enhanced Bulk Operations** - Improved bulk operations with Marquee Selection and a bulk operation context menu, providing a more intuitive, desktop-application-like user experience.
+* **New Bulk Actions** - Added bulk operations for adding tags and setting base models to multiple models simultaneously.
+
+### v0.9.0
+* **UI Overhaul for Enhanced Navigation** - Replaced the top flat folder tags with a new folder sidebar and breadcrumb navigation system for a more intuitive folder browsing and selection experience.
+* **Dual-Mode Folder Sidebar** - The new folder sidebar offers two display modes: 'List Mode,' which mirrors the classic folder view, and 'Tree Mode,' which presents a hierarchical folder structure for effortless navigation through nested directories.
+* **Internationalization Support** - Introduced multi-language support, now available in English, Simplified Chinese, Traditional Chinese, Spanish, Japanese, Korean, French, Russian, and German. Feedback from native speakers is welcome to improve the translations.
+* **Automatic Filename Conflict Resolution** - Implemented automatic file renaming (`original name + short hash`) to prevent conflicts when downloading or moving models.
+* **Performance Optimizations & Bug Fixes** - Various performance improvements and bug fixes for a more stable and responsive experience.
 
 [View Update History](./update_logs.md)
 
@@ -69,13 +117,6 @@ Watch this quick tutorial to learn how to use the new one-click LoRA integration
 - 🚀 **High Performance**
   - Fast model loading and browsing
   - Smooth scrolling through large collections
-  - Real-time updates when files change
-  
-- 📂 **Advanced Organization**
-  - Quick search with fuzzy matching
-  - Folder-based categorization
-  - Move LoRAs between folders
-  - Sort by name or date
   
 - 🌐 **Rich Model Integration**
   - Direct download from CivitAI
@@ -84,29 +125,51 @@ Watch this quick tutorial to learn how to use the new one-click LoRA integration
   - Trigger words at a glance
   - One-click workflow integration with preset values
   
+- 🔄 **Checkpoint Management**
+  - Scan and organize checkpoint models
+  - Filter and search your collection
+  - View and edit metadata
+  - Clean up and manage disk space
+  
+- 🧩 **LoRA Recipes**
+  - Save and share favorite LoRA combinations
+  - Preserve generation parameters for future reference
+  - Quick application to workflows
+  - Import/export functionality for community sharing
+  
 - 💻 **User Friendly**
   - One-click access from ComfyUI menu
   - Context menu for quick actions
   - Custom notes and usage tips
   - Multi-folder support
+  - Visual progress indicators during initialization
 
 ---
 
 ## Installation
 
-### Option 1: **ComfyUI Manager** (Recommended)
+### Option 1: **ComfyUI Manager** (Recommended for ComfyUI users)
 
 1. Open **ComfyUI**.
 2. Go to **Manager > Custom Node Manager**.
 3. Search for `lora-manager`.
 4. Click **Install**.
 
-### Option 2: **Manual Installation**
+### Option 2: **Portable Standalone Edition** (No ComfyUI required)
+
+1. Download the [Portable Package](https://github.com/willmiao/ComfyUI-Lora-Manager/releases/download/v0.9.8/lora_manager_portable.7z)
+2. Copy the provided `settings.json.example` file to create a new file named `settings.json` in `comfyui-lora-manager` folder.
+3. Edit the new `settings.json` to include your correct model folder paths and CivitAI API key
+   - Set `"use_portable_settings": true` if you want the configuration to remain inside the repository folder instead of your user settings directory.
+4. Run run.bat
+    - To change the startup port, edit `run.bat` and modify the parameter (e.g. `--port 9001`)
+
+### Option 3: **Manual Installation**
 
 ```bash
 git clone https://github.com/willmiao/ComfyUI-Lora-Manager.git
 cd ComfyUI-Lora-Manager
-pip install requirements.txt
+pip install -r requirements.txt
 ```
 
 ## Usage
@@ -127,11 +190,119 @@ pip install requirements.txt
    - Paste into the Lora Loader node's text input
    - The node will automatically apply preset strength and trigger words
 
+### Filename Format Patterns for Save Image Node
+
+The Save Image Node supports dynamic filename generation using pattern codes. You can customize how your images are named using the following format patterns:
+
+#### Available Pattern Codes
+
+- `%seed%` - Inserts the generation seed number
+- `%width%` - Inserts the image width
+- `%height%` - Inserts the image height
+- `%pprompt:N%` - Inserts the positive prompt (limited to N characters)
+- `%nprompt:N%` - Inserts the negative prompt (limited to N characters)
+- `%model:N%` - Inserts the model/checkpoint name (limited to N characters)
+- `%date%` - Inserts current date/time as "yyyyMMddhhmmss"
+- `%date:FORMAT%` - Inserts date using custom format with:
+  - `yyyy` - 4-digit year
+  - `yy` - 2-digit year
+  - `MM` - 2-digit month
+  - `dd` - 2-digit day
+  - `hh` - 2-digit hour
+  - `mm` - 2-digit minute
+  - `ss` - 2-digit second
+
+#### Examples
+
+- `image_%seed%` → `image_1234567890`
+- `gen_%width%x%height%` → `gen_512x768`
+- `%model:10%_%seed%` → `dreamshape_1234567890`
+- `%date:yyyy-MM-dd%` → `2025-04-28`
+- `%pprompt:20%_%seed%` → `beautiful landscape_1234567890`
+- `%model%_%date:yyMMdd%_%seed%` → `dreamshaper_v8_250428_1234567890`
+
+You can combine multiple patterns to create detailed, organized filenames for your generated images.
+
+### Standalone Mode
+
+You can now run LoRA Manager independently from ComfyUI:
+
+1. **For ComfyUI users**:
+   - Launch ComfyUI with LoRA Manager at least once to initialize the necessary path information in the `settings.json` file located in your user settings folder (see paths above).
+   - Make sure dependencies are installed: `pip install -r requirements.txt`
+   - From your ComfyUI root directory, run:
+     ```bash
+     python custom_nodes\comfyui-lora-manager\standalone.py
+     ```
+   - Access the interface at: `http://localhost:8188/loras`
+   - You can specify a different host or port with arguments:
+     ```bash
+     python custom_nodes\comfyui-lora-manager\standalone.py --host 127.0.0.1 --port 9000
+     ```
+
+2. **For non-ComfyUI users**:
+   - Copy the provided `settings.json.example` file to create a new file named `settings.json`. Update the API key, optional language, and folder paths only—the library registry is created automatically when LoRA Manager starts.
+   - Edit `settings.json` to include your correct model folder paths and CivitAI API key (you can leave the defaults until ready to configure them)
+   - Enable portable mode by setting `"use_portable_settings": true` if you prefer LoRA Manager to read and write the `settings.json` located in the project directory.
+   - Install required dependencies: `pip install -r requirements.txt`
+   - Run standalone mode:
+     ```bash
+     python standalone.py
+     ```
+   - Access the interface through your browser at: `http://localhost:8188/loras`
+
+   > **Note:** Existing installations automatically migrate the legacy `settings.json` from the plugin folder to the user settings directory the first time you launch this version.
+
+This standalone mode provides a lightweight option for managing your model and recipe collection without needing to run the full ComfyUI environment, making it useful even for users who primarily use other stable diffusion interfaces.
+
+## Testing & Coverage
+
+### Backend
+
+Install the development dependencies and run pytest with coverage reports:
+
+```bash
+pip install -r requirements-dev.txt
+COVERAGE_FILE=coverage/backend/.coverage pytest \
+  --cov=py \
+  --cov=standalone \
+  --cov-report=term-missing \
+  --cov-report=html:coverage/backend/html \
+  --cov-report=xml:coverage/backend/coverage.xml \
+  --cov-report=json:coverage/backend/coverage.json
+```
+
+HTML, XML, and JSON artifacts are stored under `coverage/backend/` so you can inspect hot spots locally or from CI artifacts.
+
+### Frontend
+
+Run the Vitest coverage suite to analyze widget hot spots:
+
+```bash
+npm run test:coverage
+```
+
 ---
 
 ## Contributing
 
-If you have suggestions, bug reports, or improvements, feel free to open an issue or contribute directly to the codebase. Pull requests are always welcome!
+Thank you for your interest in contributing to ComfyUI LoRA Manager! As this project is currently in its early stages and undergoing rapid development and refactoring, we are temporarily not accepting pull requests.
+
+However, your feedback and ideas are extremely valuable to us:
+- Please feel free to open issues for any bugs you encounter
+- Submit feature requests through GitHub issues
+- Share your suggestions for improvements
+
+We appreciate your understanding and look forward to potentially accepting code contributions once the project architecture stabilizes.
+
+---
+
+## Credits
+
+This project has been inspired by and benefited from other excellent ComfyUI extensions:
+
+- [ComfyUI-SaveImageWithMetaData](https://github.com/nkchocoai/ComfyUI-SaveImageWithMetaData) - For the image metadata functionality
+- [rgthree-comfy](https://github.com/rgthree/rgthree-comfy) - For the lora loader functionality
 
 ---
 
@@ -141,18 +312,16 @@ If you find this project helpful, consider supporting its development:
 
 [![ko-fi](https://ko-fi.com/img/githubbutton_sm.svg)](https://ko-fi.com/pixelpawsai)
 
+[![Patreon](https://img.shields.io/badge/Become%20a%20Patron-F96854.svg?style=for-the-badge&logo=patreon&logoColor=white)](https://patreon.com/PixelPawsAI)
+
+WeChat: [Click to view QR code](https://raw.githubusercontent.com/willmiao/ComfyUI-Lora-Manager/main/static/images/wechat-qr.webp)
+
 ## 💬 Community
 
 Join our Discord community for support, discussions, and updates:
 [Discord Server](https://discord.gg/vcqNrWVFvM)
 
 ---
+## Star History
 
-## 🗺️ Roadmap
-
-- ✅ One-click integration of LoRAs into ComfyUI workflows with preset strength values
-- 🤝 Improved usage tips retrieval from CivitAI model pages
-- 🔌 Integration with Power LoRA Loader and other management tools
-- 🛡️ Configurable NSFW level settings for content filtering
-
----
+[![Star History Chart](https://api.star-history.com/svg?repos=willmiao/ComfyUI-Lora-Manager&type=Date)](https://star-history.com/#willmiao/ComfyUI-Lora-Manager&Date)

__init__.py

@@ -1,16 +1,99 @@
-from .py.lora_manager import LoraManager
-from .py.nodes.lora_loader import LoraManagerLoader
-from .py.nodes.trigger_word_toggle import TriggerWordToggle
-from .py.nodes.lora_stacker import LoraStacker
+try:  # pragma: no cover - import fallback for pytest collection
+    from .py.lora_manager import LoraManager
+    from .py.nodes.lora_loader import LoraLoaderLM, LoraTextLoaderLM
+    from .py.nodes.trigger_word_toggle import TriggerWordToggleLM
+    from .py.nodes.prompt import PromptLM
+    from .py.nodes.text import TextLM
+    from .py.nodes.lora_stacker import LoraStackerLM
+    from .py.nodes.save_image import SaveImageLM
+    from .py.nodes.debug_metadata import DebugMetadataLM
+    from .py.nodes.wanvideo_lora_select import WanVideoLoraSelectLM
+    from .py.nodes.wanvideo_lora_select_from_text import WanVideoLoraTextSelectLM
+    from .py.nodes.lora_pool import LoraPoolLM
+    from .py.nodes.lora_randomizer import LoraRandomizerLM
+    from .py.nodes.lora_cycler import LoraCyclerLM
+    from .py.metadata_collector import init as init_metadata_collector
+except (
+    ImportError
+):  # pragma: no cover - allows running under pytest without package install
+    import importlib
+    import pathlib
+    import sys
+
+    package_root = pathlib.Path(__file__).resolve().parent
+    if str(package_root) not in sys.path:
+        sys.path.append(str(package_root))
+
+    PromptLM = importlib.import_module("py.nodes.prompt").PromptLM
+    TextLM = importlib.import_module("py.nodes.text").TextLM
+    LoraManager = importlib.import_module("py.lora_manager").LoraManager
+    LoraLoaderLM = importlib.import_module(
+        "py.nodes.lora_loader"
+    ).LoraLoaderLM
+    LoraTextLoaderLM = importlib.import_module(
+        "py.nodes.lora_loader"
+    ).LoraTextLoaderLM
+    TriggerWordToggleLM = importlib.import_module(
+        "py.nodes.trigger_word_toggle"
+    ).TriggerWordToggleLM
+    LoraStackerLM = importlib.import_module("py.nodes.lora_stacker").LoraStackerLM
+    SaveImageLM = importlib.import_module("py.nodes.save_image").SaveImageLM
+    DebugMetadataLM = importlib.import_module("py.nodes.debug_metadata").DebugMetadataLM
+    WanVideoLoraSelectLM = importlib.import_module(
+        "py.nodes.wanvideo_lora_select"
+    ).WanVideoLoraSelectLM
+    WanVideoLoraTextSelectLM = importlib.import_module(
+        "py.nodes.wanvideo_lora_select_from_text"
+    ).WanVideoLoraTextSelectLM
+    LoraPoolLM = importlib.import_module("py.nodes.lora_pool").LoraPoolLM
+    LoraRandomizerLM = importlib.import_module(
+        "py.nodes.lora_randomizer"
+    ).LoraRandomizerLM
+    LoraCyclerLM = importlib.import_module(
+        "py.nodes.lora_cycler"
+    ).LoraCyclerLM
+    init_metadata_collector = importlib.import_module("py.metadata_collector").init
 
 NODE_CLASS_MAPPINGS = {
-    LoraManagerLoader.NAME: LoraManagerLoader,
-    TriggerWordToggle.NAME: TriggerWordToggle,
-    LoraStacker.NAME: LoraStacker
+    PromptLM.NAME: PromptLM,
+    TextLM.NAME: TextLM,
+    LoraLoaderLM.NAME: LoraLoaderLM,
+    LoraTextLoaderLM.NAME: LoraTextLoaderLM,
+    TriggerWordToggleLM.NAME: TriggerWordToggleLM,
+    LoraStackerLM.NAME: LoraStackerLM,
+    SaveImageLM.NAME: SaveImageLM,
+    DebugMetadataLM.NAME: DebugMetadataLM,
+    WanVideoLoraSelectLM.NAME: WanVideoLoraSelectLM,
+    WanVideoLoraTextSelectLM.NAME: WanVideoLoraTextSelectLM,
+    LoraPoolLM.NAME: LoraPoolLM,
+    LoraRandomizerLM.NAME: LoraRandomizerLM,
+    LoraCyclerLM.NAME: LoraCyclerLM,
 }
 
 WEB_DIRECTORY = "./web/comfyui"
 
+# Check and build Vue widgets if needed (development mode)
+try:
+    from .py.vue_widget_builder import check_and_build_vue_widgets
+
+    # Auto-build in development, warn only if fails
+    check_and_build_vue_widgets(auto_build=True, warn_only=True)
+except ImportError:
+    # Fallback for pytest
+    import importlib
+
+    check_and_build_vue_widgets = importlib.import_module(
+        "py.vue_widget_builder"
+    ).check_and_build_vue_widgets
+    check_and_build_vue_widgets(auto_build=True, warn_only=True)
+except Exception as e:
+    import logging
+
+    logging.warning(f"[LoRA Manager] Vue widget build check skipped: {e}")
+
+# Initialize metadata collector
+init_metadata_collector()
+
 # Register routes on import
 LoraManager.add_routes()
-__all__ = ['NODE_CLASS_MAPPINGS', 'WEB_DIRECTORY']
+__all__ = ["NODE_CLASS_MAPPINGS", "WEB_DIRECTORY"]

docs/LM-Extension-Wiki.md

@@ -0,0 +1,180 @@
+## Overview
+
+The **LoRA Manager Civitai Extension** is a Browser extension designed to work seamlessly with [LoRA Manager](https://github.com/willmiao/ComfyUI-Lora-Manager) to significantly enhance your browsing experience on [Civitai](https://civitai.com).
+It also supports browsing on [CivArchive](https://civarchive.com/) (formerly CivitaiArchive).
+
+With this extension, you can:
+
+✅ Instantly see which models are already present in your local library  
+✅ Download new models with a single click  
+✅ Manage downloads efficiently with queue and parallel download support  
+✅ Keep your downloaded models automatically organized according to your custom settings    
+
+![Civitai Models page](https://github.com/willmiao/ComfyUI-Lora-Manager/blob/main/wiki-images/civitai-models-page.png)  
+![CivArchive Models page](https://github.com/willmiao/ComfyUI-Lora-Manager/blob/main/wiki-images/civarchive-models-page.png)  
+
+---
+
+## Why Are All Features for Supporters Only?
+
+I love building tools for the Stable Diffusion and ComfyUI communities, and LoRA Manager is a passion project that I've poured countless hours into. When I created this companion extension, my hope was to offer its core features for free, as a thank-you to all of you.
+
+Unfortunately, I've reached a point where I need to be realistic. The level of support from the free model has been far lower than what's needed to justify the continuous development and maintenance for both projects. It was a difficult decision, but I've chosen to make the extension's features exclusive to supporters.
+
+This change is crucial for me to be able to continue dedicating my time to improving the free and open-source LoRA Manager, which I'm committed to keeping available for everyone.
+
+Your support does more than just unlock a few features—it allows me to keep innovating and ensures the core LoRA Manager project thrives. I'm incredibly grateful for your understanding and any support you can offer. ❤️
+
+(_For those who previously supported me on Ko-fi with a one-time donation, I'll be sending out license keys individually as a thank-you._)
+
+
+---
+
+## Installation
+
+### Supported Browsers & Installation Methods
+
+| Browser            | Installation Method                                                                 |
+|--------------------|-------------------------------------------------------------------------------------|
+| **Google Chrome**   | [Chrome Web Store link](https://chromewebstore.google.com/detail/capigligggeijgmocnaflanlbghnamgm?utm_source=item-share-cb) |
+| **Microsoft Edge**  | Install via Chrome Web Store (compatible)                                          |
+| **Brave Browser**   | Install via Chrome Web Store (compatible)                                          |
+| **Opera**           | Install via Chrome Web Store (compatible)                                          |
+| **Firefox**         | <div id="firefox-install" class="install-ok"><a href="https://github.com/willmiao/lm-civitai-extension-firefox/releases/latest/download/extension.xpi">📦 Install Firefox Extension (reviewed and verified by Mozilla)</a></div>             |
+
+For non-Chrome browsers (e.g., Microsoft Edge), you can typically install extensions from the Chrome Web Store by following these steps: open the extension’s Chrome Web Store page, click 'Get extension', then click 'Allow' when prompted to enable installations from other stores, and finally click 'Add extension' to complete the installation.
+
+---
+
+## Privacy & Security
+
+I understand concerns around browser extensions and privacy, and I want to be fully transparent about how the **LM Civitai Extension** works:
+
+- **Reviewed and Verified**  
+  This extension has been **manually reviewed and approved by the Chrome Web Store**. The Firefox version uses the **exact same code** (only the packaging format differs) and has passed **Mozilla’s Add-on review**.
+
+- **Minimal Network Access**  
+  The only external server this extension connects to is:  
+  **`https://willmiao.shop`** — used solely for **license validation**.
+
+  It does **not collect, transmit, or store any personal or usage data**.  
+  No browsing history, no user IDs, no analytics, no hidden trackers.
+
+- **Local-Only Model Detection**  
+  Model detection and LoRA Manager communication all happen **locally** within your browser, directly interacting with your local LoRA Manager backend.
+
+I value your trust and are committed to keeping your local setup private and secure. If you have any questions, feel free to reach out!
+
+---
+
+## How to Use
+
+After installing the extension, you'll automatically receive a **7-day trial** to explore all features.
+
+When the extension is correctly installed and your license is valid:
+
+- Open **Civitai**, and you'll see visual indicators added by the extension on model cards, showing:
+  - ✅ Models already present in your local library  
+  - ⬇️ A download button for models not in your library  
+
+Clicking the download button adds the corresponding model version to the download queue, waiting to be downloaded. You can set up to **5 models to download simultaneously**.
+
+### Visual Indicators Appear On:
+
+- **Home Page** — Featured models  
+- **Models Page**  
+- **Creator Profiles** — If the creator has set their models to be visible  
+- **Recommended Resources** — On individual model pages  
+
+### Version Buttons on Model Pages
+
+On a specific model page, visual indicators also appear on version buttons, showing which versions are already in your local library.
+
+When switching to a specific version by clicking a version button:
+
+- Clicking the download button will open a dropdown:
+  - Download via **LoRA Manager**  
+  - Download via **Original Download** (browser download)  
+
+You can check **Remember my choice** to set your preferred default. You can change this setting anytime in the extension's settings.
+
+![Civitai Model Page](https://github.com/willmiao/ComfyUI-Lora-Manager/blob/main/wiki-images/civitai-model-page.png)
+
+### Resources on Image Pages (2025-08-05) — now shows in-library indicators for image resources. ‘Import image as recipe’ coming soon!
+
+![Civitai Image Page](https://github.com/willmiao/ComfyUI-Lora-Manager/blob/main/wiki-images/civitai-image-page.jpg)
+
+---
+
+## Model Download Location & LoRA Manager Settings
+
+To use the **one-click download function**, you must first set:
+
+- Your **Default LoRAs Root**  
+- Your **Default Checkpoints Root**  
+
+These are set within LoRA Manager's settings.
+
+When everything is configured, downloaded model files will be placed in:  
+
+`<Default_Models_Root>/<Base_Model_of_the_Model>/<First_Tag_of_the_Model>`
+
+
+### Update: Default Path Customization (2025-07-21)  
+
+A new setting to customize the default download path has been added in the nightly version. You can now personalize where models are saved when downloading via the LM Civitai Extension.
+
+![Default Path Customization](https://github.com/willmiao/ComfyUI-Lora-Manager/blob/main/wiki-images/default-path-customization.png)
+
+The previous YAML path mapping file will be deprecated—settings will now be unified in settings.json to simplify configuration.
+
+---
+
+## Backend Port Configuration
+
+If your **ComfyUI** or **LoRA Manager** backend is running on a port **other than the default 8188**, you must configure the backend port in the extension's settings.
+
+After correctly setting and saving the port, you'll see in the extension's header area:  
+- A **Healthy** status with the tooltip:  `Connected to LoRA Manager on port xxxx`
+
+
+---
+
+## Advanced Usage
+
+### Connecting to a Remote LoRA Manager
+
+If your LoRA Manager is running on another computer, you can still connect from your browser using port forwarding.
+
+> **Why can't you set a remote IP directly?**
+>
+> For privacy and security, the extension only requests access to `http://127.0.0.1/*`. Supporting remote IPs would require much broader permissions, which may be rejected by browser stores and could raise user concerns.
+
+**Solution: Port Forwarding with `socat`**
+
+On your browser computer, run:
+
+`socat TCP-LISTEN:8188,bind=127.0.0.1,fork TCP:REMOTE.IP.ADDRESS.HERE:8188`
+
+- Replace `REMOTE.IP.ADDRESS.HERE` with the IP of the machine running LoRA Manager.
+- Adjust the port if needed.
+
+This lets the extension connect to `127.0.0.1:8188` as usual, with traffic forwarded to your remote server.
+
+_Thanks to user **Temikus** for sharing this solution!_
+
+---
+
+## Roadmap
+
+The extension will evolve alongside **LoRA Manager** improvements. Planned features include:
+
+- [x] Support for **additional model types** (e.g., embeddings)
+- [ ] One-click **Recipe Import**  
+- [x] Display of in-library status for all resources in the **Resources Used** section of the image page  
+- [x] One-click **Auto-organize Models**
+
+**Stay tuned — and thank you for your support!**
+
+---
+

docs/architecture/example_images_routes.md

@@ -0,0 +1,93 @@
+# Example image route architecture
+
+The example image routing stack mirrors the layered model route stack described in
+[`docs/architecture/model_routes.md`](model_routes.md). HTTP wiring, controller setup,
+handler orchestration, and long-running workflows now live in clearly separated modules so
+we can extend download/import behaviour without touching the entire feature surface.
+
+```mermaid
+graph TD
+    subgraph HTTP
+        A[ExampleImagesRouteRegistrar] -->|binds| B[ExampleImagesRoutes controller]
+    end
+    subgraph Application
+        B --> C[ExampleImagesHandlerSet]
+        C --> D1[Handlers]
+        D1 --> E1[Use cases]
+        E1 --> F1[Download manager / processor / file manager]
+    end
+    subgraph Side Effects
+        F1 --> G1[Filesystem]
+        F1 --> G2[Model metadata]
+        F1 --> G3[WebSocket progress]
+    end
+```
+
+## Layer responsibilities
+
+| Layer | Module(s) | Responsibility |
+| --- | --- | --- |
+| Registrar | `py/routes/example_images_route_registrar.py` | Declarative catalogue of every example image endpoint plus helpers that bind them to an `aiohttp` router. Keeps HTTP concerns symmetrical with the model registrar. |
+| Controller | `py/routes/example_images_routes.py` | Lazily constructs `ExampleImagesHandlerSet`, injects defaults for the download manager, processor, and file manager, and exposes the registrar-ready mapping just like `BaseModelRoutes`. |
+| Handler set | `py/routes/handlers/example_images_handlers.py` | Groups HTTP adapters by concern (downloads, imports/deletes, filesystem access). Each handler translates domain errors into HTTP responses and defers to a use case or utility service. |
+| Use cases | `py/services/use_cases/example_images/*.py` | Encapsulate orchestration for downloads and imports. They validate input, translate concurrency/configuration errors, and keep handler logic declarative. |
+| Supporting services | `py/utils/example_images_download_manager.py`, `py/utils/example_images_processor.py`, `py/utils/example_images_file_manager.py` | Execute long-running work: pull assets from Civitai, persist uploads, clean metadata, expose filesystem actions with guardrails, and broadcast progress snapshots. |
+
+## Handler responsibilities & invariants
+
+`ExampleImagesHandlerSet` flattens the handler objects into the `{"handler_name": coroutine}`
+mapping consumed by the registrar. The table below outlines how each handler collaborates
+with the use cases and utilities.
+
+| Handler | Key endpoints | Collaborators | Contracts |
+| --- | --- | --- | --- |
+| `ExampleImagesDownloadHandler` | `/api/lm/download-example-images`, `/api/lm/example-images-status`, `/api/lm/pause-example-images`, `/api/lm/resume-example-images`, `/api/lm/force-download-example-images` | `DownloadExampleImagesUseCase`, `DownloadManager` | Delegates payload validation and concurrency checks to the use case; progress/status endpoints expose the same snapshot used for WebSocket broadcasts; pause/resume surface `DownloadNotRunningError` as HTTP 400 instead of 500. |
+| `ExampleImagesManagementHandler` | `/api/lm/import-example-images`, `/api/lm/delete-example-image` | `ImportExampleImagesUseCase`, `ExampleImagesProcessor` | Multipart uploads are streamed to disk via the use case; validation failures return HTTP 400 with no filesystem side effects; deletion funnels through the processor to prune metadata and cached images consistently. |
+| `ExampleImagesFileHandler` | `/api/lm/open-example-images-folder`, `/api/lm/example-image-files`, `/api/lm/has-example-images` | `ExampleImagesFileManager` | Centralises filesystem access, enforcing settings-based root paths and returning HTTP 400/404 for missing configuration or folders; responses always include `success`/`has_images` booleans for UI consumption. |
+
+## Use case boundaries
+
+| Use case | Entry point | Dependencies | Guarantees |
+| --- | --- | --- | --- |
+| `DownloadExampleImagesUseCase` | `execute(payload)` | `DownloadManager.start_download`, download configuration errors | Raises `DownloadExampleImagesInProgressError` when the manager reports an active job, rewraps configuration errors into `DownloadExampleImagesConfigurationError`, and lets `ExampleImagesDownloadError` bubble as 500s so handlers do not duplicate logging. |
+| `ImportExampleImagesUseCase` | `execute(request)` | `ExampleImagesProcessor.import_images`, temporary file helpers | Supports multipart or JSON payloads, normalises file paths into a single list, cleans up temp files even on failure, and maps validation issues to `ImportExampleImagesValidationError` for HTTP 400 responses. |
+
+## Maintaining critical invariants
+
+* **Shared progress snapshots** - The download handler returns the same snapshot built by
+  `DownloadManager`, guaranteeing parity between HTTP polling endpoints and WebSocket
+  progress events.
+* **Safe filesystem access** - All folder/file actions flow through
+  `ExampleImagesFileManager`, which validates the configured example image root and ensures
+  responses never leak absolute paths outside the allowed directory.
+* **Metadata hygiene** - Import/delete operations run through `ExampleImagesProcessor`,
+  which updates model metadata via `MetadataManager` and notifies the relevant scanners so
+  cache state stays in sync.
+
+## Migration notes
+
+The refactor brings the example image stack in line with the model/recipe stacks:
+
+1. `ExampleImagesRouteRegistrar` now owns the declarative route list. Downstream projects
+   should rely on `ExampleImagesRoutes.to_route_mapping()` instead of manually wiring
+   handler callables.
+2. `ExampleImagesRoutes` caches its `ExampleImagesHandlerSet` just like
+   `BaseModelRoutes`. If you previously instantiated handlers directly, inject custom
+   collaborators via the controller constructor (`download_manager`, `processor`,
+   `file_manager`) to keep test seams predictable.
+3. Tests that mocked `ExampleImagesRoutes.setup_routes` should switch to patching
+   `DownloadExampleImagesUseCase`/`ImportExampleImagesUseCase` at import time. The handlers
+   expect those abstractions to surface validation/concurrency errors, and bypassing them
+   will skip the HTTP-friendly error mapping.
+
+## Extending the stack
+
+1. Add the endpoint to `ROUTE_DEFINITIONS` with a unique `handler_name`.
+2. Expose the coroutine on an existing handler class (or create a new handler and extend
+   `ExampleImagesHandlerSet`).
+3. Wire additional services or factories inside `_build_handler_set` on
+   `ExampleImagesRoutes`, mirroring how the model stack introduces new use cases.
+
+`tests/routes/test_example_images_routes.py` exercises registrar binding, download pause
+flows, and import validations. Use it as a template when introducing new handler
+collaborators or error mappings.

docs/architecture/model_routes.md

@@ -0,0 +1,100 @@
+# Base model route architecture
+
+The model routing stack now splits HTTP wiring, orchestration logic, and
+business rules into discrete layers.  The goal is to make it obvious where a
+new collaborator should live and which contract it must honour.  The diagram
+below captures the end-to-end flow for a typical request:
+
+```mermaid
+graph TD
+    subgraph HTTP
+        A[ModelRouteRegistrar] -->|binds| B[BaseModelRoutes handler proxy]
+    end
+    subgraph Application
+        B --> C[ModelHandlerSet]
+        C --> D1[Handlers]
+        D1 --> E1[Use cases]
+        E1 --> F1[Services / scanners]
+    end
+    subgraph Side Effects
+        F1 --> G1[Cache & metadata]
+        F1 --> G2[Filesystem]
+        F1 --> G3[WebSocket state]
+    end
+```
+
+Every box maps to a concrete module:
+
+| Layer | Module(s) | Responsibility |
+| --- | --- | --- |
+| Registrar | `py/routes/model_route_registrar.py` | Declarative list of routes shared by every model type and helper methods for binding them to an `aiohttp` application. |
+| Route controller | `py/routes/base_model_routes.py` | Constructs the handler graph, injects shared services, exposes proxies that surface `503 Service not ready` when the model service has not been attached. |
+| Handler set | `py/routes/handlers/model_handlers.py` | Thin HTTP adapters grouped by concern (page rendering, listings, mutations, queries, downloads, CivitAI integration, move operations, auto-organize). |
+| Use cases | `py/services/use_cases/*.py` | Encapsulate long-running flows (`DownloadModelUseCase`, `BulkMetadataRefreshUseCase`, `AutoOrganizeUseCase`).  They normalise validation errors and concurrency constraints before returning control to the handlers. |
+| Services | `py/services/*.py` | Existing services and scanners that mutate caches, write metadata, move files, and broadcast WebSocket updates. |
+
+## Handler responsibilities & contracts
+
+`ModelHandlerSet` flattens the handler objects into the exact callables used by
+the registrar.  The table below highlights the separation of concerns within
+the set and the invariants that must hold after each handler returns.
+
+| Handler | Key endpoints | Collaborators | Contracts |
+| --- | --- | --- | --- |
+| `ModelPageView` | `/{prefix}` | `SettingsManager`, `server_i18n`, Jinja environment, `service.scanner` | Template is rendered with `is_initializing` flag when caches are cold; i18n filter is registered exactly once per environment instance. |
+| `ModelListingHandler` | `/api/lm/{prefix}/list` | `service.get_paginated_data`, `service.format_response` | Listings respect pagination query parameters and cap `page_size` at 100; every item is formatted before response. |
+| `ModelManagementHandler` | Mutations (delete, exclude, metadata, preview, tags, rename, bulk delete, duplicate verification) | `ModelLifecycleService`, `MetadataSyncService`, `PreviewAssetService`, `TagUpdateService`, scanner cache/index | Cache state mirrors filesystem changes: deletes prune cache & hash index, preview replacements synchronise metadata and cache NSFW levels, metadata saves trigger cache resort when names change. |
+| `ModelQueryHandler` | Read-only queries (top tags, folders, duplicates, metadata, URLs) | Service query helpers & scanner cache | Outputs always wrapped in `{"success": True}` when no error; duplicate/filename grouping omits empty entries; invalid parameters (e.g. missing `model_root`) return HTTP 400. |
+| `ModelDownloadHandler` | `/api/lm/download-model`, `/download-model-get`, `/download-progress/{id}`, `/cancel-download-get` | `DownloadModelUseCase`, `DownloadCoordinator`, `WebSocketManager` | Payload validation errors become HTTP 400 without mutating download progress cache; early-access failures surface as HTTP 401; successful downloads cache progress snapshots that back both WebSocket broadcasts and polling endpoints. |
+| `ModelCivitaiHandler` | CivitAI metadata routes | `MetadataSyncService`, metadata provider factory, `BulkMetadataRefreshUseCase` | `fetch_all_civitai` streams progress via `WebSocketBroadcastCallback`; version lookups validate model type before returning; local availability fields derive from hash lookups without mutating cache state. |
+| `ModelMoveHandler` | `move_model`, `move_models_bulk` | `ModelMoveService` | Moves execute atomically per request; bulk operations aggregate success/failure per file set. |
+| `ModelAutoOrganizeHandler` | `/api/lm/{prefix}/auto-organize` (GET/POST), `/auto-organize-progress` | `AutoOrganizeUseCase`, `WebSocketProgressCallback`, `WebSocketManager` | Enforces single-flight execution using the shared lock; progress broadcasts remain available to polling clients until explicitly cleared; conflicts return HTTP 409 with a descriptive error. |
+
+## Use case boundaries
+
+Each use case exposes a narrow asynchronous API that hides the underlying
+services.  Their error mapping is essential for predictable HTTP responses.
+
+| Use case | Entry point | Dependencies | Guarantees |
+| --- | --- | --- | --- |
+| `DownloadModelUseCase` | `execute(payload)` | `DownloadCoordinator.schedule_download` | Translates `ValueError` into `DownloadModelValidationError` for HTTP 400, recognises early-access errors (`"401"` in message) and surfaces them as `DownloadModelEarlyAccessError`, forwards success dictionaries untouched. |
+| `AutoOrganizeUseCase` | `execute(file_paths, progress_callback)` | `ModelFileService.auto_organize_models`, `WebSocketManager` lock | Guarded by `ws_manager` lock + status checks; raises `AutoOrganizeInProgressError` before invoking the file service when another run is already active. |
+| `BulkMetadataRefreshUseCase` | `execute_with_error_handling(progress_callback)` | `MetadataSyncService`, `SettingsManager`, `WebSocketBroadcastCallback` | Iterates through cached models, applies metadata sync, emits progress snapshots that handlers broadcast unchanged. |
+
+## Maintaining legacy contracts
+
+The refactor preserves the invariants called out in the previous architecture
+notes.  The most critical ones are reiterated here to emphasise the
+collaboration points:
+
+1. **Cache mutations** – Delete, exclude, rename, and bulk delete operations are
+   channelled through `ModelManagementHandler`.  The handler delegates to
+   `ModelLifecycleService` or `MetadataSyncService`, and the scanner cache is
+   mutated in-place before the handler returns.  The accompanying tests assert
+   that `scanner._cache.raw_data` and `scanner._hash_index` stay in sync after
+   each mutation.
+2. **Preview updates** – `PreviewAssetService.replace_preview` writes the new
+   asset, `MetadataSyncService` persists the JSON metadata, and
+   `scanner.update_preview_in_cache` mirrors the change.  The handler returns
+   the static URL produced by `config.get_preview_static_url`, keeping browser
+   clients in lockstep with disk state.
+3. **Download progress** – `DownloadCoordinator.schedule_download` generates the
+   download identifier, registers a WebSocket progress callback, and caches the
+   latest numeric progress via `WebSocketManager`.  Both `download_model`
+   responses and `/download-progress/{id}` polling read from the same cache to
+   guarantee consistent progress reporting across transports.
+
+## Extending the stack
+
+To add a new shared route:
+
+1. Declare it in `COMMON_ROUTE_DEFINITIONS` using a unique handler name.
+2. Implement the corresponding coroutine on one of the handlers inside
+   `ModelHandlerSet` (or introduce a new handler class when the concern does not
+   fit existing ones).
+3. Inject additional dependencies in `BaseModelRoutes._create_handler_set` by
+   wiring services or use cases through the constructor parameters.
+
+Model-specific routes should continue to be registered inside the subclass
+implementation of `setup_specific_routes`, reusing the shared registrar where
+possible.

docs/architecture/multi_library_design.md

@@ -0,0 +1,34 @@
+# Multi-Library Management for Standalone Mode
+
+## Requirements Summary
+- **Independent libraries**: In standalone mode, users can maintain multiple libraries, where each library represents a distinct set of model folders (LoRAs, checkpoints, embeddings, etc.). Only one library is active at any given time, but users need a fast way to switch between them.
+- **Library-specific settings**: The fields that vary per library are `folder_paths`, `default_lora_root`, `default_checkpoint_root`, and `default_embedding_root` inside `settings.json`.
+- **Persistent caches**: Every library must have its own SQLite persistent model cache so that metadata generated for one library does not leak into another.
+- **Backward compatibility**: Existing single-library setups should continue to work. When no multi-library configuration is provided, the application should behave exactly as before.
+
+## Proposed Design
+1. **Library registry**
+   - Extend the standalone configuration to hold a list of libraries, each identified by a unique name.
+   - Each entry stores the folder path configuration plus any library-scoped metadata (e.g. creation time, display name).
+   - The active library key is stored separately to allow quick switching without rewriting the full config.
+2. **Settings management**
+   - Update `settings_manager` to load and persist the library registry. When a library is activated, hydrate the in-memory settings object with that library's folder configuration.
+   - Provide helper methods for creating, renaming, and deleting libraries, ensuring validation for duplicate names and path collisions.
+   - Continue writing the active library settings to `settings.json` for compatibility, while storing the registry in a new section such as `libraries`.
+3. **Persistent model cache**
+   - Derive the SQLite file path from the active library, e.g. `model_cache_<library>.sqlite` or a nested directory structure like `model_cache/<library>/models.sqlite`.
+   - Update `PersistentModelCache` so it resolves the database path dynamically whenever the active library changes. Ensure connections are closed before switching to avoid locking issues.
+   - Migrate existing single cache files by treating them as the default library's cache.
+4. **Model scanning workflow**
+   - Modify `ModelScanner` and related services to react to library switches by clearing in-memory caches, re-reading folder paths, and rehydrating metadata from the library-specific SQLite cache.
+   - Provide API endpoints in standalone mode to list libraries, activate one, and trigger a rescan.
+5. **UI/UX considerations**
+   - In the standalone UI, introduce a library selector component that surfaces available libraries and offers quick switching.
+   - Offer feedback when switching libraries (e.g. spinner while rescanning) and guard destructive actions with confirmation prompts.
+
+## Implementation Notes
+- **Data migration**: On startup, detect if the old `settings.json` structure is present. If so, create a default library entry using the current folder paths and point the active library to it.
+- **Thread safety**: Ensure that any long-running scans are cancelled or awaited before switching libraries to prevent race conditions in cache writes.
+- **Testing**: Add unit tests for the settings manager to cover library CRUD operations and cache path resolution. Include integration tests that simulate switching libraries and verifying that the correct models are loaded.
+- **Documentation**: Update user guides to explain how to define libraries, switch between them, and where the new cache files are stored.
+- **Extensibility**: Keep the design open to future per-library settings (e.g. auto-refresh intervals, metadata overrides) by storing library data as objects instead of flat maps.

docs/architecture/recipe_routes.md

@@ -0,0 +1,89 @@
+# Recipe route architecture
+
+The recipe routing stack now mirrors the modular model route design. HTTP
+bindings, controller wiring, handler orchestration, and business rules live in
+separate layers so new behaviours can be added without re-threading the entire
+feature. The diagram below outlines the flow for a typical request:
+
+```mermaid
+graph TD
+    subgraph HTTP
+        A[RecipeRouteRegistrar] -->|binds| B[RecipeRoutes controller]
+    end
+    subgraph Application
+        B --> C[RecipeHandlerSet]
+        C --> D1[Handlers]
+        D1 --> E1[Use cases]
+        E1 --> F1[Services / scanners]
+    end
+    subgraph Side Effects
+        F1 --> G1[Cache & fingerprint index]
+        F1 --> G2[Metadata files]
+        F1 --> G3[Temporary shares]
+    end
+```
+
+## Layer responsibilities
+
+| Layer | Module(s) | Responsibility |
+| --- | --- | --- |
+| Registrar | `py/routes/recipe_route_registrar.py` | Declarative list of every recipe endpoint and helper methods that bind them to an `aiohttp` application. |
+| Controller | `py/routes/base_recipe_routes.py`, `py/routes/recipe_routes.py` | Lazily resolves scanners/clients from the service registry, wires shared templates/i18n, instantiates `RecipeHandlerSet`, and exposes a `{handler_name: coroutine}` mapping for the registrar. |
+| Handler set | `py/routes/handlers/recipe_handlers.py` | Thin HTTP adapters grouped by concern (page view, listings, queries, mutations, sharing). They normalise responses and translate service exceptions into HTTP status codes. |
+| Services & scanners | `py/services/recipes/*.py`, `py/services/recipe_scanner.py`, `py/services/service_registry.py` | Concrete business logic: metadata parsing, persistence, sharing, fingerprint/index maintenance, and cache refresh. |
+
+## Handler responsibilities & invariants
+
+`RecipeHandlerSet` flattens purpose-built handler objects into the callables the
+registrar binds. Each handler is responsible for a narrow concern and enforces a
+set of invariants before returning:
+
+| Handler | Key endpoints | Collaborators | Contracts |
+| --- | --- | --- | --- |
+| `RecipePageView` | `/loras/recipes` | `SettingsManager`, `server_i18n`, Jinja environment, recipe scanner getter | Template rendered with `is_initializing` flag when caches are still warming; i18n filter registered exactly once per environment instance. |
+| `RecipeListingHandler` | `/api/lm/recipes`, `/api/lm/recipe/{id}` | `recipe_scanner.get_paginated_data`, `recipe_scanner.get_recipe_by_id` | Listings respect pagination and search filters; every item receives a `file_url` fallback even when metadata is incomplete; missing recipes become HTTP 404. |
+| `RecipeQueryHandler` | Tag/base-model stats, syntax, LoRA lookups | Recipe scanner cache, `format_recipe_file_url` helper | Cache snapshots are reused without forcing refresh; duplicate lookups collapse groups by fingerprint; syntax lookups return helpful errors when LoRAs are absent. |
+| `RecipeManagementHandler` | Save, update, reconnect, bulk delete, widget ingest | `RecipePersistenceService`, `RecipeAnalysisService`, recipe scanner | Persistence results propagate HTTP status codes; fingerprint/index updates flow through the scanner before returning; validation errors surface as HTTP 400 without touching disk. |
+| `RecipeAnalysisHandler` | Uploaded/local/remote analysis | `RecipeAnalysisService`, `civitai_client`, recipe scanner | Unsupported content types map to HTTP 400; download errors (`RecipeDownloadError`) are not retried; every response includes a `loras` array for client compatibility. |
+| `RecipeSharingHandler` | Share + download | `RecipeSharingService`, recipe scanner | Share responses provide a stable download URL and filename; expired shares surface as HTTP 404; downloads stream via `web.FileResponse` with attachment headers. |
+
+## Use case boundaries
+
+The dedicated services encapsulate long-running work so handlers stay thin.
+
+| Use case | Entry point | Dependencies | Guarantees |
+| --- | --- | --- | --- |
+| `RecipeAnalysisService` | `analyze_uploaded_image`, `analyze_remote_image`, `analyze_local_image`, `analyze_widget_metadata` | `ExifUtils`, `RecipeParserFactory`, downloader factory, optional metadata collector/processor | Normalises missing/invalid payloads into `RecipeValidationError`; generates consistent fingerprint data to keep duplicate detection stable; temporary files are cleaned up after every analysis path. |
+| `RecipePersistenceService` | `save_recipe`, `delete_recipe`, `update_recipe`, `reconnect_lora`, `bulk_delete`, `save_recipe_from_widget` | `ExifUtils`, recipe scanner, card preview sizing constants | Writes images/JSON metadata atomically; updates scanner caches and hash indices before returning; recalculates fingerprints whenever LoRA assignments change. |
+| `RecipeSharingService` | `share_recipe`, `prepare_download` | `tempfile`, recipe scanner | Copies originals to TTL-managed temp files; metadata lookups re-use the scanner; expired shares trigger cleanup and `RecipeNotFoundError`. |
+
+## Maintaining critical invariants
+
+* **Cache updates** – Mutations (`save`, `delete`, `bulk_delete`, `update`) call
+  back into the recipe scanner to mutate the in-memory cache and fingerprint
+  index before returning a response. Tests assert that these methods are invoked
+  even when stubbing persistence.
+* **Fingerprint management** – `RecipePersistenceService` recomputes
+  fingerprints whenever LoRA metadata changes and duplicate lookups use those
+  fingerprints to group recipes. Handlers bubble the resulting IDs so clients
+  can merge duplicates without an extra fetch.
+* **Metadata synchronisation** – Saving or reconnecting a recipe updates the
+  JSON sidecar, refreshes embedded metadata via `ExifUtils`, and instructs the
+  scanner to resort its cache. Sharing relies on this metadata to generate
+  filenames and ensure downloads stay in sync with on-disk state.
+
+## Extending the stack
+
+1. Declare the new endpoint in `ROUTE_DEFINITIONS` with a unique handler name.
+2. Implement the coroutine on an existing handler or introduce a new handler
+   class inside `py/routes/handlers/recipe_handlers.py` when the concern does
+   not fit existing ones.
+3. Wire additional collaborators inside
+   `BaseRecipeRoutes._create_handler_set` (inject new services or factories) and
+   expose helper getters on the handler owner if the handler needs to share
+   utilities.
+
+Integration tests in `tests/routes/test_recipe_routes.py` exercise the listing,
+mutation, analysis-error, and sharing paths end-to-end, ensuring the controller
+and handler wiring remains valid as new capabilities are added.
+

docs/custom_priority_tags_format.md

@@ -0,0 +1,46 @@
+# Custom Priority Tag Format Proposal
+
+To support user-defined priority tags with flexible aliasing across different model types, the configuration will be stored as editable strings. The format balances readability with enough structure for parsing on both the backend and frontend.
+
+## Format Overview
+
+- Each model type is declared on its own line: `model_type: entries`.
+- Entries are comma-separated and ordered by priority from highest to lowest.
+- An entry may be a single canonical tag (e.g., `realistic`) or a canonical tag with aliases.
+- Canonical tags define the final folder name that should be used when matching that entry.
+- Aliases are enclosed in parentheses and separated by `|` (vertical bar).
+- All matching is case-insensitive; stored canonical names preserve the user-specified casing for folder creation and UI suggestions.
+
+### Grammar
+
+```
+priority-config := model-config { "\n" model-config }
+model-config    := model-type ":" entry-list
+model-type      := <identifier without spaces>
+entry-list      := entry { "," entry }
+entry           := canonical [ "(" alias { "|" alias } ")" ]
+canonical       := <tag text without parentheses or commas>
+alias           := <tag text without parentheses, commas, or pipes>
+```
+
+Examples:
+
+```
+lora: celebrity(celeb|celebrity), stylized, character(char)
+checkpoint: realistic(realism|realistic), anime(anime-style|toon)
+embedding: face, celeb(celebrity|celeb)
+```
+
+## Parsing Notes
+
+- Whitespace around separators is ignored to make manual editing more forgiving.
+- Duplicate canonical tags within the same model type collapse to a single entry; the first definition wins.
+- Aliases map to their canonical tag. When generating folder names, the canonical form is used.
+- Tags that do not match any alias or canonical entry fall back to the first tag in the model's tag list, preserving current behavior.
+
+## Usage
+
+- **Backend:** Convert each model type's string into an ordered list of canonical tags with alias sets. During path generation, iterate by priority order and match tags against both canonical names and their aliases.
+- **Frontend:** Surface canonical tags as suggestions, optionally displaying aliases in tooltips or secondary text. Input validation should warn about duplicate aliases within the same model type.
+
+This format allows users to customize priority tag handling per model type while keeping editing simple and avoiding proliferation of folder names through alias normalization.

docs/dom-widgets/README.md

@@ -0,0 +1,28 @@
+# DOM Widgets Documentation
+
+Documentation for custom DOM widget development in ComfyUI LoRA Manager.
+
+## Files
+
+- **[Value Persistence Best Practices](value-persistence-best-practices.md)** - Essential guide for implementing text input DOM widgets that persist values correctly
+
+## Key Lessons
+
+### Common Anti-Patterns
+
+❌ **Don't**: Create internal state variables  
+❌ **Don't**: Use v-model for text inputs  
+❌ **Don't**: Add serializeValue, onSetValue callbacks  
+❌ **Don't**: Watch props.widget.value  
+
+### Best Practices
+
+✅ **Do**: Use DOM element as single source of truth  
+✅ **Do**: Store DOM reference on widget.inputEl  
+✅ **Do**: Direct getValue/setValue to DOM  
+✅ **Do**: Clean up reference on unmount
+
+## Related Documentation
+
+- [DOM Widget Development Guide](../dom_widget_dev_guide.md) - Comprehensive guide for building DOM widgets
+- [ComfyUI Built-in Example](../../../../code/ComfyUI_frontend/src/renderer/extensions/vueNodes/widgets/composables/useStringWidget.ts) - Reference implementation

docs/dom-widgets/value-persistence-best-practices.md

@@ -0,0 +1,225 @@
+# DOM Widget Value Persistence - Best Practices
+
+## Overview
+
+DOM widgets require different persistence patterns depending on their complexity. This document covers two patterns:
+
+1. **Simple Text Widgets**: DOM element as source of truth (e.g., textarea, input)
+2. **Complex Widgets**: Internal value with `widget.callback` (e.g., LoraPoolWidget, RandomizerWidget)
+
+## Understanding ComfyUI's Built-in Callback Mechanism
+
+When `widget.value` is set (e.g., during workflow load), ComfyUI's `domWidget.ts` triggers this flow:
+
+```typescript
+// From ComfyUI_frontend/src/scripts/domWidget.ts:146-149
+set value(v: V) {
+  this.options.setValue?.(v)      // 1. Update internal state
+  this.callback?.(this.value)     // 2. Notify listeners for UI updates
+}
+```
+
+This means:
+- `setValue()` handles storing the value
+- `widget.callback()` is automatically called to notify the UI
+- You don't need custom callback mechanisms like `onSetValue`
+
+---
+
+## Pattern 1: Simple Text Input Widgets
+
+For widgets where the value IS the DOM element's text content (textarea, input fields).
+
+### When to Use
+
+- Single text input/textarea widgets
+- Value is a simple string
+- No complex state management needed
+
+### Implementation
+
+**main.ts:**
+```typescript
+const widget = node.addDOMWidget(name, type, container, {
+  getValue() {
+    return widget.inputEl?.value ?? ''
+  },
+  setValue(v: string) {
+    if (widget.inputEl) {
+      widget.inputEl.value = v ?? ''
+    }
+  }
+})
+```
+
+**Vue Component:**
+```typescript
+onMounted(() => {
+  if (textareaRef.value) {
+    props.widget.inputEl = textareaRef.value
+  }
+})
+
+onUnmounted(() => {
+  if (props.widget.inputEl === textareaRef.value) {
+    props.widget.inputEl = undefined
+  }
+})
+```
+
+### Why This Works
+
+- Single source of truth: the DOM element
+- `getValue()` reads directly from DOM
+- `setValue()` writes directly to DOM
+- No sync issues between multiple state variables
+
+---
+
+## Pattern 2: Complex Widgets
+
+For widgets with structured data (JSON configs, arrays, objects) where the value cannot be stored in a DOM element.
+
+### When to Use
+
+- Value is a complex object/array (e.g., `{ loras: [...], settings: {...} }`)
+- Multiple UI elements contribute to the value
+- Vue reactive state manages the UI
+
+### Implementation
+
+**main.ts:**
+```typescript
+let internalValue: MyConfig | undefined
+
+const widget = node.addDOMWidget(name, type, container, {
+  getValue() {
+    return internalValue
+  },
+  setValue(v: MyConfig) {
+    internalValue = v
+    // NO custom onSetValue needed - widget.callback is called automatically
+  },
+  serialize: true  // Ensure value is saved with workflow
+})
+```
+
+**Vue Component:**
+```typescript
+const config = ref<MyConfig>(getDefaultConfig())
+
+onMounted(() => {
+  // Set up callback for UI updates when widget.value changes externally
+  // (e.g., workflow load, undo/redo)
+  props.widget.callback = (newValue: MyConfig) => {
+    if (newValue) {
+      config.value = newValue
+    }
+  }
+
+  // Restore initial value if workflow was already loaded
+  if (props.widget.value) {
+    config.value = props.widget.value
+  }
+})
+
+// When UI changes, update widget value
+function onConfigChange(newConfig: MyConfig) {
+  config.value = newConfig
+  props.widget.value = newConfig  // This also triggers callback
+}
+```
+
+### Why This Works
+
+1. **Clear separation**: `internalValue` stores the data, Vue ref manages the UI
+2. **Built-in callback**: ComfyUI calls `widget.callback()` automatically after `setValue()`
+3. **Bidirectional sync**:
+   - External → UI: `setValue()` updates `internalValue`, `callback()` updates Vue ref
+   - UI → External: User interaction updates Vue ref, which updates `widget.value`
+
+---
+
+## Common Mistakes
+
+### ❌ Creating custom callback mechanisms
+
+```typescript
+// Wrong - unnecessary complexity
+setValue(v: MyConfig) {
+  internalValue = v
+  widget.onSetValue?.(v)  // Don't add this - use widget.callback instead
+}
+```
+
+Use the built-in `widget.callback` instead.
+
+### ❌ Using v-model for simple text inputs in DOM widgets
+
+```html
+<!-- Wrong - creates sync issues -->
+<textarea v-model="textValue" />
+
+<!-- Right for simple text widgets -->
+<textarea ref="textareaRef" @input="onInput" />
+```
+
+### ❌ Watching props.widget.value
+
+```typescript
+// Wrong - creates race conditions
+watch(() => props.widget.value, (newValue) => {
+  config.value = newValue
+})
+```
+
+Use `widget.callback` instead - it's called at the right time in the lifecycle.
+
+### ❌ Multiple sources of truth
+
+```typescript
+// Wrong - who is the source of truth?
+let internalValue = ''        // State 1
+const textValue = ref('')     // State 2
+const domElement = textarea   // State 3
+props.widget.value            // State 4
+```
+
+Choose ONE source of truth:
+- **Simple widgets**: DOM element
+- **Complex widgets**: `internalValue` (with Vue ref as derived UI state)
+
+### ❌ Adding serializeValue for simple widgets
+
+```typescript
+// Wrong - getValue/setValue handle serialization
+props.widget.serializeValue = async () => textValue.value
+```
+
+---
+
+## Decision Guide
+
+| Widget Type | Source of Truth | Use `widget.callback` | Example |
+|-------------|-----------------|----------------------|---------|
+| Simple text input | DOM element (`inputEl`) | Optional | AutocompleteTextWidget |
+| Complex config | `internalValue` | Yes, for UI sync | LoraPoolWidget |
+| Vue component widget | Vue ref + `internalValue` | Yes | RandomizerWidget |
+
+---
+
+## Testing Checklist
+
+- [ ] Load workflow - value restores correctly
+- [ ] Switch workflow - value persists
+- [ ] Reload page - value persists
+- [ ] UI interaction - value updates
+- [ ] Undo/redo - value syncs with UI
+- [ ] No console errors
+
+---
+
+## References
+
+- ComfyUI DOMWidget implementation: `ComfyUI_frontend/src/scripts/domWidget.ts`
+- Simple text widget example: `ComfyUI_frontend/src/renderer/extensions/vueNodes/widgets/composables/useStringWidget.ts`

docs/dom_widget_dev_guide.md

@@ -0,0 +1,546 @@
+# DOMWidget Development Guide
+
+This document provides a comprehensive guide for developing custom DOMWidgets in ComfyUI using Vanilla JavaScript. DOMWidgets allow you to embed standard HTML elements (div, video, canvas, input, etc.) into ComfyUI nodes while benefitting from the frontend's automatic layout and zoom management.
+
+## 1. Core Concepts
+
+In ComfyUI, a `DOMWidget` extends the default LiteGraph Canvas rendering logic. It maintains an HTML layer on top of the Canvas, making complex interactions and media displays significantly easier to implement than pure Canvas drawing.
+
+### Key APIs
+*   **`app.registerExtension`**: The entry point for registering extensions.
+*   **`getCustomWidgets`**: A hook for defining new widget types associated with specific input types.
+*   **`node.addDOMWidget`**: The core method to add HTML elements to a node.
+
+---
+
+## 2. Basic Structure
+
+A standard custom DOMWidget extension typically follows this structure:
+
+```javascript
+import { app } from "../../scripts/app.js";
+
+app.registerExtension({
+    name: "My.Custom.Extension",
+    async getCustomWidgets() {
+        return {
+            // Define a new widget type named "MY_WIDGET_TYPE"
+            MY_WIDGET_TYPE(node, inputName, inputData, app) {
+                // 1. Create the HTML element
+                const container = document.createElement("div");
+                container.innerHTML = "Hello <b>DOMWidget</b>!";
+                
+                // 2. Setup styles (Optional but recommended)
+                container.style.color = "white";
+                container.style.backgroundColor = "#222";
+                container.style.padding = "5px";
+
+                // 3. Add the DOMWidget and return the result
+                const widget = node.addDOMWidget(inputName, "MY_WIDGET_TYPE", container, {
+                    // Configuration options
+                    getValue() {
+                        return container.innerText;
+                    },
+                    setValue(v) {
+                        container.innerText = v;
+                    }
+                });
+
+                // 4. Return in the standard format
+                return { widget };
+            }
+        };
+    }
+});
+```
+
+---
+
+## ComfyUI Dual Rendering Modes
+
+ComfyUI frontend supports two rendering modes:
+
+| Mode | Description | DOM Structure |
+| :--- | :--- | :--- |
+| **Canvas Mode** | Traditional rendering where widgets are rendered on top of canvas using absolute positioning | Uses `.dom-widget` class on containers |
+| **Vue DOM Mode** | New rendering mode where nodes and widgets are rendered as Vue components | Uses `.lg-node-widget` class on containers with dynamic IDs (e.g., `v-1-0`) |
+
+### Mode Switching
+
+The frontend switches between modes via `LiteGraph.vueNodesMode` boolean:
+- `LiteGraph.vueNodesMode = true` → Vue DOM Mode
+- `LiteGraph.vueNodesMode = false` → Canvas Mode
+
+**Key Behavior**: Mode switching triggers DOM re-rendering WITHOUT page reload. Widget elements are destroyed and recreated, so any event listeners or references to old DOM elements become invalid.
+
+### Testing Mode Switches via Chrome DevTools MCP
+
+```javascript
+// Trigger render mode change
+LiteGraph.vueNodesMode = !LiteGraph.vueNodesMode;
+
+// Force canvas redraw (optional but helps trigger re-render)
+if (app.canvas) {
+    app.canvas.draw(true, true);
+}
+```
+
+### Development Notes
+
+When implementing widgets that attach event listeners or maintain external references:
+1. **Use `node.onRemoved`** to clean up when node is deleted
+2. **Detect DOM changes** by checking if widget input element is still in document: `document.body.contains(inputElement)`
+3. **Poll for mode changes** by watching `LiteGraph.vueNodesMode` and re-initializing when it changes
+4. **Use `loadedGraphNode` hook** for initial setup (guarantees DOM is fully rendered)
+
+
+---
+
+## 3. The `addDOMWidget` API
+
+```javascript
+node.addDOMWidget(name, type, element, options)
+```
+
+### Parameters
+1.  **`name`**: The internal name of the widget (usually matches the input name).
+2.  **`type`**: The type identifier for the widget.
+3.  **`element`**: The actual HTMLElement to embed.
+4.  **`options`**: (Object) Configuration for lifecycle, sizing, and persistence.
+
+### Common `options` Fields
+| Field | Type | Description |
+| :--- | :--- | :--- |
+| `getValue` | `Function` | Defines how to retrieve the widget's value for serialization. |
+| `setValue` | `Function` | Defines how to restore the widget's state from workflow data. |
+| `getMinHeight` | `Function` | Returns the minimum height in pixels. |
+| `getHeight` | `Function` | Returns the preferred height (supports numbers or percentage strings like `"50%"`). |
+| `onResize` | `Function` | Callback triggered when the widget is resized. |
+| `hideOnZoom`| `Boolean` | Whether to hide the DOM element when zoomed out to improve performance (default: `true`). |
+| `selectOn` | `string[]` | Events on the element that should trigger node selection (default: `['focus', 'click']`). |
+
+---
+
+## 4. Size Control
+
+Custom DOMWidgets must actively inform the parent Node of their size requirements to ensure the Node layout is calculated correctly and connection wires remain aligned.
+
+### 4.1 Core Mechanism
+
+Whether in Canvas Mode or Vue Mode, the underlying logic model (`LGraphNode`) calls the widget's `computeLayoutSize` method to determine dimensions. This logic is used to calculate the Node's total size and the position of input/output slots.
+
+### 4.2 Controlling Height
+
+It is recommended to use the `options` parameter to define height behavior.
+
+**Performance Note:** providing `getMinHeight` and `getHeight` via `options` allows the system to skip expensive DOM measurements (`getComputedStyle`) during rendering loop. This significantly improves performance and prevents FPS drops during node resizing.
+
+**Method 1: Using `options` (Recommended)**
+
+```javascript
+const widget = node.addDOMWidget("MyWidget", "custom", element, {
+    // Specify minimum height in pixels
+    getMinHeight: () => 150,
+    
+    // Or specify preferred height (pixels or percentage string)
+    // getHeight: () => "50%", 
+});
+```
+
+**Method 2: Using CSS Variables**
+
+You can also set specific CSS variables on the root element:
+
+```javascript
+element.style.setProperty("--comfy-widget-min-height", "150px");
+// or --comfy-widget-height
+```
+
+### 4.3 Controlling Width
+
+By default, a DOMWidget's width automatically stretches to fit the Node's width (which is determined by the Title or other Input Slots).
+
+If you must **force the Node to be wider** to accommodate your widget, you need to override the widget instance's `computeLayoutSize` method:
+
+```javascript
+const widget = node.addDOMWidget("WideWidget", "custom", element);
+
+// Override the default layout calculation
+widget.computeLayoutSize = (targetNode) => {
+    return {
+        minHeight: 150, // Must return height
+        minWidth: 300   // Force the Node to be at least 300px wide
+    };
+};
+```
+
+### 4.4 Dynamic Resizing
+
+If your widget's content changes dynamically (e.g., expanding sections, loading images, or CSS changes), the DOM element will resize, but the Canvas-rendered Node background and Slots will not automatically follow. You must manually trigger a synchronization.
+
+**The Update Sequence:**
+Whenever the **actual rendering height** of your DOM element changes, execute the following "three-step combo":
+
+```javascript
+// 1. Calculate the new optimal size for the node based on current widget requirements
+const newSize = node.computeSize();
+
+// 2. Apply the new size to the node model (updates bounding box and slot positions)
+node.setSize(newSize);
+
+// 3. Mark the canvas as dirty to trigger a redraw in the next animation frame
+node.setDirtyCanvas(true, true);
+```
+
+**Common Scenarios:**
+
+| Scenario | Actual Height Change? | Update Required? |
+| :--- | :--- | :--- |
+| **Expand/Collapse content** | **Yes** | ✅ **Yes**. Prevents widget from overflowing node boundaries. |
+| **Image/Video finished loading** | **Yes** | ✅ **Yes**. Initial height might be 0 until the media loads. |
+| **Changing `minHeight`** | **Maybe** | ❓ **Only if** the change causes the element's actual height to shift. |
+| **Changing font size/styles** | **Yes** | ✅ **Yes**. Text reflow often changes the total height. |
+| **User dragging node corner** | **Yes** | ❌ **No**. LiteGraph handles this internally. |
+
+---
+
+## 5. State Persistence (Serialization)
+
+### 5.1 Default Behavior
+
+DOMWidgets have **serialization enabled** by default (`serialize` property is `true`).
+*   **Saving**: ComfyUI attempts to read the widget's value to save into the Workflow file.
+*   **Loading**: ComfyUI reads the value from the Workflow file and assigns it to the widget.
+
+### 5.2 Custom Serialization
+
+To make persistence work effectively (saving internal DOM state and restoring it), you must implement `getValue` and `setValue` in the `options`:
+
+*   **`getValue`**: Returns the state to be saved (Number, String, or Object).
+*   **`setValue`**: Receives the restored value and updates the DOM element.
+
+**Example:**
+
+```javascript
+const inputEl = document.createElement("input");
+const widget = node.addDOMWidget("MyInput", "custom", inputEl, {
+    // 1. Called during Save
+    getValue: () => {
+        return inputEl.value;
+    },
+    // 2. Called during Load or Copy/Paste
+    setValue: (value) => {
+        inputEl.value = value || "";
+    }
+});
+
+// Optional: Listen for changes to update widget.value immediately
+inputEl.addEventListener("change", () => {
+    widget.value = inputEl.value; // Triggers callbacks
+});
+```
+
+> **⚠️ Important**: For Vue-based DOM widgets with text inputs, follow the [Value Persistence Best Practices](dom-widgets/value-persistence-best-practices.md) to avoid sync issues. Key takeaway: use DOM element as single source of truth, avoid internal state variables and v-model.
+
+### 5.3 The Restoration Mechanism (`configure`)
+
+*   **`configure(data)`**: When a Workflow is loaded, `LGraphNode` calls its `configure(data)` method.
+*   **`setValue` Chain**: During `configure`, the Node iterates over the saved `widgets_values` array and assigns each value (`widget.value = savedValue`). For DOMWidgets, this assignment triggers the `setValue` callback defined in your options.
+
+Therefore, `options.setValue` is the critical hook for restoring widget state.
+
+### 5.4 Disabling Serialization
+
+If your widget is purely for display (e.g., a real-time monitor or generated chart) and doesn't need to save state, disable serialization to reduce workflow file size.
+
+**Note**: You cannot set this via `options`. You must modify the widget instance directly.
+
+```javascript
+const widget = node.addDOMWidget("DisplayOnly", "custom", element);
+widget.serialize = false; // Explicitly disable
+```
+
+---
+
+## 6. Lifecycle & Events
+
+### 6.1 `onResize`
+
+When the Node size changes (e.g., user drags the corner), the widget can receive a notification via `options`:
+
+```javascript
+const widget = node.addDOMWidget("ResizingWidget", "custom", element, {
+    onResize: (w) => {
+        // 'w' is the widget instance
+        // Adjust internal DOM layout here if necessary
+        console.log("Widget resized");
+    }
+});
+```
+
+### 6.2 Construction & Mounting
+
+*   **Construction**: Occurs immediately when `addDOMWidget` is called.
+*   **Mounting**:
+    *   **Canvas Mode**: Appended to `.dom-widget-container` via `DomWidget.vue`.
+    *   **Vue Mode**: Appended inside the Node component via `WidgetDOM.vue`.
+    *   **Caution**: When `addDOMWidget` returns, the element may not be in the `document.body` yet. If you need to access layout properties like `getBoundingClientRect`, use `setTimeout` or wait for the first `onResize`.
+
+### 6.3 Cleanup
+
+If you create external references (like `setInterval` or global event listeners), ensure you clean them up using `node.onRemoved`:
+
+```javascript
+node.onRemoved = function() {
+    clearInterval(myInterval);
+    // Call original onRemoved if it existed
+};
+```
+
+---
+
+## 7. Styling & Best Practices
+
+### 7.1 Styling
+Since DOMWidgets are placed in absolute positioned containers or managed by Vue, ensure your container handles sizing gracefully:
+
+```javascript
+container.style.width = "100%";
+container.style.boxSizing = "border-box";
+```
+
+### 7.2 Path References
+When importing `app`, adjust the path based on your extension's folder depth. Typically:
+`import { app } from "../../scripts/app.js";`
+
+### 7.3 Security
+If setting `innerHTML` dynamically, ensure the content is sanitized or trusted to prevent XSS attacks.
+
+### 7.4 UI Constraints for ComfyUI Custom Node Widgets
+
+When developing DOMWidgets as internal UI widgets for ComfyUI custom nodes, keep the following constraints in mind:
+
+#### 7.4.1 Minimize Vertical Space
+
+ComfyUI nodes are often displayed in a compact graph view with many nodes visible simultaneously. Avoid excessive vertical spacing that could clutter the workspace.
+
+- Keep layouts compact and efficient
+- Use appropriate padding and margins (4-8px typically)
+- Stack related controls vertically but avoid unnecessary spacing
+
+#### 7.4.2 Avoid Dynamic Height Changes
+
+Dynamic height changes (expand/collapse sections, showing/hiding content) can cause node layout recalculations and affect connection wire positioning.
+
+- Prefer static layouts over expandable/collapsible sections
+- Use tooltips or overlays for additional information instead
+- If dynamic height is unavoidable, manually trigger layout updates (see Section 4.4)
+
+#### 7.4.3 Keep UI Simple and Intuitive
+
+As internal widgets for ComfyUI custom nodes, the UI should be accessible to users without technical implementation details.
+
+- Use clear, user-friendly terminology (avoid "frontend/backend roll" in favor of "fixed/always randomize")
+- Focus on user intent rather than implementation details
+- Avoid complex interactions that may confuse users
+
+#### 7.4.4 Forward Middle Mouse Events to Canvas
+
+By default, when a DOM widget receives pointer events (e.g., mouse clicks, drags), these events are captured by the widget and not forwarded to the ComfyUI canvas. This prevents users from panning the workflow using the middle mouse button when the cursor is over a DOM widget.
+
+To enable workflow panning over your widget, you should forward middle mouse events (button 1) to the canvas using the `forwardMiddleMouseToCanvas` utility function:
+
+```javascript
+import { forwardMiddleMouseToCanvas } from "./utils.js";
+
+// In your widget creation function
+const container = document.createElement("div");
+container.style.width = "100%";
+container.style.height = "100%";
+// ... other styles ...
+
+// Forward middle mouse events to canvas for panning
+forwardMiddleMouseToCanvas(container);
+
+const widget = node.addDOMWidget(name, type, container, { ... });
+```
+
+The `forwardMiddleMouseToCanvas` function:
+- Forwards `pointerdown` events with button 1 (middle mouse button) to `app.canvas.processMouseDown`
+- Forwards `pointermove` events while middle mouse button is pressed to `app.canvas.processMouseMove`
+- Forwards `pointerup` events with button 1 to `app.canvas.processMouseUp`
+
+This allows users to pan the workflow canvas even when their mouse cursor is hovering over your DOM widget.
+
+---
+
+## 8. Event Handling in Vue DOM Render Mode
+
+ComfyUI frontend supports two rendering modes for nodes:
+- **Legacy Canvas Mode**: Traditional rendering where widgets are rendered on top of the canvas using absolute positioning
+- **Vue DOM Render Mode**: New rendering mode where nodes and widgets are rendered as Vue components
+
+In Vue DOM render mode, event handling works differently. The frontend uses `useCanvasInteractions` composable to manage event forwarding to the canvas. This can cause custom event handlers in your widgets (e.g., mouse wheel for sliders, custom drag operations) to be intercepted by the canvas.
+
+### 8.1 Wheel Event Handling
+
+By default in Vue DOM render mode, wheel events on widgets may be forwarded to the canvas for workflow zoom, overriding your custom wheel handlers (e.g., adjusting slider values with mouse wheel).
+
+To fix this, use the `data-capture-wheel="true"` attribute on elements that should capture wheel events:
+
+```vue
+<!-- Vue component template -->
+<div class="my-slider" data-capture-wheel="true" @wheel="onWheel">
+  <!-- Slider content -->
+</div>
+
+<script setup lang="ts">
+const onWheel = (event: WheelEvent) => {
+  event.preventDefault()
+  // Custom wheel handling logic here
+}
+</script>
+```
+
+**How it works:**
+- ComfyUI's `useCanvasInteractions.ts` checks `target?.closest('[data-capture-wheel="true"]')` before forwarding wheel events
+- If an element (or its ancestor) has this attribute, wheel events are not forwarded to canvas
+- Your custom `@wheel` handler will work as expected
+
+**Granular control:**
+- Apply `data-capture-wheel="true"` to specific interactive elements (e.g., sliders, scrollable areas)
+- Widget container without this attribute will allow workflow zoom when wheel is used elsewhere
+- This allows users to both: adjust widget values with wheel, and zoom workflow with wheel in widget's non-interactive areas
+
+**Example from DualRangeSlider.vue:**
+```vue
+<template>
+  <div
+    class="dual-range-slider"
+    :class="{ disabled, 'is-dragging': dragging !== null }"
+    data-capture-wheel="true"
+    @wheel="onWheel"
+  >
+    <!-- Slider tracks and handles -->
+  </div>
+</template>
+```
+
+### 8.2 Pointer Event Handling
+
+In Vue DOM render mode, pointer events (click, drag, etc.) may also be captured by the canvas system. For custom drag operations:
+
+1. **Use event modifiers to stop propagation:**
+   ```vue
+   <div
+     @pointerdown.stop="startDrag"
+     @pointermove.stop="onDrag"
+     @pointerup.stop="stopDrag"
+   >
+   ```
+
+2. **Use pointer capture for reliable drag tracking:**
+   ```javascript
+   const startDrag = (event: PointerEvent) => {
+     const target = event.currentTarget as HTMLElement
+     target.setPointerCapture(event.pointerId)
+     // ... drag initialization
+   }
+
+   const stopDrag = (event: PointerEvent) => {
+     const target = event.currentTarget as HTMLElement
+     target.releasePointerCapture(event.pointerId)
+     // ... drag cleanup
+   }
+   ```
+
+3. **Use `touch-action: none` CSS for touch devices:**
+   ```css
+   .my-draggable {
+     touch-action: none;
+   }
+   ```
+
+### 8.3 Compatibility Checklist
+
+Ensure your widget works in both rendering modes:
+
+| Feature | Canvas Mode | Vue DOM Mode | Solution |
+|---------|-------------|--------------|----------|
+| Mouse wheel on sliders | Works by default | Needs `data-capture-wheel` | Add `data-capture-wheel="true"` to slider elements |
+| Custom drag operations | Works with `stopPropagation()` | Needs `stopPropagation()` | Use `.stop` modifier and pointer capture |
+| Middle mouse panning | Manual forwarding required | Manual forwarding required | Use `forwardMiddleMouseToCanvas()` |
+| Workflow zoom on widget edges | Works by default | Works by default | No action needed (works by default) |
+
+### 8.4 Testing Recommendations
+
+Test your widget in both rendering modes:
+1. Toggle between Canvas Mode and Vue DOM Mode in ComfyUI settings
+2. Verify custom interactions (wheel, drag, etc.) work in both modes
+3. Verify canvas interactions (zoom, pan) still work when cursor is over non-interactive widget areas
+4. Test with touch devices if applicable
+
+---
+
+## 9. Complete Example: Text Counter
+
+This example implements a simple widget that displays the character count of another text widget in the same node.
+
+```javascript
+import { app } from "../../scripts/app.js";
+
+app.registerExtension({
+    name: "Comfy.TextCounter",
+    getCustomWidgets() {
+        return {
+            TEXT_COUNTER(node, inputName) {
+                const el = document.createElement("div");
+                Object.assign(el.style, {
+                    background: "#222",
+                    border: "1px solid #444",
+                    padding: "8px",
+                    borderRadius: "4px",
+                    fontSize: "12px",
+                    color: "#eee"
+                });
+                
+                const label = document.createElement("span");
+                label.innerText = "Characters: 0";
+                el.appendChild(label);
+
+                const widget = node.addDOMWidget(inputName, "TEXT_COUNTER", el, {
+                    getValue() { return ""; }, // Nothing to save
+                    setValue(v) { },           // Nothing to restore
+                    getMinHeight() { return 40; }
+                });
+                
+                // Disable serialization for this display-only widget
+                widget.serialize = false;
+
+                // Custom method to update UI
+                widget.updateCount = (text) => {
+                    label.innerText = `Characters: ${text.length}`;
+                };
+
+                return { widget };
+            }
+        };
+    },
+    nodeCreated(node) {
+        // Logic to link widgets after the node is initialized
+        if (node.comfyClass === "MyTextNode") {
+            const counterWidget = node.widgets.find(w => w.type === "TEXT_COUNTER");
+            const textWidget = node.widgets.find(w => w.name === "text");
+            
+            if (counterWidget && textWidget) {
+                // Hook into the text widget's callback
+                const oldCallback = textWidget.callback;
+                textWidget.callback = function(v) {
+                    if (oldCallback) oldCallback.apply(this, arguments);
+                    counterWidget.updateCount(v);
+                };
+            }
+        }
+    }
+});
+```

docs/frontend-dom-fixtures.md

@@ -0,0 +1,51 @@
+# Frontend DOM Fixture Strategy
+
+This guide outlines how to reproduce the markup emitted by the Django templates while running Vitest in jsdom.  The aim is to make it straightforward to write integration-style unit tests for managers and UI helpers without having to duplicate template fragments inline.
+
+## Loading Template Markup
+
+Vitest executes inside Node, so we can read the same HTML templates that ship with the extension:
+
+1. Use the helper utilities from `tests/frontend/utils/domFixtures.js` to read files under the `templates/` directory.
+2. Mount the returned markup into `document.body` (or any custom container) before importing the module under test so its query selectors resolve correctly.
+
+```js
+import { renderTemplate } from '../utils/domFixtures.js'; // adjust the relative path to your spec
+
+beforeEach(() => {
+  renderTemplate('loras.html', {
+    dataset: { page: 'loras' }
+  });
+});
+```
+
+The helper ensures the dataset is applied to the container, which mirrors how Django sets `data-page` in production.
+
+## Working with Partial Components
+
+Many features are implemented as template partials located under `templates/components/`.  When a test only needs a fragment (for example, the progress panel or context menu markup), load the component file directly:
+
+```js
+const container = renderTemplate('components/progress_panel.html');
+
+const progressPanel = container.querySelector('#progress-panel');
+```
+
+This pattern avoids hand-written fixture strings and keeps the tests aligned with the actual markup.
+
+## Resetting Between Tests
+
+The shared Vitest setup clears `document.body` and storage APIs before each test.  If a suite adds additional DOM nodes outside of the body or needs to reset custom attributes mid-test, use `resetDom()` exported from `domFixtures.js`.
+
+```js
+import { resetDom } from '../utils/domFixtures.js';
+
+afterEach(() => {
+  resetDom();
+});
+```
+
+## Future Enhancements
+
+- Provide typed helpers for injecting mock script tags (e.g., replicating ComfyUI globals).
+- Compose higher-level fixtures that mimic specific pages (loras, checkpoints, recipes) once those managers receive dedicated suites.

docs/frontend-filtering-test-matrix.md

@@ -0,0 +1,44 @@
+# LoRA & Checkpoints Filtering/Sorting Test Matrix
+
+This matrix captures the scenarios that Phase 3 frontend tests should cover for the LoRA and Checkpoint managers. It focuses on how search, filter, sort, and duplicate badge toggles interact so future specs can share fixtures and expectations.
+
+## Scope
+
+- **Components**: `PageControls`, `FilterManager`, `SearchManager`, and `ModelDuplicatesManager` wiring invoked through `CheckpointsPageManager` and `LorasPageManager`.
+- **Templates**: `templates/loras.html` and `templates/checkpoints.html` along with shared filter panel and toolbar partials.
+- **APIs**: Requests issued through `baseModelApi.fetchModels` (via `resetAndReload`/`refreshModels`) and duplicates badge updates.
+
+## Shared Setup Considerations
+
+1. Render full page templates using `renderLorasPage` / `renderCheckpointsPage` helpers before importing modules so DOM queries resolve.
+2. Stub storage helpers (`getStorageItem`, `setStorageItem`, `getSessionItem`, `setSessionItem`) to observe persistence behavior without mutating real storage.
+3. Mock `sidebarManager` to capture refresh calls triggered after sort/filter actions.
+4. Provide fake API implementations exposing `resetAndReload`, `refreshModels`, `fetchFromCivitai`, `toggleBulkMode`, and `clearCustomFilter` so control events remain asynchronous but deterministic.
+5. Supply a minimal `ModelDuplicatesManager` mock exposing `toggleDuplicateMode`, `checkDuplicatesCount`, and `updateDuplicatesBadgeAfterRefresh` to validate duplicate badge wiring.
+
+## Scenario Matrix
+
+| ID | Feature | Scenario | LoRAs Expectations | Checkpoints Expectations | Notes |
+| --- | --- | --- | --- | --- | --- |
+| F-01 | Search filter | Typing a query updates `pageState.filters.search`, persists to session, and triggers `resetAndReload` on submit | Validate `SearchManager` writes query and reloads via API stub; confirm LoRA cards pass query downstream | Same as LoRAs | Cover `enter` press and clicking search icon |
+| F-02 | Tag filter | Selecting a tag chip cycles include ➜ exclude ➜ clear, updates storage, and reloads results | Tag state stored under `filters.tags[tagName] = 'include'|'exclude'`; `FilterManager.applyFilters` persists and triggers `resetAndReload(true)` | Same; ensure base model tag set is scoped to checkpoints dataset | Include removal path |
+| F-03 | Base model filter | Toggling base model checkboxes updates `filters.baseModel`, persists, and reloads | Ensure only LoRA-supported models show; toggle multi-select | Ensure SDXL/Flux base models appear as expected | Capture UI state restored from storage on next init |
+| F-04 | Favorites-only | Clicking favorites toggle updates session flag and calls `resetAndReload(true)` | Button gains `.active` class and API called | Same | Verify duplicates badge refresh when active |
+| F-05 | Sort selection | Changing sort select saves preference (legacy + new format) and reloads | Confirm `PageControls.saveSortPreference` invoked with option and API called | Same with checkpoints-specific defaults | Cover `convertLegacySortFormat` branch |
+| F-06 | Filter persistence | Re-initializing manager loads stored filters/sort and updates DOM | Filters pre-populate chips/checkboxes; favorites state restored | Same | Requires simulating repeated construction |
+| F-07 | Combined filters | Applying search + tag + base model yields aggregated query params for fetch | Assert API receives merged filter payload | Same | Validate toast messaging for active filters |
+| F-08 | Clearing filters | Using "Clear filters" resets state, storage, and reloads list | `FilterManager.clearFilters` empties `filters`, removes active class, shows toast | Same | Ensure favorites-only toggle unaffected |
+| F-09 | Duplicate badge toggle | Pressing "Find duplicates" toggles duplicate mode and updates badge counts post-refresh | `ModelDuplicatesManager.toggleDuplicateMode` invoked and badge refresh called after API rebuild | Same plus checkpoint-specific duplicate badge dataset | Connects to future duplicate-specific specs |
+| F-10 | Bulk actions menu | Opening bulk dropdown keeps filters intact and closes on outside click | Validate dropdown class toggling and no unintended reload | Same | Guard against regression when dropdown interacts with filters |
+
+## Automation Coverage Status
+
+- ✅ F-01 Search filter, F-02 Tag filter, F-03 Base model filter, F-04 Favorites-only toggle, F-05 Sort selection, and F-09 Duplicate badge toggle are covered by `tests/frontend/components/pageControls.filtering.test.js` for both LoRA and checkpoint pages.
+- ⏳ F-06 Filter persistence, F-07 Combined filters, F-08 Clearing filters, and F-10 Bulk actions remain to be automated alongside upcoming bulk mode refinements.
+
+## Coverage Gaps & Follow-Ups
+
+- Write Vitest suites that exercise the matrix for both managers, sharing fixtures through page helpers to avoid duplication.
+- Capture API parameter assertions by inspecting `baseModelApi.fetchModels` mocks rather than relying solely on state mutations.
+- Add regression cases for legacy storage migrations (old filter keys) once fixtures exist for older payloads.
+- Extend duplicate badge coverage with scenarios where `checkDuplicatesCount` signals zero duplicates versus pending calculations.

docs/frontend-testing-roadmap.md

@@ -0,0 +1,33 @@
+# Frontend Automation Testing Roadmap
+
+This roadmap tracks the planned rollout of automated testing for the ComfyUI LoRA Manager frontend. Each phase builds on the infrastructure introduced in this change set and records progress so future contributors can quickly identify the next tasks.
+
+## Phase Overview
+
+| Phase | Goal | Primary Focus | Status | Notes |
+| --- | --- | --- | --- | --- |
+| Phase 0 | Establish baseline tooling | Add Node test runner, jsdom environment, and seed smoke tests | ✅ Complete | Vitest + jsdom configured, example state tests committed |
+| Phase 1 | Cover state management logic | Unit test selectors, derived data helpers, and storage utilities under `static/js/state` and `static/js/utils` | ✅ Complete | Storage helpers and state selectors now exercised via deterministic suites |
+| Phase 2 | Test AppCore orchestration | Simulate page bootstrapping, infinite scroll hooks, and manager registration using JSDOM DOM fixtures | ✅ Complete | AppCore initialization + page feature suites now validate manager wiring, infinite scroll hooks, and onboarding gating |
+| Phase 3 | Validate page-specific managers | Add focused suites for `loras`, `checkpoints`, `embeddings`, and `recipes` managers covering filtering, sorting, and bulk actions | ✅ Complete | LoRA/checkpoint suites expanded; embeddings + recipes managers now covered with initialization, filtering, and duplicate workflows |
+| Phase 4 | Interaction-level regression tests | Exercise template fragments, modals, and menus to ensure UI wiring remains intact | ✅ Complete | Vitest DOM suites cover NSFW selector, recipe modal editing, and global context menus |
+| Phase 5 | Continuous integration & coverage | Integrate frontend tests into CI workflow and track coverage metrics | ✅ Complete | CI workflow runs Vitest and aggregates V8 coverage into `coverage/frontend` via a dedicated script |
+
+## Next Steps Checklist
+
+- [x] Expand unit tests for `storageHelpers` covering migrations and namespace behavior.
+- [x] Document DOM fixture strategy for reproducing template structures in tests.
+- [x] Prototype AppCore initialization test that verifies manager bootstrapping with stubbed dependencies.
+- [x] Add AppCore page feature suite exercising context menu creation and infinite scroll registration via DOM fixtures.
+- [x] Extend AppCore orchestration tests to cover manager wiring, bulk menu setup, and onboarding gating scenarios.
+- [x] Add interaction regression suites for context menus and recipe modals to complete Phase 4.
+- [x] Evaluate integrating coverage reporting once test surface grows (> 20 specs).
+- [x] Create shared fixtures for the loras and checkpoints pages once dedicated manager suites are added.
+- [x] Draft focused test matrix for loras/checkpoints manager filtering and sorting paths ahead of Phase 3.
+- [x] Implement LoRAs manager filtering/sorting specs for scenarios F-01–F-05 & F-09; queue remaining edge cases after duplicate/bulk flows stabilize.
+- [x] Implement checkpoints manager filtering/sorting specs for scenarios F-01–F-05 & F-09; cover remaining paths alongside bulk action work.
+- [x] Implement checkpoints page manager smoke tests covering initialization and duplicate badge wiring.
+- [x] Outline focused checkpoints scenarios (filtering, sorting, duplicate badge toggles) to feed into the shared test matrix.
+- [ ] Add duplicate badge regression coverage for zero/pending states after API refreshes.
+
+Maintaining this roadmap alongside code changes will make it easier to append new automated test tasks and update their progress.

docs/library-switching.md

@@ -0,0 +1,28 @@
+# Library Switching and Preview Routes
+
+Library switching no longer requires restarting the backend. The preview
+thumbnails shown in the UI are now served through a dynamic endpoint that
+resolves files against the folders registered for the active library at request
+time. This allows the multi-library flow to update model roots without touching
+the aiohttp router, so previews remain available immediately after a switch.
+
+## How the dynamic preview endpoint works
+
+* `config.get_preview_static_url()` now returns `/api/lm/previews?path=<encoded>`
+  for any preview path. The raw filesystem location is URL encoded so that it
+  can be passed through the query string without leaking directory structure in
+  the route itself.【F:py/config.py†L398-L404】
+* `PreviewRoutes` exposes the `/api/lm/previews` handler which validates the
+  decoded path against the directories registered for the current library. The
+  request is rejected if it falls outside those roots or if the file does not
+  exist.【F:py/routes/preview_routes.py†L5-L21】【F:py/routes/handlers/preview_handlers.py†L9-L48】
+* `Config` keeps an up-to-date cache of allowed preview roots. Every time a
+  library is applied the cache is rebuilt using the declared LoRA, checkpoint
+  and embedding directories (including symlink targets). The validation logic
+  checks preview requests against this cache.【F:py/config.py†L51-L68】【F:py/config.py†L180-L248】【F:py/config.py†L332-L346】
+
+Both the ComfyUI runtime (`LoraManager.add_routes`) and the standalone launcher
+(`StandaloneLoraManager.add_routes`) register the new preview routes instead of
+mounting a static directory per root. Switching libraries therefore works
+without restarting the application, and preview URLs generated before or after a
+switch continue to resolve correctly.【F:py/lora_manager.py†L21-L82】【F:standalone.py†L302-L315】

docs/priority_tags_help.md

@@ -0,0 +1,71 @@
+# Priority Tags Configuration Guide
+
+This guide explains how to tailor the tag priority order that powers folder naming and tag suggestions in the LoRA Manager. You only need to edit the comma-separated list of entries shown in the **Priority Tags** field for each model type.
+
+## 1. Pick the Model Type
+
+In the **Priority Tags** dialog you will find one tab per model type (LoRA, Checkpoint, Embedding). Select the tab you want to update; changes on one tab do not affect the others.
+
+## 2. Edit the Entry List
+
+Inside the textarea you will see a line similar to:
+
+```
+character, concept, style(toon|toon_style)
+```
+
+This entire line is the **entry list**. Replace it with your own ordered list.
+
+### Entry Rules
+
+Each entry is separated by a comma, in order from highest to lowest priority:
+
+- **Canonical tag only:** `realistic`
+- **Canonical tag with aliases:** `character(char|chars)`
+
+Aliases live inside `()` and are separated with `|`. The canonical name is what appears in folder names and UI suggestions when any of the aliases are detected. Matching is case-insensitive.
+
+## Use `{first_tag}` in Path Templates
+
+When your path template contains `{first_tag}`, the app picks a folder name based on your priority list and the model’s own tags:
+
+- It checks the priority list from top to bottom. If a canonical tag or any of its aliases appear in the model tags, that canonical name becomes the folder name.
+- If no priority tags are found but the model has tags, the very first model tag is used.
+- If the model has no tags at all, the folder falls back to `no tags`.
+
+### Example
+
+With a template like `/{model_type}/{first_tag}` and the priority entry list `character(char|chars), style(anime|toon)`:
+
+| Model Tags | Folder Name | Why |
+| --- | --- | --- |
+| `["chars", "female"]` | `character` | `chars` matches the `character` alias, so the canonical wins. |
+| `["anime", "portrait"]` | `style` | `anime` hits the `style` entry, so its canonical label is used. |
+| `["portrait", "bw"]` | `portrait` | No priority match, so the first model tag is used. |
+| `[]` | `no tags` | Nothing to match, so the fallback is applied. |
+
+## 3. Save the Settings
+
+After editing the entry list, press **Enter** to save. Use **Shift+Enter** whenever you need a new line. Clicking outside the field also saves automatically. A success toast confirms the update.
+
+## Examples
+
+| Goal | Entry List |
+| --- | --- |
+| Prefer people over styles | `character, portraits, style(anime\|toon)` |
+| Group sci-fi variants | `sci-fi(scifi\|science_fiction), cyberpunk(cyber\|punk)` |
+| Alias shorthand tags | `realistic(real\|realisim), photorealistic(photo_real)` |
+
+## Tips
+
+- Keep canonical names short and meaningful—they become folder names.
+- Place the most important categories first; the first match wins.
+- Avoid duplicate canonical names within the same list; only the first instance is used.
+
+## Troubleshooting
+
+- **Unexpected folder name?** Check that the canonical name you want is placed before other matches.
+- **Alias not working?** Ensure the alias is inside parentheses and separated with `|`, e.g. `character(char|chars)`.
+- **Validation error?** Look for missing parentheses or stray commas. Each entry must follow the `canonical(alias|alias)` pattern or just `canonical`.
+
+With these basics you can quickly adapt Priority Tags to match your library’s organization style.

docs/reference/danbooru_e621_categories.md

@@ -0,0 +1,69 @@
+# Danbooru/E621 Tag Categories Reference
+
+Reference for category values used in `danbooru_e621_merged.csv` tag files.
+
+## Category Value Mapping
+
+### Danbooru Categories
+
+| Value | Description |
+|-------|-------------|
+| 0 | General |
+| 1 | Artist |
+| 2 | *(unused)* |
+| 3 | Copyright |
+| 4 | Character |
+| 5 | Meta |
+
+### e621 Categories
+
+| Value | Description |
+|-------|-------------|
+| 6 | *(unused)* |
+| 7 | General |
+| 8 | Artist |
+| 9 | Contributor |
+| 10 | Copyright |
+| 11 | Character |
+| 12 | Species |
+| 13 | *(unused)* |
+| 14 | Meta |
+| 15 | Lore |
+
+## Danbooru Category Colors
+
+| Description | Normal Color | Hover Color |
+|-------------|--------------|-------------|
+| General | #009be6 | #4bb4ff |
+| Artist | #ff8a8b | #ffc3c3 |
+| Copyright | #c797ff | #ddc9fb |
+| Character | #35c64a | #93e49a |
+| Meta | #ead084 | #f7e7c3 |
+
+## CSV Column Structure
+
+Each row in the merged CSV file contains 4 columns:
+
+| Column | Description | Example |
+|--------|-------------|---------|
+| 1 | Tag name | `1girl`, `highres`, `solo` |
+| 2 | Category value (0-15) | `0`, `5`, `7` |
+| 3 | Post count | `6008644`, `5256195` |
+| 4 | Aliases (comma-separated, quoted) | `"1girls,sole_female"`, empty string |
+
+### Sample Data
+
+```
+1girl,0,6008644,"1girls,sole_female"
+highres,5,5256195,"high_res,high_resolution,hires"
+solo,0,5000954,"alone,female_solo,single,solo_female"
+long_hair,0,4350743,"/lh,longhair"
+mammal,12,3437444,"cetancodont,cetancodontamorph,feralmammal"
+anthro,7,3381927,"adult_anthro,anhtro,antho,anthro_horse"
+skirt,0,1557883,
+```
+
+## Source
+
+- [PR #312: Add danbooru_e621_merged.csv](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/pull/312)
+- [DraconicDragon/dbr-e621-lists-archive](https://github.com/DraconicDragon/dbr-e621-lists-archive)

docs/technical/model_type_refactoring_todo.md

@@ -0,0 +1,191 @@
+# Model Type 字段重构 - 遗留工作清单
+
+> **状态**: Phase 1-4 已完成 | **创建日期**: 2026-01-30  
+> **相关文件**: `py/utils/models.py`, `py/services/model_query.py`, `py/services/checkpoint_scanner.py`, etc.
+
+---
+
+## 概述
+
+本次重构旨在解决 `model_type` 字段语义不统一的问题。系统中有两个层面的"类型"概念：
+
+1. **Scanner Type** (`scanner_type`): 架构层面的大类 - `lora`, `checkpoint`, `embedding`
+2. **Sub Type** (`sub_type`): 业务层面的细分类型 - `lora`/`locon`/`dora`, `checkpoint`/`diffusion_model`, `embedding`
+
+重构目标是统一使用 `sub_type` 表示细分类型，保留 `model_type` 作为向后兼容的别名。
+
+---
+
+## 已完成工作 ✅
+
+### Phase 1: 后端字段重命名
+- [x] `CheckpointMetadata.model_type` → `sub_type`
+- [x] `EmbeddingMetadata.model_type` → `sub_type`
+- [x] `model_scanner.py` `_build_cache_entry()` 同时处理 `sub_type` 和 `model_type`
+
+### Phase 2: 查询逻辑更新
+- [x] `model_query.py` 新增 `resolve_sub_type()` 和 `normalize_sub_type()`
+- [x] ~~保持向后兼容的别名 `resolve_civitai_model_type`, `normalize_civitai_model_type`~~ (已在 Phase 5 移除)
+- [x] `ModelFilterSet.apply()` 更新为使用新的解析函数
+
+### Phase 3: API 响应更新
+- [x] `LoraService.format_response()` 返回 `sub_type` ~~+ `model_type`~~ (已移除 `model_type`)
+- [x] `CheckpointService.format_response()` 返回 `sub_type` ~~+ `model_type`~~ (已移除 `model_type`)
+- [x] `EmbeddingService.format_response()` 返回 `sub_type` ~~+ `model_type`~~ (已移除 `model_type`)
+
+### Phase 4: 前端更新
+- [x] `constants.js` 新增 `MODEL_SUBTYPE_DISPLAY_NAMES`
+- [x] `MODEL_TYPE_DISPLAY_NAMES` 作为别名保留
+
+### Phase 5: 清理废弃代码 ✅
+- [x] 从 `ModelScanner._build_cache_entry()` 中移除 `model_type` 向后兼容代码
+- [x] 从 `CheckpointScanner` 中移除 `model_type` 兼容处理
+- [x] 从 `model_query.py` 中移除 `resolve_civitai_model_type` 和 `normalize_civitai_model_type` 别名
+- [x] 更新前端 `FilterManager.js` 使用 `sub_type` (已在使用 `MODEL_SUBTYPE_DISPLAY_NAMES`)
+- [x] 更新所有相关测试
+
+---
+
+## 遗留工作 ⏳
+
+### Phase 5: 清理废弃代码 ✅ **已完成**
+
+所有 Phase 5 的清理工作已完成：
+
+#### 5.1 移除 `model_type` 字段的向后兼容代码 ✅
+- 从 `ModelScanner._build_cache_entry()` 中移除了 `model_type` 的设置
+- 现在只设置 `sub_type` 字段
+
+#### 5.2 移除 CheckpointScanner 的 model_type 兼容处理 ✅
+- `adjust_metadata()` 现在只处理 `sub_type`
+- `adjust_cached_entry()` 现在只设置 `sub_type`
+
+#### 5.3 移除 model_query 中的向后兼容别名 ✅
+- 移除了 `resolve_civitai_model_type = resolve_sub_type`
+- 移除了 `normalize_civitai_model_type = normalize_sub_type`
+
+#### 5.4 前端清理 ✅
+- `FilterManager.js` 已经在使用 `MODEL_SUBTYPE_DISPLAY_NAMES` (通过别名 `MODEL_TYPE_DISPLAY_NAMES`)
+- API list endpoint 现在只返回 `sub_type`，不再返回 `model_type`
+- `ModelCard.js` 现在设置 `card.dataset.sub_type` (所有模型类型通用)
+- `CheckpointContextMenu.js` 现在读取 `card.dataset.sub_type`
+- `MoveManager.js` 现在处理 `cache_entry.sub_type`
+- `RecipeModal.js` 现在读取 `checkpoint.sub_type`
+
+---
+
+## 数据库迁移评估
+
+### 当前状态
+- `persistent_model_cache.py` 使用 `civitai_model_type` 列存储 CivitAI 原始类型
+- 缓存 entry 中的 `sub_type` 在运行期动态计算
+- 数据库 schema **无需立即修改**
+
+### 未来可选优化
+```sql
+-- 可选：在 models 表中添加 sub_type 列（与 civitai_model_type 保持一致但语义更清晰）
+ALTER TABLE models ADD COLUMN sub_type TEXT;
+
+-- 数据迁移
+UPDATE models SET sub_type = civitai_model_type WHERE sub_type IS NULL;
+```
+
+**建议**: 如果决定添加 `sub_type` 列，应与 Phase 5 一起进行。
+
+---
+
+## 测试覆盖率
+
+### 新增/更新测试文件（已全部通过 ✅）
+
+| 测试文件 | 数量 | 覆盖内容 |
+|---------|------|---------|
+| `tests/utils/test_models_sub_type.py` | 7 | Metadata sub_type 字段 |
+| `tests/services/test_model_query_sub_type.py` | 19 | sub_type 解析和过滤 |
+| `tests/services/test_checkpoint_scanner_sub_type.py` | 6 | CheckpointScanner sub_type |
+| `tests/services/test_service_format_response_sub_type.py` | 6 | API 响应 sub_type 包含 |
+| `tests/services/test_checkpoint_scanner.py` | 1 | Checkpoint 缓存 sub_type |
+| `tests/services/test_model_scanner.py` | 1 | adjust_cached_entry hook |
+| `tests/services/test_download_manager.py` | 1 | Checkpoint 下载 sub_type |
+
+### 需要补充的测试（可选）
+
+- [ ] 集成测试：验证前端过滤使用 sub_type 字段
+- [ ] 数据库迁移测试（如果执行可选优化）
+- [ ] 性能测试：确认 resolve_sub_type 的优先级查找没有显著性能影响
+
+---
+
+## 兼容性检查清单
+
+### 已完成 ✅
+
+- [x] 前端代码已全部改用 `sub_type` 字段
+- [x] API list endpoint 已移除 `model_type`，只返回 `sub_type`
+- [x] 后端 cache entry 已移除 `model_type`，只保留 `sub_type`
+- [x] 所有测试已更新通过
+- [x] 文档已更新
+
+---
+
+## 相关文件清单
+
+### 核心文件
+```
+py/utils/models.py
+py/utils/constants.py
+py/services/model_scanner.py
+py/services/model_query.py
+py/services/checkpoint_scanner.py
+py/services/base_model_service.py
+py/services/lora_service.py
+py/services/checkpoint_service.py
+py/services/embedding_service.py
+```
+
+### 前端文件
+```
+static/js/utils/constants.js
+static/js/managers/FilterManager.js
+static/js/managers/MoveManager.js
+static/js/components/shared/ModelCard.js
+static/js/components/ContextMenu/CheckpointContextMenu.js
+static/js/components/RecipeModal.js
+```
+
+### 测试文件
+```
+tests/utils/test_models_sub_type.py
+tests/services/test_model_query_sub_type.py
+tests/services/test_checkpoint_scanner_sub_type.py
+tests/services/test_service_format_response_sub_type.py
+```
+
+---
+
+## 风险评估
+
+| 风险项 | 影响 | 缓解措施 |
+|-------|------|---------|
+| ~~第三方代码依赖 `model_type`~~ | ~~高~~ | ~~保持别名至少 1 个 major 版本~~ ✅ 已完成移除 |
+| ~~数据库 schema 变更~~ | ~~中~~ | ~~暂缓 schema 变更，仅运行时计算~~ ✅ 无需变更 |
+| ~~前端过滤失效~~ | ~~中~~ | ~~全面的集成测试覆盖~~ ✅ 测试通过 |
+| CivitAI API 变化 | 低 | 保持多源解析策略 |
+
+---
+
+## 时间线
+
+- **v1.x**: Phase 1-4 已完成，保持向后兼容
+- **v2.0 (当前)**: ✅ Phase 5 已完成 - `model_type` 兼容代码已移除
+  - API list endpoint 只返回 `sub_type`
+  - Cache entry 只保留 `sub_type`
+  - 移除了 `resolve_civitai_model_type` 和 `normalize_civitai_model_type` 别名
+
+---
+
+## 备注
+
+- 重构期间发现 `civitai_model_type` 数据库列命名尚可，但语义上应理解为存储 CivitAI API 返回的原始类型值
+- Checkpoint 的 `diffusion_model` sub_type 不能通过 CivitAI API 获取，必须通过文件路径（model root）判断
+- LoRA 的 sub_type（lora/locon/dora）直接来自 CivitAI API 的 `version_info.model.type`

docs/testing/coverage_analysis.md

@@ -0,0 +1,26 @@
+# Backend Test Coverage Notes
+
+## Pytest Execution
+- Command: `python -m pytest`
+- Result: All 283 collected tests passed in the current environment.
+- Coverage tooling (``pytest-cov``/``coverage``) is unavailable in the offline sandbox, so line-level metrics could not be generated. The earlier attempt to install ``pytest-cov`` failed because the package index cannot be reached from the container.
+
+## High-Priority Gaps to Address
+
+### 1. Standalone server bootstrapping
+* **Source:** [`standalone.py`](../../standalone.py)
+* **Why it matters:** The standalone entry point wires together the aiohttp application, static asset routes, model-route registration, and configuration validation. None of these behaviours are covered by automated tests, leaving regressions in bootstrapping logic undetected.
+* **Suggested coverage:** Add integration-style tests that instantiate `StandaloneServer`/`StandaloneLoraManager` with temporary settings and assert that routes (HTTP + websocket) are registered, configuration warnings fire for missing paths, and the mock ComfyUI shims behave as expected.
+
+### 2. Model service registration factory
+* **Source:** [`py/services/model_service_factory.py`](../../py/services/model_service_factory.py)
+* **Why it matters:** The factory coordinates which model services and routes the API exposes, including error handling when unknown model types are requested. No current tests verify registration, memoization of route instances, or the logging path on failures.
+* **Suggested coverage:** Unit tests that exercise `register_model_type`, `get_route_instance`, error branches in `get_service_class`/`get_route_class`, and `setup_all_routes` when a route setup raises. Use lightweight fakes to confirm the logger is called and state is cleared via `clear_registrations`.
+
+### 3. Server-side i18n helper
+* **Source:** [`py/services/server_i18n.py`](../../py/services/server_i18n.py)
+* **Why it matters:** Template rendering relies on the `ServerI18nManager` to load locale JSON, perform key lookups, and format parameters. The fallback logic (dot-notation lookup, English fallbacks, placeholder substitution) is untested, so malformed locale files or regressions in placeholder handling would slip through.
+* **Suggested coverage:** Tests that load fixture locale dictionaries, assert `set_locale` fallbacks, verify nested key resolution and placeholder substitution, and ensure missing keys return the original identifier.
+
+## Next Steps
+Prioritize creating focused unit tests around these modules, then re-run pytest once coverage tooling is available to confirm the new tests close the identified gaps.

package.json

@@ -0,0 +1,17 @@
+{
+  "name": "comfyui-lora-manager-frontend",
+  "version": "0.1.0",
+  "private": true,
+  "type": "module",
+  "scripts": {
+    "test": "npm run test:js && npm run test:vue",
+    "test:js": "vitest run",
+    "test:vue": "cd vue-widgets && npx vitest run",
+    "test:watch": "vitest",
+    "test:coverage": "node scripts/run_frontend_coverage.js"
+  },
+  "devDependencies": {
+    "jsdom": "^24.0.0",
+    "vitest": "^1.6.0"
+  }
+}

py/__init__.py

@@ -0,0 +1,12 @@
+"""Project namespace package."""
+
+# pytest's internal compatibility layer still imports ``py.path.local`` from the
+# historical ``py`` dependency.  Because this project reuses the ``py`` package
+# name, we expose a minimal shim so ``py.path.local`` resolves to ``pathlib.Path``
+# during test runs without pulling in the external dependency.
+from pathlib import Path
+from types import SimpleNamespace
+
+path = SimpleNamespace(local=Path)
+
+__all__ = ["path"]

py/config.py

@@ -1,24 +1,211 @@
 import os
 import platform
+import threading
+from pathlib import Path
 import folder_paths # type: ignore
-from typing import List
+from typing import Any, Dict, Iterable, List, Mapping, Optional, Set, Tuple
 import logging
+import json
+import urllib.parse
+import time
+
+from .utils.cache_paths import CacheType, get_cache_file_path, get_legacy_cache_paths
+from .utils.settings_paths import ensure_settings_file, get_settings_dir, load_settings_template
+
+# Use an environment variable to control standalone mode
+standalone_mode = os.environ.get("LORA_MANAGER_STANDALONE", "0") == "1" or os.environ.get("HF_HUB_DISABLE_TELEMETRY", "0") == "0"
 
 logger = logging.getLogger(__name__)
 
+
+def _normalize_folder_paths_for_comparison(
+    folder_paths: Mapping[str, Iterable[str]]
+) -> Dict[str, Set[str]]:
+    """Normalize folder paths for comparison across libraries."""
+
+    normalized: Dict[str, Set[str]] = {}
+    for key, values in folder_paths.items():
+        if isinstance(values, str):
+            candidate_values: Iterable[str] = [values]
+        else:
+            try:
+                candidate_values = iter(values)
+            except TypeError:
+                continue
+
+        normalized_values: Set[str] = set()
+        for value in candidate_values:
+            if not isinstance(value, str):
+                continue
+            stripped = value.strip()
+            if not stripped:
+                continue
+            normalized_values.add(os.path.normcase(os.path.normpath(stripped)))
+
+        if normalized_values:
+            normalized[key] = normalized_values
+
+    return normalized
+
+
+def _normalize_library_folder_paths(
+    library_payload: Mapping[str, Any]
+) -> Dict[str, Set[str]]:
+    """Return normalized folder paths extracted from a library payload."""
+
+    folder_paths = library_payload.get("folder_paths")
+    if isinstance(folder_paths, Mapping):
+        return _normalize_folder_paths_for_comparison(folder_paths)
+    return {}
+
+
+def _get_template_folder_paths() -> Dict[str, Set[str]]:
+    """Return normalized folder paths defined in the bundled template."""
+
+    template_payload = load_settings_template()
+    if not template_payload:
+        return {}
+
+    folder_paths = template_payload.get("folder_paths")
+    if isinstance(folder_paths, Mapping):
+        return _normalize_folder_paths_for_comparison(folder_paths)
+    return {}
+
+
 class Config:
     """Global configuration for LoRA Manager"""
     
     def __init__(self):
         self.templates_path = os.path.join(os.path.dirname(os.path.dirname(__file__)), 'templates')
         self.static_path = os.path.join(os.path.dirname(os.path.dirname(__file__)), 'static')
-        # 路径映射字典, target to link mapping
-        self._path_mappings = {}
-        # 静态路由映射字典, target to route mapping
-        self._route_mappings = {}
+        self.i18n_path = os.path.join(os.path.dirname(os.path.dirname(__file__)), 'locales')
+        # Path mapping dictionary, target to link mapping
+        self._path_mappings: Dict[str, str] = {}
+        # Normalized preview root directories used to validate preview access
+        self._preview_root_paths: Set[Path] = set()
+        # Fingerprint of the symlink layout from the last successful scan
+        self._cached_fingerprint: Optional[Dict[str, object]] = None
         self.loras_roots = self._init_lora_paths()
-        # 在初始化时扫描符号链接
-        self._scan_symbolic_links()
+        self.checkpoints_roots = None
+        self.unet_roots = None
+        self.embeddings_roots = None
+        self.base_models_roots = self._init_checkpoint_paths()
+        self.embeddings_roots = self._init_embedding_paths()
+        # Scan symbolic links during initialization
+        self._initialize_symlink_mappings()
+        
+        if not standalone_mode:
+            # Save the paths to settings.json when running in ComfyUI mode
+            self.save_folder_paths_to_settings()
+
+    def save_folder_paths_to_settings(self):
+        """Persist ComfyUI-derived folder paths to the multi-library settings."""
+        try:
+            ensure_settings_file(logger)
+            from .services.settings_manager import get_settings_manager
+
+            settings_service = get_settings_manager()
+            libraries = settings_service.get_libraries()
+            comfy_library = libraries.get("comfyui", {})
+            default_library = libraries.get("default", {})
+
+            template_folder_paths = _get_template_folder_paths()
+            default_library_paths: Dict[str, Set[str]] = {}
+            if isinstance(default_library, Mapping):
+                default_library_paths = _normalize_library_folder_paths(default_library)
+
+            libraries_changed = False
+            if (
+                isinstance(default_library, Mapping)
+                and template_folder_paths
+                and default_library_paths == template_folder_paths
+            ):
+                if "comfyui" in libraries:
+                    try:
+                        settings_service.delete_library("default")
+                        libraries_changed = True
+                        logger.info("Removed template 'default' library entry")
+                    except Exception as delete_error:
+                        logger.debug(
+                            "Failed to delete template 'default' library: %s",
+                            delete_error,
+                        )
+                else:
+                    try:
+                        settings_service.rename_library("default", "comfyui")
+                        libraries_changed = True
+                        logger.info("Renamed template 'default' library to 'comfyui'")
+                    except Exception as rename_error:
+                        logger.debug(
+                            "Failed to rename template 'default' library: %s",
+                            rename_error,
+                        )
+
+            if libraries_changed:
+                libraries = settings_service.get_libraries()
+                comfy_library = libraries.get("comfyui", {})
+                default_library = libraries.get("default", {})
+
+            target_folder_paths = {
+                'loras': list(self.loras_roots),
+                'checkpoints': list(self.checkpoints_roots or []),
+                'unet': list(self.unet_roots or []),
+                'embeddings': list(self.embeddings_roots or []),
+            }
+
+            normalized_target_paths = _normalize_folder_paths_for_comparison(target_folder_paths)
+
+            normalized_default_paths: Optional[Dict[str, Set[str]]] = None
+            if isinstance(default_library, Mapping):
+                normalized_default_paths = _normalize_library_folder_paths(default_library)
+
+            if (
+                not comfy_library
+                and default_library
+                and normalized_target_paths
+                and normalized_default_paths == normalized_target_paths
+            ):
+                try:
+                    settings_service.rename_library("default", "comfyui")
+                    logger.info("Renamed legacy 'default' library to 'comfyui'")
+                    libraries = settings_service.get_libraries()
+                    comfy_library = libraries.get("comfyui", {})
+                except Exception as rename_error:
+                    logger.debug(
+                        "Failed to rename legacy 'default' library: %s", rename_error
+                    )
+
+            default_lora_root = comfy_library.get("default_lora_root", "")
+            if not default_lora_root and len(self.loras_roots) == 1:
+                default_lora_root = self.loras_roots[0]
+
+            default_checkpoint_root = comfy_library.get("default_checkpoint_root", "")
+            if (not default_checkpoint_root and self.checkpoints_roots and
+                    len(self.checkpoints_roots) == 1):
+                default_checkpoint_root = self.checkpoints_roots[0]
+
+            default_embedding_root = comfy_library.get("default_embedding_root", "")
+            if (not default_embedding_root and self.embeddings_roots and
+                    len(self.embeddings_roots) == 1):
+                default_embedding_root = self.embeddings_roots[0]
+
+            metadata = dict(comfy_library.get("metadata", {}))
+            metadata.setdefault("display_name", "ComfyUI")
+            metadata["source"] = "comfyui"
+
+            settings_service.upsert_library(
+                "comfyui",
+                folder_paths=target_folder_paths,
+                default_lora_root=default_lora_root,
+                default_checkpoint_root=default_checkpoint_root,
+                default_embedding_root=default_embedding_root,
+                metadata=metadata,
+                activate=True,
+            )
+
+            logger.info("Updated 'comfyui' library with current folder paths")
+        except Exception as e:
+            logger.warning(f"Failed to save folder paths: {e}")
 
     def _is_link(self, path: str) -> bool:
         try:
@@ -37,85 +224,674 @@ class Config:
             logger.error(f"Error checking link status for {path}: {e}")
             return False
 
-    def _scan_symbolic_links(self):
-        """扫描所有 LoRA 根目录中的符号链接"""
-        for root in self.loras_roots:
-            self._scan_directory_links(root)
+    def _entry_is_symlink(self, entry: os.DirEntry) -> bool:
+        """Check if a directory entry is a symlink, including Windows junctions."""
+        if entry.is_symlink():
+            return True
+        if platform.system() == 'Windows':
+            try:
+                import ctypes
+                FILE_ATTRIBUTE_REPARSE_POINT = 0x400
+                attrs = ctypes.windll.kernel32.GetFileAttributesW(entry.path)
+                return attrs != -1 and (attrs & FILE_ATTRIBUTE_REPARSE_POINT)
+            except Exception:
+                pass
+        return False
 
-    def _scan_directory_links(self, root: str):
-        """递归扫描目录中的符号链接"""
+    def _normalize_path(self, path: str) -> str:
+        return os.path.normpath(path).replace(os.sep, '/')
+
+    def _get_symlink_cache_path(self) -> Path:
+        canonical_path = get_cache_file_path(CacheType.SYMLINK, create_dir=True)
+        return Path(canonical_path)
+
+    def _symlink_roots(self) -> List[str]:
+        roots: List[str] = []
+        roots.extend(self.loras_roots or [])
+        roots.extend(self.base_models_roots or [])
+        roots.extend(self.embeddings_roots or [])
+        return roots
+
+    def _build_symlink_fingerprint(self) -> Dict[str, object]:
+        roots = [self._normalize_path(path) for path in self._symlink_roots() if path]
+        unique_roots = sorted(set(roots))
+
+        # Include first-level symlinks in fingerprint for change detection.
+        # This ensures new symlinks under roots trigger a cache invalidation.
+        # Use lists (not tuples) for JSON serialization compatibility.
+        direct_symlinks: List[List[str]] = []
+        for root in unique_roots:
+            try:
+                if os.path.isdir(root):
+                    with os.scandir(root) as it:
+                        for entry in it:
+                            if self._entry_is_symlink(entry):
+                                try:
+                                    target = os.path.realpath(entry.path)
+                                    direct_symlinks.append([
+                                        self._normalize_path(entry.path),
+                                        self._normalize_path(target)
+                                    ])
+                                except OSError:
+                                    pass
+            except (OSError, PermissionError):
+                pass
+
+        return {
+            "roots": unique_roots,
+            "direct_symlinks": sorted(direct_symlinks)
+        }
+
+    def _initialize_symlink_mappings(self) -> None:
+        start = time.perf_counter()
+        cache_loaded = self._load_persisted_cache_into_mappings()
+
+        if cache_loaded:
+            logger.info(
+                "Symlink mappings restored from cache in %.2f ms",
+                (time.perf_counter() - start) * 1000,
+            )
+            self._rebuild_preview_roots()
+
+            current_fingerprint = self._build_symlink_fingerprint()
+            cached_fingerprint = self._cached_fingerprint
+
+            # Check 1: First-level symlinks unchanged (catches new symlinks at root)
+            fingerprint_valid = cached_fingerprint and current_fingerprint == cached_fingerprint
+
+            # Check 2: All cached mappings still valid (catches changes at any depth)
+            mappings_valid = self._validate_cached_mappings() if fingerprint_valid else False
+
+            if fingerprint_valid and mappings_valid:
+                return
+
+            logger.info("Symlink configuration changed; rescanning symbolic links")
+
+        self.rebuild_symlink_cache()
+        logger.info(
+            "Symlink mappings rebuilt and cached in %.2f ms",
+            (time.perf_counter() - start) * 1000,
+        )
+
+    def rebuild_symlink_cache(self) -> None:
+        """Force a fresh scan of all symbolic links and update the persistent cache."""
+        self._scan_symbolic_links()
+        self._save_symlink_cache()
+        self._rebuild_preview_roots()
+
+    def _load_persisted_cache_into_mappings(self) -> bool:
+        """Load the symlink cache and store its fingerprint for comparison."""
+        cache_path = self._get_symlink_cache_path()
+
+        # Check canonical path first, then legacy paths for migration
+        paths_to_check = [cache_path]
+        legacy_paths = get_legacy_cache_paths(CacheType.SYMLINK)
+        paths_to_check.extend(Path(p) for p in legacy_paths if p != str(cache_path))
+
+        loaded_path = None
+        payload = None
+
+        for check_path in paths_to_check:
+            if not check_path.exists():
+                continue
+            try:
+                with check_path.open("r", encoding="utf-8") as handle:
+                    payload = json.load(handle)
+                    loaded_path = check_path
+                    break
+            except Exception as exc:
+                logger.info("Failed to load symlink cache %s: %s", check_path, exc)
+                continue
+
+        if payload is None:
+            return False
+
+        if not isinstance(payload, dict):
+            return False
+
+        cached_mappings = payload.get("path_mappings")
+        if not isinstance(cached_mappings, Mapping):
+            return False
+
+        # Store the cached fingerprint for comparison during initialization
+        self._cached_fingerprint = payload.get("fingerprint")
+
+        normalized_mappings: Dict[str, str] = {}
+        for target, link in cached_mappings.items():
+            if not isinstance(target, str) or not isinstance(link, str):
+                continue
+            normalized_mappings[self._normalize_path(target)] = self._normalize_path(link)
+
+        self._path_mappings = normalized_mappings
+
+        # Log migration if loaded from legacy path
+        if loaded_path is not None and loaded_path != cache_path:
+            logger.info(
+                "Symlink cache migrated from %s (will save to %s)",
+                loaded_path,
+                cache_path,
+            )
+
+            try:
+                if loaded_path.exists():
+                    loaded_path.unlink()
+                    logger.info("Cleaned up legacy symlink cache: %s", loaded_path)
+
+                    try:
+                        parent_dir = loaded_path.parent
+                        if parent_dir.name == "cache" and not any(parent_dir.iterdir()):
+                            parent_dir.rmdir()
+                            logger.info("Removed empty legacy cache directory: %s", parent_dir)
+                    except Exception:
+                        pass
+
+            except Exception as exc:
+                logger.warning(
+                    "Failed to cleanup legacy symlink cache %s: %s",
+                    loaded_path,
+                    exc,
+                )
+        else:
+            logger.info("Symlink cache loaded with %d mappings", len(self._path_mappings))
+
+        return True
+
+    def _validate_cached_mappings(self) -> bool:
+        """Verify all cached symlink mappings are still valid.
+
+        Returns True if all mappings are valid, False if rescan is needed.
+        This catches removed or retargeted symlinks at ANY depth.
+        """
+        for target, link in self._path_mappings.items():
+            # Convert normalized paths back to OS paths
+            link_path = link.replace('/', os.sep)
+
+            # Check if symlink still exists
+            if not self._is_link(link_path):
+                logger.debug("Cached symlink no longer exists: %s", link_path)
+                return False
+
+            # Check if target is still the same
+            try:
+                actual_target = self._normalize_path(os.path.realpath(link_path))
+                if actual_target != target:
+                    logger.debug(
+                        "Symlink target changed: %s -> %s (cached: %s)",
+                        link_path, actual_target, target
+                    )
+                    return False
+            except OSError:
+                logger.debug("Cannot resolve symlink: %s", link_path)
+                return False
+
+        return True
+
+    def _save_symlink_cache(self) -> None:
+        cache_path = self._get_symlink_cache_path()
+        payload = {
+            "fingerprint": self._build_symlink_fingerprint(),
+            "path_mappings": self._path_mappings,
+        }
+
+        try:
+            with cache_path.open("w", encoding="utf-8") as handle:
+                json.dump(payload, handle, ensure_ascii=False, indent=2)
+            logger.debug("Symlink cache saved to %s with %d mappings", cache_path, len(self._path_mappings))
+        except Exception as exc:
+            logger.info("Failed to write symlink cache %s: %s", cache_path, exc)
+
+    def _scan_symbolic_links(self):
+        """Scan symbolic links in LoRA, Checkpoint, and Embedding root directories.
+
+        Only scans the first level of each root directory to avoid performance
+        issues with large file systems. Detects symlinks and Windows junctions
+        at the root level only (not nested symlinks in subdirectories).
+        """
+        start = time.perf_counter()
+        
+        # Reset mappings before rescanning to avoid stale entries
+        self._path_mappings.clear()
+        self._seed_root_symlink_mappings()
+        for root in self._symlink_roots():
+            self._scan_first_level_symlinks(root)
+        logger.debug(
+            "Symlink scan finished in %.2f ms with %d mappings",
+            (time.perf_counter() - start) * 1000,
+            len(self._path_mappings),
+        )
+
+    def _scan_first_level_symlinks(self, root: str):
+        """Scan only the first level of a directory for symlinks.
+        
+        This avoids traversing the entire directory tree which can be extremely
+        slow for large model collections. Only symlinks directly under the root
+        are detected.
+        """
         try:
             with os.scandir(root) as it:
                 for entry in it:
-                    if self._is_link(entry.path):
+                    try:
+                        # Only detect symlinks including Windows junctions
+                        # Skip normal directories to avoid deep traversal
+                        if not self._entry_is_symlink(entry):
+                            continue
+
+                        # Resolve the symlink target
                         target_path = os.path.realpath(entry.path)
-                        if os.path.isdir(target_path):
-                            self.add_path_mapping(entry.path, target_path)
-                            self._scan_directory_links(target_path)
-                    elif entry.is_dir(follow_symlinks=False):
-                        self._scan_directory_links(entry.path)
+                        if not os.path.isdir(target_path):
+                            continue
+
+                        self.add_path_mapping(entry.path, target_path)
+                    except Exception as inner_exc:
+                        logger.debug(
+                            "Error processing directory entry %s: %s", entry.path, inner_exc
+                        )
         except Exception as e:
             logger.error(f"Error scanning links in {root}: {e}")
 
+
+
     def add_path_mapping(self, link_path: str, target_path: str):
-        """添加符号链接路径映射
-        target_path: 实际目标路径
-        link_path: 符号链接路径
+        """Add a symbolic link path mapping
+        target_path: actual target path
+        link_path: symbolic link path
         """
-        normalized_link = os.path.normpath(link_path).replace(os.sep, '/')
-        normalized_target = os.path.normpath(target_path).replace(os.sep, '/')
-        # 保持原有的映射关系：目标路径 -> 链接路径
+        normalized_link = self._normalize_path(link_path)
+        normalized_target = self._normalize_path(target_path)
+        # Keep the original mapping: target path -> link path
         self._path_mappings[normalized_target] = normalized_link
         logger.info(f"Added path mapping: {normalized_target} -> {normalized_link}")
+        self._preview_root_paths.update(self._expand_preview_root(normalized_target))
+        self._preview_root_paths.update(self._expand_preview_root(normalized_link))
 
-    def add_route_mapping(self, path: str, route: str):
-        """添加静态路由映射"""
-        normalized_path = os.path.normpath(path).replace(os.sep, '/')
-        self._route_mappings[normalized_path] = route
-        logger.info(f"Added route mapping: {normalized_path} -> {route}")
+    def _seed_root_symlink_mappings(self) -> None:
+        """Ensure symlinked root folders are recorded before deep scanning."""
+
+        for root in self._symlink_roots():
+            if not root:
+                continue
+            try:
+                if not self._is_link(root):
+                    continue
+                target_path = os.path.realpath(root)
+                if not os.path.isdir(target_path):
+                    continue
+                self.add_path_mapping(root, target_path)
+            except Exception as exc:
+                logger.debug("Skipping root symlink %s: %s", root, exc)
+
+    def _expand_preview_root(self, path: str) -> Set[Path]:
+        """Return normalized ``Path`` objects representing a preview root."""
+
+        roots: Set[Path] = set()
+        if not path:
+            return roots
+
+        try:
+            raw_path = Path(path).expanduser()
+        except Exception:
+            return roots
+
+        if raw_path.is_absolute():
+            roots.add(raw_path)
+
+        try:
+            resolved = raw_path.resolve(strict=False)
+        except RuntimeError:
+            resolved = raw_path.absolute()
+        roots.add(resolved)
+
+        try:
+            real_path = raw_path.resolve()
+        except (FileNotFoundError, RuntimeError):
+            real_path = resolved
+        roots.add(real_path)
+
+        normalized: Set[Path] = set()
+        for candidate in roots:
+            if candidate.is_absolute():
+                normalized.add(candidate)
+            else:
+                try:
+                    normalized.add(candidate.resolve(strict=False))
+                except RuntimeError:
+                    normalized.add(candidate.absolute())
+
+        return normalized
+
+    def _rebuild_preview_roots(self) -> None:
+        """Recompute the cache of directories permitted for previews."""
+
+        preview_roots: Set[Path] = set()
+
+        for root in self.loras_roots or []:
+            preview_roots.update(self._expand_preview_root(root))
+        for root in self.base_models_roots or []:
+            preview_roots.update(self._expand_preview_root(root))
+        for root in self.embeddings_roots or []:
+            preview_roots.update(self._expand_preview_root(root))
+
+        for target, link in self._path_mappings.items():
+            preview_roots.update(self._expand_preview_root(target))
+            preview_roots.update(self._expand_preview_root(link))
+
+        self._preview_root_paths = {path for path in preview_roots if path.is_absolute()}
+        logger.debug(
+            "Preview roots rebuilt: %d paths from %d lora roots, %d checkpoint roots, %d embedding roots, %d symlink mappings",
+            len(self._preview_root_paths),
+            len(self.loras_roots or []),
+            len(self.base_models_roots or []),
+            len(self.embeddings_roots or []),
+            len(self._path_mappings),
+        )
 
     def map_path_to_link(self, path: str) -> str:
-        """将目标路径映射回符号链接路径"""
+        """Map a target path back to its symbolic link path"""
         normalized_path = os.path.normpath(path).replace(os.sep, '/')
-        # 检查路径是否包含在任何映射的目标路径中
+        # Check if the path is contained in any mapped target path
         for target_path, link_path in self._path_mappings.items():
-            if normalized_path.startswith(target_path):
-                # 如果路径以目标路径开头，则替换为链接路径
+            # Match whole path components to avoid prefix collisions (e.g., /a/b vs /a/bc)
+            if normalized_path == target_path:
+                return link_path
+            
+            if normalized_path.startswith(target_path + '/'):
+                # If the path starts with the target path, replace with link path
                 mapped_path = normalized_path.replace(target_path, link_path, 1)
                 return mapped_path
-        return path
+        return normalized_path
+    
+    def map_link_to_path(self, link_path: str) -> str:
+        """Map a symbolic link path back to the actual path"""
+        normalized_link = os.path.normpath(link_path).replace(os.sep, '/')
+        # Check if the path is contained in any mapped target path
+        for target_path, link_path_mapped in self._path_mappings.items():
+            # Match whole path components
+            if normalized_link == link_path_mapped:
+                return target_path
+
+            if normalized_link.startswith(link_path_mapped + '/'):
+                # If the path starts with the link path, replace with actual path
+                mapped_path = normalized_link.replace(link_path_mapped, target_path, 1)
+                return mapped_path
+        return normalized_link
+
+    def _dedupe_existing_paths(self, raw_paths: Iterable[str]) -> Dict[str, str]:
+        dedup: Dict[str, str] = {}
+        for path in raw_paths:
+            if not isinstance(path, str):
+                continue
+            if not os.path.exists(path):
+                continue
+            real_path = os.path.normpath(os.path.realpath(path)).replace(os.sep, '/')
+            normalized = os.path.normpath(path).replace(os.sep, '/')
+            if real_path not in dedup:
+                dedup[real_path] = normalized
+        return dedup
+
+    def _prepare_lora_paths(self, raw_paths: Iterable[str]) -> List[str]:
+        path_map = self._dedupe_existing_paths(raw_paths)
+        unique_paths = sorted(path_map.values(), key=lambda p: p.lower())
+
+        for original_path in unique_paths:
+            real_path = os.path.normpath(os.path.realpath(original_path)).replace(os.sep, '/')
+            if real_path != original_path:
+                self.add_path_mapping(original_path, real_path)
+
+        return unique_paths
+
+    def _prepare_checkpoint_paths(
+        self, checkpoint_paths: Iterable[str], unet_paths: Iterable[str]
+    ) -> List[str]:
+        checkpoint_map = self._dedupe_existing_paths(checkpoint_paths)
+        unet_map = self._dedupe_existing_paths(unet_paths)
+
+        # Detect when checkpoints and unet share the same physical location
+        # This is a configuration issue that can cause duplicate model entries
+        overlapping_real_paths = set(checkpoint_map.keys()) & set(unet_map.keys())
+        if overlapping_real_paths:
+            logger.warning(
+                "Detected overlapping paths between 'checkpoints' and 'diffusion_models' (unet). "
+                "They should not point to the same physical folder as they are different model types. "
+                "Please fix your ComfyUI path configuration to separate these folders. "
+                "Falling back to 'checkpoints' for backward compatibility. "
+                "Overlapping real paths: %s",
+                [checkpoint_map.get(rp, rp) for rp in overlapping_real_paths]
+            )
+            # Remove overlapping paths from unet_map to prioritize checkpoints
+            for rp in overlapping_real_paths:
+                if rp in unet_map:
+                    del unet_map[rp]
+
+        merged_map: Dict[str, str] = {}
+        for real_path, original in {**checkpoint_map, **unet_map}.items():
+            if real_path not in merged_map:
+                merged_map[real_path] = original
+
+        unique_paths = sorted(merged_map.values(), key=lambda p: p.lower())
+
+        checkpoint_values = set(checkpoint_map.values())
+        unet_values = set(unet_map.values())
+        self.checkpoints_roots = [p for p in unique_paths if p in checkpoint_values]
+        self.unet_roots = [p for p in unique_paths if p in unet_values]
+
+        for original_path in unique_paths:
+            real_path = os.path.normpath(os.path.realpath(original_path)).replace(os.sep, '/')
+            if real_path != original_path:
+                self.add_path_mapping(original_path, real_path)
+
+        return unique_paths
+
+    def _prepare_embedding_paths(self, raw_paths: Iterable[str]) -> List[str]:
+        path_map = self._dedupe_existing_paths(raw_paths)
+        unique_paths = sorted(path_map.values(), key=lambda p: p.lower())
+
+        for original_path in unique_paths:
+            real_path = os.path.normpath(os.path.realpath(original_path)).replace(os.sep, '/')
+            if real_path != original_path:
+                self.add_path_mapping(original_path, real_path)
+
+        return unique_paths
+
+    def _apply_library_paths(self, folder_paths: Mapping[str, Iterable[str]]) -> None:
+        self._path_mappings.clear()
+        self._preview_root_paths = set()
+
+        lora_paths = folder_paths.get('loras', []) or []
+        checkpoint_paths = folder_paths.get('checkpoints', []) or []
+        unet_paths = folder_paths.get('unet', []) or []
+        embedding_paths = folder_paths.get('embeddings', []) or []
+
+        self.loras_roots = self._prepare_lora_paths(lora_paths)
+        self.base_models_roots = self._prepare_checkpoint_paths(checkpoint_paths, unet_paths)
+        self.embeddings_roots = self._prepare_embedding_paths(embedding_paths)
+
+        self._initialize_symlink_mappings()
 
     def _init_lora_paths(self) -> List[str]:
         """Initialize and validate LoRA paths from ComfyUI settings"""
-        paths = list(set(path.replace(os.sep, "/") 
-                for path in folder_paths.get_folder_paths("loras") 
-                if os.path.exists(path)))
-        print("Found LoRA roots:", "\n - " + "\n - ".join(paths))
-        
-        if not paths:
-            raise ValueError("No valid loras folders found in ComfyUI configuration")
-        
-        # 初始化路径映射
-        for path in paths:
-            real_path = os.path.normpath(os.path.realpath(path)).replace(os.sep, '/')
-            if real_path != path:
-                self.add_path_mapping(path, real_path)
-        
-        return paths
+        try:
+            raw_paths = folder_paths.get_folder_paths("loras")
+            unique_paths = self._prepare_lora_paths(raw_paths)
+            logger.info("Found LoRA roots:" + ("\n - " + "\n - ".join(unique_paths) if unique_paths else "[]"))
+
+            if not unique_paths:
+                logger.warning("No valid loras folders found in ComfyUI configuration")
+                return []
+
+            return unique_paths
+        except Exception as e:
+            logger.warning(f"Error initializing LoRA paths: {e}")
+            return []
+
+    def _init_checkpoint_paths(self) -> List[str]:
+        """Initialize and validate checkpoint paths from ComfyUI settings"""
+        try:
+            raw_checkpoint_paths = folder_paths.get_folder_paths("checkpoints")
+            raw_unet_paths = folder_paths.get_folder_paths("unet")
+            unique_paths = self._prepare_checkpoint_paths(raw_checkpoint_paths, raw_unet_paths)
+
+            logger.info("Found checkpoint roots:" + ("\n - " + "\n - ".join(unique_paths) if unique_paths else "[]"))
+
+            if not unique_paths:
+                logger.warning("No valid checkpoint folders found in ComfyUI configuration")
+                return []
+
+            return unique_paths
+        except Exception as e:
+            logger.warning(f"Error initializing checkpoint paths: {e}")
+            return []
+
+    def _init_embedding_paths(self) -> List[str]:
+        """Initialize and validate embedding paths from ComfyUI settings"""
+        try:
+            raw_paths = folder_paths.get_folder_paths("embeddings")
+            unique_paths = self._prepare_embedding_paths(raw_paths)
+            logger.info("Found embedding roots:" + ("\n - " + "\n - ".join(unique_paths) if unique_paths else "[]"))
+
+            if not unique_paths:
+                logger.warning("No valid embeddings folders found in ComfyUI configuration")
+                return []
+
+            return unique_paths
+        except Exception as e:
+            logger.warning(f"Error initializing embedding paths: {e}")
+            return []
 
     def get_preview_static_url(self, preview_path: str) -> str:
-        """Convert local preview path to static URL"""
         if not preview_path:
             return ""
+
+        normalized = os.path.normpath(preview_path).replace(os.sep, '/')
+        encoded_path = urllib.parse.quote(normalized, safe='')
+        return f'/api/lm/previews?path={encoded_path}'
+
+    def is_preview_path_allowed(self, preview_path: str) -> bool:
+        """Return ``True`` if ``preview_path`` is within an allowed directory.
         
-        real_path = os.path.realpath(preview_path).replace(os.sep, '/')
+        If the path is initially rejected, attempts to discover deep symlinks
+        that were not scanned during initialization. If a symlink is found,
+        updates the in-memory path mappings and retries the check.
+        """
 
-        for path, route in self._route_mappings.items():
-            if real_path.startswith(path):
-                relative_path = os.path.relpath(real_path, path)
-                return f'{route}/{relative_path.replace(os.sep, "/")}'
+        if self._is_path_in_allowed_roots(preview_path):
+            return True
 
-        return ""
+        if self._try_discover_deep_symlink(preview_path):
+            return self._is_path_in_allowed_roots(preview_path)
+
+        return False
+
+    def _is_path_in_allowed_roots(self, preview_path: str) -> bool:
+        """Check if preview_path is within allowed preview roots without modification."""
+
+        if not preview_path:
+            return False
+
+        try:
+            candidate = Path(preview_path).expanduser().resolve(strict=False)
+        except Exception:
+            return False
+
+        candidate_str = os.path.normcase(str(candidate))
+        for root in self._preview_root_paths:
+            root_str = os.path.normcase(str(root))
+            if candidate_str == root_str or candidate_str.startswith(root_str + os.sep):
+                return True
+
+        logger.debug(
+            "Path not in allowed roots: %s (candidate=%s, num_roots=%d)",
+            preview_path,
+            candidate_str,
+            len(self._preview_root_paths),
+        )
+
+        return False
+
+    def _try_discover_deep_symlink(self, preview_path: str) -> bool:
+        """Attempt to discover a deep symlink that contains the preview_path.
+
+        Walks up from the preview path to the root directories, checking each
+        parent directory for symlinks. If a symlink is found, updates the
+        in-memory path mappings and preview roots.
+
+        Only updates in-memory state (self._path_mappings and self._preview_root_paths),
+        does not modify the persistent cache file.
+
+        Returns:
+            True if a symlink was discovered and mappings updated, False otherwise.
+        """
+        if not preview_path:
+            return False
+
+        try:
+            candidate = Path(preview_path).expanduser()
+        except Exception:
+            return False
+
+        current = candidate
+        while True:
+            try:
+                if self._is_link(str(current)):
+                    try:
+                        target = os.path.realpath(str(current))
+                        normalized_target = self._normalize_path(target)
+                        normalized_link = self._normalize_path(str(current))
+
+                        self._path_mappings[normalized_target] = normalized_link
+                        self._preview_root_paths.update(self._expand_preview_root(normalized_target))
+                        self._preview_root_paths.update(self._expand_preview_root(normalized_link))
+
+                        logger.debug(
+                            "Discovered deep symlink: %s -> %s (preview path: %s)",
+                            normalized_link,
+                            normalized_target,
+                            preview_path
+                        )
+
+                        return True
+                    except OSError:
+                        pass
+            except OSError:
+                pass
+
+            parent = current.parent
+            if parent == current:
+                break
+            current = parent
+
+        return False
+
+    def apply_library_settings(self, library_config: Mapping[str, object]) -> None:
+        """Update runtime paths to match the provided library configuration."""
+        folder_paths = library_config.get('folder_paths') if isinstance(library_config, Mapping) else {}
+        if not isinstance(folder_paths, Mapping):
+            folder_paths = {}
+
+        self._apply_library_paths(folder_paths)
+
+        logger.info(
+            "Applied library settings with %d lora roots, %d checkpoint roots, and %d embedding roots",
+            len(self.loras_roots or []),
+            len(self.base_models_roots or []),
+            len(self.embeddings_roots or []),
+        )
+
+    def get_library_registry_snapshot(self) -> Dict[str, object]:
+        """Return the current library registry and active library name."""
+
+        try:
+            from .services.settings_manager import get_settings_manager
+
+            settings_service = get_settings_manager()
+            libraries = settings_service.get_libraries()
+            active_library = settings_service.get_active_library_name()
+            return {
+                "active_library": active_library,
+                "libraries": libraries,
+            }
+        except Exception as exc:  # pragma: no cover - defensive logging
+            logger.debug("Failed to collect library registry snapshot: %s", exc)
+            return {"active_library": "", "libraries": {}}
 
 # Global config instance
 config = Config()

py/lora_manager.py

@@ -1,107 +1,376 @@
 import asyncio
+import sys
 import os
-from server import PromptServer # type: ignore
-from .config import config
-from .routes.lora_routes import LoraRoutes
-from .routes.api_routes import ApiRoutes
-from .services.lora_scanner import LoraScanner
-from .services.file_monitor import LoraFileMonitor
-from .services.lora_cache import LoraCache
 import logging
+from .utils.logging_config import setup_logging
+
+# Check if we're in standalone mode
+standalone_mode = os.environ.get("LORA_MANAGER_STANDALONE", "0") == "1" or os.environ.get("HF_HUB_DISABLE_TELEMETRY", "0") == "0"
+
+# Only setup logging prefix if not in standalone mode
+if not standalone_mode:
+    setup_logging()
+
+from server import PromptServer # type: ignore
+
+from .config import config
+from .services.model_service_factory import ModelServiceFactory, register_default_model_types
+from .routes.recipe_routes import RecipeRoutes
+from .routes.stats_routes import StatsRoutes
+from .routes.update_routes import UpdateRoutes
+from .routes.misc_routes import MiscRoutes
+from .routes.preview_routes import PreviewRoutes
+from .routes.example_images_routes import ExampleImagesRoutes
+from .services.service_registry import ServiceRegistry
+from .services.settings_manager import get_settings_manager
+from .utils.example_images_migration import ExampleImagesMigration
+from .services.websocket_manager import ws_manager
+from .services.example_images_cleanup_service import ExampleImagesCleanupService
+from .middleware.csp_middleware import relax_csp_for_remote_media
 
 logger = logging.getLogger(__name__)
 
+HEADER_SIZE_LIMIT = 16384
+
+
+def _sanitize_size_limit(value):
+    """Return a non-negative integer size for ``handler_args`` comparisons."""
+
+    try:
+        coerced = int(value)
+    except (TypeError, ValueError):
+        return 0
+    return coerced if coerced >= 0 else 0
+
+
+class _SettingsProxy:
+    def __init__(self):
+        self._manager = None
+
+    def _resolve(self):
+        if self._manager is None:
+            self._manager = get_settings_manager()
+        return self._manager
+
+    def get(self, *args, **kwargs):
+        return self._resolve().get(*args, **kwargs)
+
+    def __getattr__(self, item):
+        return getattr(self._resolve(), item)
+
+
+settings = _SettingsProxy()
+
 class LoraManager:
     """Main entry point for LoRA Manager plugin"""
     
     @classmethod
     def add_routes(cls):
-        """Initialize and register all routes"""
+        """Initialize and register all routes using the new refactored architecture"""
         app = PromptServer.instance.app
 
-        added_targets = set()  # 用于跟踪已添加的目标路径
-        
-        # Add static routes for each lora root
-        for idx, root in enumerate(config.loras_roots, start=1):
-            preview_path = f'/loras_static/root{idx}/preview'
-            
-            real_root = root
-            if root in config._path_mappings.values():
-                for target, link in config._path_mappings.items():
-                    if link == root:
-                        real_root = target
-                        break
-            # 为原始路径添加静态路由
-            app.router.add_static(preview_path, real_root)
-            logger.info(f"Added static route {preview_path} -> {real_root}")
-            
-            # 记录路由映射
-            config.add_route_mapping(real_root, preview_path)
-            added_targets.add(real_root)
-        
-        # 为符号链接的目标路径添加额外的静态路由
-        link_idx = 1
-        
-        for target_path, link_path in config._path_mappings.items():
-            if target_path not in added_targets:
-                route_path = f'/loras_static/link_{link_idx}/preview'
-                app.router.add_static(route_path, target_path)
-                logger.info(f"Added static route for link target {route_path} -> {target_path}")
-                config.add_route_mapping(target_path, route_path)
-                added_targets.add(target_path)
-                link_idx += 1
-        
+        if relax_csp_for_remote_media not in app.middlewares:
+            # Ensure CSP relaxer executes after ComfyUI's block_external_middleware so it can
+            # see and extend the restrictive header instead of being overwritten by it.
+            block_middleware_index = next(
+                (
+                    idx
+                    for idx, middleware in enumerate(app.middlewares)
+                    if getattr(middleware, "__name__", "") == "block_external_middleware"
+                ),
+                None,
+            )
+
+            if block_middleware_index is None:
+                app.middlewares.append(relax_csp_for_remote_media)
+            else:
+                app.middlewares.insert(block_middleware_index, relax_csp_for_remote_media)
+
+        # Increase allowed header sizes so browsers with large localhost cookie
+        # jars (multiple UIs on 127.0.0.1) don't trip aiohttp's 8KB default
+        # limits. Cookies for unrelated apps are still sent to the plugin and
+        # may otherwise raise LineTooLong errors when the request parser reads
+        # them. Preserve any previously configured handler arguments while
+        # ensuring our minimum sizes are applied.
+        handler_args = getattr(app, "_handler_args", {}) or {}
+        updated_handler_args = dict(handler_args)
+        updated_handler_args["max_field_size"] = max(
+            _sanitize_size_limit(handler_args.get("max_field_size", 0)),
+            HEADER_SIZE_LIMIT,
+        )
+        updated_handler_args["max_line_size"] = max(
+            _sanitize_size_limit(handler_args.get("max_line_size", 0)),
+            HEADER_SIZE_LIMIT,
+        )
+        app._handler_args = updated_handler_args
+
+        # Configure aiohttp access logger to be less verbose
+        logging.getLogger('aiohttp.access').setLevel(logging.WARNING)
+
+        # Add specific suppression for connection reset errors
+        class ConnectionResetFilter(logging.Filter):
+            def filter(self, record):
+                # Filter out connection reset errors that are not critical
+                if "ConnectionResetError" in str(record.getMessage()):
+                    return False
+                if "_call_connection_lost" in str(record.getMessage()):
+                    return False
+                if "WinError 10054" in str(record.getMessage()):
+                    return False
+                return True
+
+        # Apply the filter to asyncio logger
+        asyncio_logger = logging.getLogger("asyncio")
+        asyncio_logger.addFilter(ConnectionResetFilter())
+
+        # Add static route for example images if the path exists in settings
+        example_images_path = settings.get('example_images_path')
+        logger.info(f"Example images path: {example_images_path}")
+        if example_images_path and os.path.exists(example_images_path):
+            app.router.add_static('/example_images_static', example_images_path)
+            logger.info(f"Added static route for example images: /example_images_static -> {example_images_path}")
+
+        # Add static route for locales JSON files
+        if os.path.exists(config.i18n_path):
+            app.router.add_static('/locales', config.i18n_path)
+            logger.info(f"Added static route for locales: /locales -> {config.i18n_path}")
+
         # Add static route for plugin assets
         app.router.add_static('/loras_static', config.static_path)
         
-        # Setup feature routes
-        routes = LoraRoutes()
+        # Register default model types with the factory
+        register_default_model_types()
         
-        # Setup file monitoring
-        monitor = LoraFileMonitor(routes.scanner, config.loras_roots)
-        monitor.start()
+        # Setup all model routes using the factory
+        ModelServiceFactory.setup_all_routes(app)
         
-        routes.setup_routes(app)
-        ApiRoutes.setup_routes(app, monitor)
+        # Setup non-model-specific routes
+        stats_routes = StatsRoutes()
+        stats_routes.setup_routes(app)
+        RecipeRoutes.setup_routes(app)
+        UpdateRoutes.setup_routes(app)  
+        MiscRoutes.setup_routes(app)
+        ExampleImagesRoutes.setup_routes(app, ws_manager=ws_manager)
+        PreviewRoutes.setup_routes(app)
         
-        # Store monitor in app for cleanup
-        app['lora_monitor'] = monitor
+        # Setup WebSocket routes that are shared across all model types
+        app.router.add_get('/ws/fetch-progress', ws_manager.handle_connection)
+        app.router.add_get('/ws/download-progress', ws_manager.handle_download_connection)
+        app.router.add_get('/ws/init-progress', ws_manager.handle_init_connection)
         
-        # Schedule cache initialization using the application's startup handler
-        app.on_startup.append(lambda app: cls._schedule_cache_init(routes.scanner))
+        # Schedule service initialization 
+        app.on_startup.append(lambda app: cls._initialize_services())
         
         # Add cleanup
         app.on_shutdown.append(cls._cleanup)
-        app.on_shutdown.append(ApiRoutes.cleanup)
-    
+        
     @classmethod
-    async def _schedule_cache_init(cls, scanner: LoraScanner):
-        """Schedule cache initialization in the running event loop"""
+    async def _initialize_services(cls):
+        """Initialize all services using the ServiceRegistry"""
         try:
-            # 创建低优先级的初始化任务
-            asyncio.create_task(cls._initialize_cache(scanner), name='lora_cache_init')
-        except Exception as e:
-            print(f"LoRA Manager: Error scheduling cache initialization: {e}")
-    
-    @classmethod
-    async def _initialize_cache(cls, scanner: LoraScanner):
-        """Initialize cache in background"""
-        try:
-            # 设置初始缓存占位
-            scanner._cache = LoraCache(
-                raw_data=[],
-                sorted_by_name=[],
-                sorted_by_date=[],
-                folders=[]
+            # Initialize CivitaiClient first to ensure it's ready for other services
+            await ServiceRegistry.get_civitai_client()
+
+            # Register DownloadManager with ServiceRegistry
+            await ServiceRegistry.get_download_manager()
+
+            from .services.metadata_service import initialize_metadata_providers
+            await initialize_metadata_providers()
+            
+            # Initialize WebSocket manager
+            await ServiceRegistry.get_websocket_manager()
+            
+            # Initialize scanners in background
+            lora_scanner = await ServiceRegistry.get_lora_scanner()
+            checkpoint_scanner = await ServiceRegistry.get_checkpoint_scanner()
+            embedding_scanner = await ServiceRegistry.get_embedding_scanner()
+            
+            # Initialize recipe scanner if needed
+            recipe_scanner = await ServiceRegistry.get_recipe_scanner()
+            
+            # Create low-priority initialization tasks
+            init_tasks = [
+                asyncio.create_task(lora_scanner.initialize_in_background(), name='lora_cache_init'),
+                asyncio.create_task(checkpoint_scanner.initialize_in_background(), name='checkpoint_cache_init'),
+                asyncio.create_task(embedding_scanner.initialize_in_background(), name='embedding_cache_init'),
+                asyncio.create_task(recipe_scanner.initialize_in_background(), name='recipe_cache_init')
+            ]
+
+            await ExampleImagesMigration.check_and_run_migrations()
+            
+            # Schedule post-initialization tasks to run after scanners complete
+            asyncio.create_task(
+                cls._run_post_initialization_tasks(init_tasks), 
+                name='post_init_tasks'
             )
             
-            # 分阶段加载缓存
-            await scanner.get_cached_data(force_refresh=True)
+            logger.debug("LoRA Manager: All services initialized and background tasks scheduled")
+                
         except Exception as e:
-            print(f"LoRA Manager: Error initializing cache: {e}")
+            logger.error(f"LoRA Manager: Error initializing services: {e}", exc_info=True)
     
+    @classmethod
+    async def _run_post_initialization_tasks(cls, init_tasks):
+        """Run post-initialization tasks after all scanners complete"""
+        try:
+            logger.debug("LoRA Manager: Waiting for scanner initialization to complete...")
+            
+            # Wait for all scanner initialization tasks to complete
+            await asyncio.gather(*init_tasks, return_exceptions=True)
+
+            logger.debug("LoRA Manager: Scanner initialization completed, starting post-initialization tasks...")
+            
+            # Run post-initialization tasks
+            post_tasks = [
+                asyncio.create_task(cls._cleanup_backup_files(), name='cleanup_bak_files'),
+                # Add more post-initialization tasks here as needed
+                # asyncio.create_task(cls._another_post_task(), name='another_task'),
+            ]
+            
+            # Run all post-initialization tasks
+            results = await asyncio.gather(*post_tasks, return_exceptions=True)
+            
+            # Log results
+            for i, result in enumerate(results):
+                task_name = post_tasks[i].get_name()
+                if isinstance(result, Exception):
+                    logger.error(f"Post-initialization task '{task_name}' failed: {result}")
+                else:
+                    logger.debug(f"Post-initialization task '{task_name}' completed successfully")
+                    
+            logger.debug("LoRA Manager: All post-initialization tasks completed")
+            
+        except Exception as e:
+            logger.error(f"LoRA Manager: Error in post-initialization tasks: {e}", exc_info=True)
+    
+    @classmethod
+    async def _cleanup_backup_files(cls):
+        """Clean up .bak files in all model roots"""
+        try:
+            logger.debug("Starting cleanup of .bak files in model directories...")
+            
+            # Collect all model roots
+            all_roots = set()
+            all_roots.update(config.loras_roots)
+            all_roots.update(config.base_models_roots) 
+            all_roots.update(config.embeddings_roots)
+            
+            total_deleted = 0
+            total_size_freed = 0
+            
+            for root_path in all_roots:
+                if not os.path.exists(root_path):
+                    continue
+                    
+                try:
+                    deleted_count, size_freed = await cls._cleanup_backup_files_in_directory(root_path)
+                    total_deleted += deleted_count
+                    total_size_freed += size_freed
+                    
+                    if deleted_count > 0:
+                        logger.debug(f"Cleaned up {deleted_count} .bak files in {root_path} (freed {size_freed / (1024*1024):.2f} MB)")
+                        
+                except Exception as e:
+                    logger.error(f"Error cleaning up .bak files in {root_path}: {e}")
+                    
+                # Yield control periodically
+                await asyncio.sleep(0.01)
+            
+            if total_deleted > 0:
+                logger.debug(f"Backup cleanup completed: removed {total_deleted} .bak files, freed {total_size_freed / (1024*1024):.2f} MB total")
+            else:
+                logger.debug("Backup cleanup completed: no .bak files found")
+
+        except Exception as e:
+            logger.error(f"Error during backup file cleanup: {e}", exc_info=True)
+    
+    @classmethod
+    async def _cleanup_backup_files_in_directory(cls, directory_path: str):
+        """Clean up .bak files in a specific directory recursively
+        
+        Args:
+            directory_path: Path to the directory to clean
+            
+        Returns:
+            Tuple[int, int]: (number of files deleted, total size freed in bytes)
+        """
+        deleted_count = 0
+        size_freed = 0
+        visited_paths = set()
+        
+        def cleanup_recursive(path):
+            nonlocal deleted_count, size_freed
+            
+            try:
+                real_path = os.path.realpath(path)
+                if real_path in visited_paths:
+                    return
+                visited_paths.add(real_path)
+                
+                with os.scandir(path) as it:
+                    for entry in it:
+                        try:
+                            if entry.is_file(follow_symlinks=True) and entry.name.endswith('.bak'):
+                                file_size = entry.stat().st_size
+                                os.remove(entry.path)
+                                deleted_count += 1
+                                size_freed += file_size
+                                logger.debug(f"Deleted .bak file: {entry.path}")
+                                
+                            elif entry.is_dir(follow_symlinks=True):
+                                cleanup_recursive(entry.path)
+                                
+                        except Exception as e:
+                            logger.warning(f"Could not delete .bak file {entry.path}: {e}")
+                            
+            except Exception as e:
+                logger.error(f"Error scanning directory {path} for .bak files: {e}")
+        
+        # Run the recursive cleanup in a thread pool to avoid blocking
+        loop = asyncio.get_event_loop()
+        await loop.run_in_executor(None, cleanup_recursive, directory_path)
+        
+        return deleted_count, size_freed
+    
+    @classmethod
+    async def _cleanup_example_images_folders(cls):
+        """Invoke the example images cleanup service for manual execution."""
+        try:
+            service = ExampleImagesCleanupService()
+            result = await service.cleanup_example_image_folders()
+
+            if result.get('success'):
+                logger.debug(
+                    "Manual example images cleanup completed: moved=%s",
+                    result.get('moved_total'),
+                )
+            elif result.get('partial_success'):
+                logger.warning(
+                    "Manual example images cleanup partially succeeded: moved=%s failures=%s",
+                    result.get('moved_total'),
+                    result.get('move_failures'),
+                )
+            else:
+                logger.debug(
+                    "Manual example images cleanup skipped or failed: %s",
+                    result.get('error', 'no changes'),
+                )
+
+            return result
+
+        except Exception as e:  # pragma: no cover - defensive guard
+            logger.error(f"Error during example images cleanup: {e}", exc_info=True)
+            return {
+                'success': False,
+                'error': str(e),
+                'error_code': 'unexpected_error',
+            }
+
     @classmethod
     async def _cleanup(cls, app):
-        """Cleanup resources"""
-        if 'lora_monitor' in app:
-            app['lora_monitor'].stop()
+        """Cleanup resources using ServiceRegistry"""
+        try:
+            logger.info("LoRA Manager: Cleaning up services")
+                
+        except Exception as e:
+            logger.error(f"Error during cleanup: {e}", exc_info=True)

py/metadata_collector/__init__.py

@@ -0,0 +1,33 @@
+import os
+import logging
+
+logger = logging.getLogger(__name__)
+
+# Check if running in standalone mode
+standalone_mode = os.environ.get("LORA_MANAGER_STANDALONE", "0") == "1" or os.environ.get("HF_HUB_DISABLE_TELEMETRY", "0") == "0"
+
+if not standalone_mode:
+    from .metadata_hook import MetadataHook
+    from .metadata_registry import MetadataRegistry
+
+    def init():
+        # Install hooks to collect metadata during execution
+        MetadataHook.install()
+        
+        # Initialize registry
+        registry = MetadataRegistry()
+        
+        logger.info("ComfyUI Metadata Collector initialized")
+        
+    def get_metadata(prompt_id=None):
+        """Helper function to get metadata from the registry"""
+        registry = MetadataRegistry()
+        return registry.get_metadata(prompt_id)
+else:
+    # Standalone mode - provide dummy implementations
+    def init():
+        logger.info("ComfyUI Metadata Collector disabled in standalone mode")
+        
+    def get_metadata(prompt_id=None):
+        """Dummy implementation for standalone mode"""
+        return {}

py/metadata_collector/constants.py

@@ -0,0 +1,13 @@
+"""Constants used by the metadata collector"""
+
+# Metadata categories
+MODELS = "models"
+PROMPTS = "prompts"
+SAMPLING = "sampling"
+LORAS = "loras"
+SIZE = "size"
+IMAGES = "images"
+IS_SAMPLER = "is_sampler"  # New constant to mark sampler nodes
+
+# Complete list of categories to track
+METADATA_CATEGORIES = [MODELS, PROMPTS, SAMPLING, LORAS, SIZE, IMAGES]

py/metadata_collector/metadata_hook.py

@@ -0,0 +1,207 @@
+import sys
+import inspect
+import logging
+from .metadata_registry import MetadataRegistry
+
+logger = logging.getLogger(__name__)
+
+class MetadataHook:
+    """Install hooks for metadata collection"""
+    
+    @staticmethod
+    def install():
+        """Install hooks to collect metadata during execution"""
+        try:
+            # Import ComfyUI's execution module
+            execution = None
+            try:
+                # Try direct import first
+                import execution # type: ignore
+            except ImportError:
+                # Try to locate from system modules
+                for module_name in sys.modules:
+                    if module_name.endswith('.execution'):
+                        execution = sys.modules[module_name]
+                        break
+                    
+            # If we can't find the execution module, we can't install hooks
+            if execution is None:
+                logger.warning("Could not locate ComfyUI execution module, metadata collection disabled")
+                return
+            
+            # Detect whether we're using the new async version of ComfyUI
+            is_async = False
+            map_node_func_name = '_map_node_over_list'
+            
+            if hasattr(execution, '_async_map_node_over_list'):
+                is_async = inspect.iscoroutinefunction(execution._async_map_node_over_list)
+                map_node_func_name = '_async_map_node_over_list'
+            elif hasattr(execution, '_map_node_over_list'):
+                is_async = inspect.iscoroutinefunction(execution._map_node_over_list)
+            
+            if is_async:
+                logger.info("Detected async ComfyUI execution, installing async metadata hooks")
+                MetadataHook._install_async_hooks(execution, map_node_func_name)
+            else:
+                logger.info("Detected sync ComfyUI execution, installing sync metadata hooks")
+                MetadataHook._install_sync_hooks(execution)
+            
+            logger.info("Metadata collection hooks installed for runtime values")
+            
+        except Exception as e:
+            logger.error(f"Error installing metadata hooks: {str(e)}")
+    
+    @staticmethod
+    def _install_sync_hooks(execution):
+        """Install hooks for synchronous execution model"""
+        # Store the original _map_node_over_list function
+        original_map_node_over_list = execution._map_node_over_list
+        
+        # Define the wrapped _map_node_over_list function
+        def map_node_over_list_with_metadata(obj, input_data_all, func, allow_interrupt=False, execution_block_cb=None, pre_execute_cb=None):
+            # Only collect metadata when calling the main function of nodes
+            if func == obj.FUNCTION and hasattr(obj, '__class__'):
+                try:
+                    # Get the current prompt_id from the registry
+                    registry = MetadataRegistry()
+                    prompt_id = registry.current_prompt_id
+                    
+                    if prompt_id is not None:
+                        # Get node class type
+                        class_type = obj.__class__.__name__
+                        
+                        # Unique ID might be available through the obj if it has a unique_id field
+                        node_id = getattr(obj, 'unique_id', None)
+                        if node_id is None and pre_execute_cb:
+                            # Try to extract node_id through reflection on GraphBuilder.set_default_prefix
+                            frame = inspect.currentframe()
+                            while frame:
+                                if 'unique_id' in frame.f_locals:
+                                    node_id = frame.f_locals['unique_id']
+                                    break
+                                frame = frame.f_back
+                        
+                        # Record inputs before execution
+                        if node_id is not None:
+                            registry.record_node_execution(node_id, class_type, input_data_all, None)
+                except Exception as e:
+                    logger.error(f"Error collecting metadata (pre-execution): {str(e)}")
+            
+            # Execute the original function
+            results = original_map_node_over_list(obj, input_data_all, func, allow_interrupt, execution_block_cb, pre_execute_cb)
+            
+            # After execution, collect outputs for relevant nodes
+            if func == obj.FUNCTION and hasattr(obj, '__class__'):
+                try:
+                    # Get the current prompt_id from the registry
+                    registry = MetadataRegistry()
+                    prompt_id = registry.current_prompt_id
+                    
+                    if prompt_id is not None:
+                        # Get node class type
+                        class_type = obj.__class__.__name__
+                        
+                        # Unique ID might be available through the obj if it has a unique_id field
+                        node_id = getattr(obj, 'unique_id', None)
+                        if node_id is None and pre_execute_cb:
+                            # Try to extract node_id through reflection
+                            frame = inspect.currentframe()
+                            while frame:
+                                if 'unique_id' in frame.f_locals:
+                                    node_id = frame.f_locals['unique_id']
+                                    break
+                                frame = frame.f_back
+                        
+                        # Record outputs after execution
+                        if node_id is not None:
+                            registry.update_node_execution(node_id, class_type, results)
+                except Exception as e:
+                    logger.error(f"Error collecting metadata (post-execution): {str(e)}")
+            
+            return results
+            
+        # Also hook the execute function to track the current prompt_id
+        original_execute = execution.execute
+        
+        def execute_with_prompt_tracking(*args, **kwargs):
+            if len(args) >= 7:  # Check if we have enough arguments
+                server, prompt, caches, node_id, extra_data, executed, prompt_id = args[:7]
+                registry = MetadataRegistry()
+                
+                # Start collection if this is a new prompt
+                if not registry.current_prompt_id or registry.current_prompt_id != prompt_id:
+                    registry.start_collection(prompt_id)
+                    
+                # Store the dynprompt reference for node lookups
+                if hasattr(prompt, 'original_prompt'):
+                    registry.set_current_prompt(prompt)
+            
+            # Execute the original function
+            return original_execute(*args, **kwargs)
+            
+        # Replace the functions
+        execution._map_node_over_list = map_node_over_list_with_metadata
+        execution.execute = execute_with_prompt_tracking
+    
+    @staticmethod
+    def _install_async_hooks(execution, map_node_func_name='_async_map_node_over_list'):
+        """Install hooks for asynchronous execution model"""
+        # Store the original _async_map_node_over_list function
+        original_map_node_over_list = getattr(execution, map_node_func_name)
+        
+        # Wrapped async function, compatible with both stable and nightly
+        async def async_map_node_over_list_with_metadata(prompt_id, unique_id, obj, input_data_all, func, allow_interrupt=False, execution_block_cb=None, pre_execute_cb=None, *args, **kwargs):
+            hidden_inputs = kwargs.get('hidden_inputs', None)
+            # Only collect metadata when calling the main function of nodes
+            if func == obj.FUNCTION and hasattr(obj, '__class__'):
+                try:
+                    registry = MetadataRegistry()
+                    if prompt_id is not None:
+                        class_type = obj.__class__.__name__
+                        node_id = unique_id
+                        if node_id is not None:
+                            registry.record_node_execution(node_id, class_type, input_data_all, None)
+                except Exception as e:
+                    logger.error(f"Error collecting metadata (pre-execution): {str(e)}")
+            
+            # Call original function with all args/kwargs
+            results = await original_map_node_over_list(
+                prompt_id, unique_id, obj, input_data_all, func,
+                allow_interrupt, execution_block_cb, pre_execute_cb, *args, **kwargs
+            )
+            
+            if func == obj.FUNCTION and hasattr(obj, '__class__'):
+                try:
+                    registry = MetadataRegistry()
+                    if prompt_id is not None:
+                        class_type = obj.__class__.__name__
+                        node_id = unique_id
+                        if node_id is not None:
+                            registry.update_node_execution(node_id, class_type, results)
+                except Exception as e:
+                    logger.error(f"Error collecting metadata (post-execution): {str(e)}")
+            
+            return results
+        
+        # Also hook the execute function to track the current prompt_id
+        original_execute = execution.execute
+        
+        async def async_execute_with_prompt_tracking(*args, **kwargs):
+            if len(args) >= 7:  # Check if we have enough arguments
+                server, prompt, caches, node_id, extra_data, executed, prompt_id = args[:7]
+                registry = MetadataRegistry()
+                
+                # Start collection if this is a new prompt
+                if not registry.current_prompt_id or registry.current_prompt_id != prompt_id:
+                    registry.start_collection(prompt_id)
+                    
+                # Store the dynprompt reference for node lookups
+                if hasattr(prompt, 'original_prompt'):
+                    registry.set_current_prompt(prompt)
+            
+            # Execute the original function
+            return await original_execute(*args, **kwargs)
+            
+        # Replace the functions with async versions
+        setattr(execution, map_node_func_name, async_map_node_over_list_with_metadata)
+        execution.execute = async_execute_with_prompt_tracking

py/metadata_collector/metadata_processor.py

@@ -0,0 +1,600 @@
+import json
+import os
+from .constants import IMAGES
+
+# Check if running in standalone mode
+standalone_mode = os.environ.get("LORA_MANAGER_STANDALONE", "0") == "1" or os.environ.get("HF_HUB_DISABLE_TELEMETRY", "0") == "0"
+
+from .constants import MODELS, PROMPTS, SAMPLING, LORAS, SIZE, IS_SAMPLER
+
+class MetadataProcessor:
+    """Process and format collected metadata"""
+    
+    @staticmethod
+    def find_primary_sampler(metadata, downstream_id=None):
+        """
+        Find the primary KSampler node that executed before the given downstream node
+        
+        Parameters:
+        - metadata: The workflow metadata
+        - downstream_id: Optional ID of a downstream node to help identify the specific primary sampler
+        """
+        if downstream_id is None:
+            if IMAGES in metadata and "first_decode" in metadata[IMAGES]:
+                downstream_id = metadata[IMAGES]["first_decode"]["node_id"]
+                
+        # If we have a downstream_id and execution_order, use it to narrow down potential samplers
+        if downstream_id and "execution_order" in metadata:
+            execution_order = metadata["execution_order"]
+            
+            # Find the index of the downstream node in the execution order
+            if downstream_id in execution_order:
+                downstream_index = execution_order.index(downstream_id)
+                
+                # Extract all sampler nodes that executed before the downstream node
+                candidate_samplers = {}
+                for i in range(downstream_index):
+                    node_id = execution_order[i]
+                    # Use IS_SAMPLER flag to identify true sampler nodes
+                    if node_id in metadata.get(SAMPLING, {}) and metadata[SAMPLING][node_id].get(IS_SAMPLER, False):
+                        candidate_samplers[node_id] = metadata[SAMPLING][node_id]
+                
+                    # If we found candidate samplers, apply primary sampler logic to these candidates only
+                    
+                    # PRE-PROCESS: Ensure all candidate samplers have their parameters populated
+                    # This is especially important for SamplerCustomAdvanced which needs tracing
+                    prompt = metadata.get("current_prompt")
+                    for node_id in candidate_samplers:
+                        # If a sampler is missing common parameters like steps or denoise, 
+                        # try to populate them using tracing before ranking
+                        sampler_info = candidate_samplers[node_id]
+                        params = sampler_info.get("parameters", {})
+                        
+                        if prompt and (params.get("steps") is None or params.get("denoise") is None):
+                            # Create a temporary params dict to use the handler
+                            temp_params = {
+                                "steps": params.get("steps"),
+                                "denoise": params.get("denoise"),
+                                "sampler": params.get("sampler_name"),
+                                "scheduler": params.get("scheduler")
+                            }
+                            
+                            # Check if it's SamplerCustomAdvanced
+                            if prompt.original_prompt and node_id in prompt.original_prompt:
+                                if prompt.original_prompt[node_id].get("class_type") == "SamplerCustomAdvanced":
+                                    MetadataProcessor.handle_custom_advanced_sampler(metadata, prompt, node_id, temp_params)
+                                    
+                                    # Update the actual parameters with found values
+                                    params["steps"] = temp_params.get("steps")
+                                    params["denoise"] = temp_params.get("denoise")
+                                    if temp_params.get("sampler"):
+                                        params["sampler_name"] = temp_params.get("sampler")
+                                    if temp_params.get("scheduler"):
+                                        params["scheduler"] = temp_params.get("scheduler")
+
+                    # Collect potential primary samplers based on different criteria
+                    custom_advanced_samplers = []
+                    advanced_add_noise_samplers = []
+                    high_denoise_samplers = []
+                    max_denoise = -1
+                    high_denoise_id = None
+                    
+                    # First, check for SamplerCustomAdvanced among candidates
+                    if prompt and prompt.original_prompt:
+                        for node_id in candidate_samplers:
+                            node_info = prompt.original_prompt.get(node_id, {})
+                            if node_info.get("class_type") == "SamplerCustomAdvanced":
+                                custom_advanced_samplers.append(node_id)
+                    
+                    # Next, check for KSamplerAdvanced with add_noise="enable" among candidates
+                    for node_id, sampler_info in candidate_samplers.items():
+                        parameters = sampler_info.get("parameters", {})
+                        add_noise = parameters.get("add_noise")
+                        if add_noise == "enable":
+                            advanced_add_noise_samplers.append(node_id)
+                    
+                    # Find the sampler with highest denoise value among candidates
+                    for node_id, sampler_info in candidate_samplers.items():
+                        parameters = sampler_info.get("parameters", {})
+                        denoise = parameters.get("denoise")
+                        if denoise is not None and denoise > max_denoise:
+                            max_denoise = denoise
+                            high_denoise_id = node_id
+                    
+                    if high_denoise_id:
+                        high_denoise_samplers.append(high_denoise_id)
+                    
+                    # Combine all potential primary samplers
+                    potential_samplers = custom_advanced_samplers + advanced_add_noise_samplers + high_denoise_samplers
+                    
+                    # Find the first potential primary sampler (prefer base sampler over refine)
+                    # Use forward search to prioritize the first one in execution order
+                    for i in range(downstream_index):
+                        node_id = execution_order[i]
+                        if node_id in potential_samplers:
+                            return node_id, candidate_samplers[node_id]
+                    
+                    # If no potential sampler found from our criteria, return the first sampler
+                    if candidate_samplers:
+                        for i in range(downstream_index):
+                            node_id = execution_order[i]
+                            if node_id in candidate_samplers:
+                                return node_id, candidate_samplers[node_id]
+        
+        # If no downstream_id provided or no suitable sampler found, fall back to original logic
+        primary_sampler = None
+        primary_sampler_id = None
+        max_denoise = -1
+        
+        # First, check for SamplerCustomAdvanced
+        prompt = metadata.get("current_prompt")
+        if prompt and prompt.original_prompt:
+            for node_id, node_info in prompt.original_prompt.items():
+                if node_info.get("class_type") == "SamplerCustomAdvanced":
+                    # Check if the node is in SAMPLING and has IS_SAMPLER flag
+                    if node_id in metadata.get(SAMPLING, {}) and metadata[SAMPLING][node_id].get(IS_SAMPLER, False):
+                        return node_id, metadata[SAMPLING][node_id]
+        
+        # Next, check for KSamplerAdvanced with add_noise="enable" using IS_SAMPLER flag
+        for node_id, sampler_info in metadata.get(SAMPLING, {}).items():
+            # Skip if not marked as a sampler
+            if not sampler_info.get(IS_SAMPLER, False):
+                continue
+                
+            parameters = sampler_info.get("parameters", {})
+            add_noise = parameters.get("add_noise")
+            if add_noise == "enable":
+                primary_sampler = sampler_info
+                primary_sampler_id = node_id
+                break
+        
+        # If no specialized sampler found, find the sampler with highest denoise value
+        if primary_sampler is None:
+            for node_id, sampler_info in metadata.get(SAMPLING, {}).items():
+                # Skip if not marked as a sampler
+                if not sampler_info.get(IS_SAMPLER, False):
+                    continue
+                    
+                parameters = sampler_info.get("parameters", {})
+                denoise = parameters.get("denoise")
+                if denoise is not None and denoise > max_denoise:
+                    max_denoise = denoise
+                    primary_sampler = sampler_info
+                    primary_sampler_id = node_id
+                
+        return primary_sampler_id, primary_sampler
+    
+    @staticmethod
+    def trace_node_input(prompt, node_id, input_name, target_class=None, max_depth=10):
+        """
+        Trace an input connection from a node to find the source node
+        
+        Parameters:
+        - prompt: The prompt object containing node connections
+        - node_id: ID of the starting node
+        - input_name: Name of the input to trace
+        - target_class: Optional class name to search for (e.g., "CLIPTextEncode")
+        - max_depth: Maximum depth to follow the node chain to prevent infinite loops
+        
+        Returns:
+        - node_id of the found node, or None if not found
+        """
+        if not prompt or not prompt.original_prompt or node_id not in prompt.original_prompt:
+            return None
+            
+        # For depth tracking
+        current_depth = 0
+        
+        current_node_id = node_id
+        current_input = input_name
+        
+        # If we're just tracing to origin (no target_class), keep track of the last valid node
+        last_valid_node = None
+        
+        while current_depth < max_depth:
+            if current_node_id not in prompt.original_prompt:
+                return last_valid_node if not target_class else None
+                
+            node_inputs = prompt.original_prompt[current_node_id].get("inputs", {})
+            if current_input not in node_inputs:
+                # We've reached a node without the specified input - this is our origin node
+                # if we're not looking for a specific target_class
+                return current_node_id if not target_class else None
+                
+            input_value = node_inputs[current_input]
+            # Input connections are formatted as [node_id, output_index]
+            if isinstance(input_value, list) and len(input_value) >= 2:
+                found_node_id = input_value[0]  # Connected node_id
+                
+                # If we're looking for a specific node class
+                if target_class:
+                    if found_node_id not in prompt.original_prompt:
+                        return None
+                    if prompt.original_prompt[found_node_id].get("class_type") == target_class:
+                        return found_node_id
+                
+                # If we're not looking for a specific class, update the last valid node
+                if not target_class:
+                    last_valid_node = found_node_id
+                
+                # Continue tracing through intermediate nodes
+                current_node_id = found_node_id
+                
+                # Check if current source node exists
+                if current_node_id not in prompt.original_prompt:
+                    return found_node_id if not target_class else None
+                    
+                # Determine which input to follow next on the source node
+                source_node_inputs = prompt.original_prompt[current_node_id].get("inputs", {})
+                if input_name in source_node_inputs:
+                    current_input = input_name
+                elif "conditioning" in source_node_inputs:
+                    current_input = "conditioning"
+                else:
+                    # If there's no suitable input to follow, return the current node
+                    # if we're not looking for a specific target_class
+                    return found_node_id if not target_class else None
+            else:
+                # We've reached a node with no further connections
+                return last_valid_node if not target_class else None
+            
+            current_depth += 1
+            
+        # If we've reached max depth without finding target_class
+        return last_valid_node if not target_class else None
+    
+    @staticmethod
+    def trace_model_path(metadata, prompt, start_node_id):
+        """
+        Trace the model connection path upstream to find the checkpoint
+        """
+        if not prompt or not prompt.original_prompt:
+            return None
+            
+        current_node_id = start_node_id
+        depth = 0
+        max_depth = 50
+        
+        while depth < max_depth:
+            # Check if current node is a registered checkpoint in our metadata
+            # This handles cached nodes correctly because metadata contains info for all nodes in the graph
+            if current_node_id in metadata.get(MODELS, {}):
+                if metadata[MODELS][current_node_id].get("type") == "checkpoint":
+                    return current_node_id
+            
+            if current_node_id not in prompt.original_prompt:
+                return None
+                
+            node = prompt.original_prompt[current_node_id]
+            inputs = node.get("inputs", {})
+            class_type = node.get("class_type", "")
+            
+            # Determine which input to follow next
+            next_input_name = "model"
+            
+            # Special handling for initial node
+            if depth == 0:
+                if class_type == "SamplerCustomAdvanced":
+                    next_input_name = "guider"
+            
+            # If the specific input doesn't exist, try generic 'model' 
+            if next_input_name not in inputs:
+                if "model" in inputs:
+                    next_input_name = "model"
+                elif "basic_pipe" in inputs:
+                    # Handle pipe nodes like FromBasicPipe by following the pipeline
+                    next_input_name = "basic_pipe"
+                else:
+                    # Dead end - no model input to follow
+                    return None
+            
+            # Get connected node
+            input_val = inputs[next_input_name]
+            if isinstance(input_val, list) and len(input_val) > 0:
+                current_node_id = input_val[0]
+            else:
+                return None
+                
+            depth += 1
+            
+        return None
+
+    @staticmethod
+    def find_primary_checkpoint(metadata, downstream_id=None, primary_sampler_id=None):
+        """
+        Find the primary checkpoint model in the workflow
+        
+        Parameters:
+        - metadata: The workflow metadata
+        - downstream_id: Optional ID of a downstream node to help identify the specific primary sampler
+        - primary_sampler_id: Optional ID of the primary sampler if already known
+        """
+        if not metadata.get(MODELS):
+            return None
+        
+        # Method 1: Topology-based tracing (More accurate for complex workflows)
+        # First, find the primary sampler if not provided
+        if not primary_sampler_id:
+            primary_sampler_id, _ = MetadataProcessor.find_primary_sampler(metadata, downstream_id)
+        
+        if primary_sampler_id:
+            prompt = metadata.get("current_prompt")
+            if prompt:
+                # Trace back from the sampler to find the checkpoint
+                checkpoint_id = MetadataProcessor.trace_model_path(metadata, prompt, primary_sampler_id)
+                if checkpoint_id and checkpoint_id in metadata.get(MODELS, {}):
+                    return metadata[MODELS][checkpoint_id].get("name")
+            
+        # Method 2: Fallback to the first available checkpoint (Original behavior)
+        # In most simple workflows, there's only one checkpoint, so we can just take the first one
+        for node_id, model_info in metadata.get(MODELS, {}).items():
+            if model_info.get("type") == "checkpoint":
+                return model_info.get("name")
+                
+        return None
+    
+    @staticmethod
+    def match_conditioning_to_prompts(metadata, sampler_id):
+        """
+        Match conditioning objects from a sampler to prompts in metadata
+        
+        Parameters:
+        - metadata: The workflow metadata
+        - sampler_id: ID of the sampler node to match
+        
+        Returns:
+        - Dictionary with 'prompt' and 'negative_prompt' if found
+        """
+        result = {
+            "prompt": "",
+            "negative_prompt": ""
+        }
+        
+        # Check if we have stored conditioning objects for this sampler
+        if sampler_id in metadata.get(PROMPTS, {}) and (
+            "pos_conditioning" in metadata[PROMPTS][sampler_id] or 
+            "neg_conditioning" in metadata[PROMPTS][sampler_id]):
+            
+            pos_conditioning = metadata[PROMPTS][sampler_id].get("pos_conditioning")
+            neg_conditioning = metadata[PROMPTS][sampler_id].get("neg_conditioning")
+            
+            # Helper function to recursively find prompt text for a conditioning object
+            def find_prompt_text_for_conditioning(conditioning_obj, is_positive=True):
+                if conditioning_obj is None:
+                    return ""
+                    
+                # Try to match conditioning objects with those stored by extractors
+                for prompt_node_id, prompt_data in metadata[PROMPTS].items():
+                    # For nodes with single conditioning output
+                    if "conditioning" in prompt_data:
+                        if id(prompt_data["conditioning"]) == id(conditioning_obj):
+                            return prompt_data.get("text", "")
+                    
+                    # For nodes with separate pos_conditioning and neg_conditioning outputs (like TSC_EfficientLoader)
+                    if is_positive and "positive_encoded" in prompt_data:
+                        if id(prompt_data["positive_encoded"]) == id(conditioning_obj):
+                            if "positive_text" in prompt_data:
+                                return prompt_data["positive_text"]
+                            else:
+                                orig_conditioning = prompt_data.get("orig_pos_cond", None)
+                                if orig_conditioning is not None:
+                                    # Recursively find the prompt text for the original conditioning
+                                    return find_prompt_text_for_conditioning(orig_conditioning, is_positive=True)
+                    
+                    if not is_positive and "negative_encoded" in prompt_data:
+                        if id(prompt_data["negative_encoded"]) == id(conditioning_obj):
+                            if "negative_text" in prompt_data:
+                                return prompt_data["negative_text"]
+                            else:
+                                orig_conditioning = prompt_data.get("orig_neg_cond", None)
+                                if orig_conditioning is not None:
+                                    # Recursively find the prompt text for the original conditioning
+                                    return find_prompt_text_for_conditioning(orig_conditioning, is_positive=False)
+                
+                return ""
+            
+            # Find prompt texts using the helper function
+            result["prompt"] = find_prompt_text_for_conditioning(pos_conditioning, is_positive=True)
+            result["negative_prompt"] = find_prompt_text_for_conditioning(neg_conditioning, is_positive=False)
+            
+        return result
+    
+    @staticmethod
+    def extract_generation_params(metadata, id=None):
+        """
+        Extract generation parameters from metadata using node relationships
+        
+        Parameters:
+        - metadata: The workflow metadata
+        - id: Optional ID of a downstream node to help identify the specific primary sampler
+        """
+        params = {
+            "prompt": "",
+            "negative_prompt": "",
+            "seed": None,
+            "steps": None,
+            "cfg_scale": None,
+            # "guidance": None,  # Add guidance parameter
+            "sampler": None,
+            "scheduler": None,
+            "checkpoint": None,
+            "loras": "",
+            "size": None,
+            "clip_skip": None
+        }
+        
+        # Get the prompt object for node relationship tracing
+        prompt = metadata.get("current_prompt")
+        
+        # Find the primary KSampler node
+        primary_sampler_id, primary_sampler = MetadataProcessor.find_primary_sampler(metadata, id)
+        
+        # Directly get checkpoint from metadata instead of tracing
+        # Pass primary_sampler_id to avoid redundant calculation
+        checkpoint = MetadataProcessor.find_primary_checkpoint(metadata, id, primary_sampler_id)
+        if checkpoint:
+            params["checkpoint"] = checkpoint
+        
+        # Check if guidance parameter exists in any sampling node
+        for node_id, sampler_info in metadata.get(SAMPLING, {}).items():
+            parameters = sampler_info.get("parameters", {})
+            if "guidance" in parameters and parameters["guidance"] is not None:
+                params["guidance"] = parameters["guidance"]
+                break
+        
+        if primary_sampler:
+            # Extract sampling parameters
+            sampling_params = primary_sampler.get("parameters", {})
+            # Handle both seed and noise_seed
+            params["seed"] = sampling_params.get("seed") if sampling_params.get("seed") is not None else sampling_params.get("noise_seed")
+            params["steps"] = sampling_params.get("steps")
+            params["cfg_scale"] = sampling_params.get("cfg")
+            params["sampler"] = sampling_params.get("sampler_name")
+            params["scheduler"] = sampling_params.get("scheduler")
+            
+            if prompt and primary_sampler_id:
+                # Check if this is a SamplerCustomAdvanced node
+                is_custom_advanced = False
+                if prompt.original_prompt and primary_sampler_id in prompt.original_prompt:
+                    is_custom_advanced = prompt.original_prompt[primary_sampler_id].get("class_type") == "SamplerCustomAdvanced"
+                
+                if is_custom_advanced:
+                    # For SamplerCustomAdvanced, use the new handler method
+                    MetadataProcessor.handle_custom_advanced_sampler(metadata, prompt, primary_sampler_id, params)
+                
+                else:
+                    # For standard samplers, match conditioning objects to prompts
+                    prompt_results = MetadataProcessor.match_conditioning_to_prompts(metadata, primary_sampler_id)
+                    params["prompt"] = prompt_results["prompt"]
+                    params["negative_prompt"] = prompt_results["negative_prompt"]
+
+                    # If prompts were still not found, fall back to tracing connections
+                    if not params["prompt"]:
+                        # Original tracing for standard samplers
+                        # Trace positive prompt - look specifically for CLIPTextEncode
+                        positive_node_id = MetadataProcessor.trace_node_input(prompt, primary_sampler_id, "positive", max_depth=10)
+                        if positive_node_id and positive_node_id in metadata.get(PROMPTS, {}):
+                            params["prompt"] = metadata[PROMPTS][positive_node_id].get("text", "")
+                        else:
+                            # If CLIPTextEncode is not found, try to find CLIPTextEncodeFlux
+                            positive_flux_node_id = MetadataProcessor.trace_node_input(prompt, primary_sampler_id, "positive", "CLIPTextEncodeFlux", max_depth=10)
+                            if positive_flux_node_id and positive_flux_node_id in metadata.get(PROMPTS, {}):
+                                params["prompt"] = metadata[PROMPTS][positive_flux_node_id].get("text", "")
+                        
+                        # Trace negative prompt - look specifically for CLIPTextEncode
+                        negative_node_id = MetadataProcessor.trace_node_input(prompt, primary_sampler_id, "negative", max_depth=10)
+                        if negative_node_id and negative_node_id in metadata.get(PROMPTS, {}):
+                            params["negative_prompt"] = metadata[PROMPTS][negative_node_id].get("text", "")
+                    
+                    # For SamplerCustom, handle any additional parameters
+                    MetadataProcessor.handle_custom_advanced_sampler(metadata, prompt, primary_sampler_id, params)
+                
+            # Size extraction is same for all sampler types
+            # Check if the sampler itself has size information (from latent_image)
+            if primary_sampler_id in metadata.get(SIZE, {}):
+                width = metadata[SIZE][primary_sampler_id].get("width")
+                height = metadata[SIZE][primary_sampler_id].get("height")
+                if width and height:
+                    params["size"] = f"{width}x{height}"
+        
+        # Extract LoRAs using the standardized format
+        lora_parts = []
+        for node_id, lora_info in metadata.get(LORAS, {}).items():
+            # Access the lora_list from the standardized format
+            lora_list = lora_info.get("lora_list", [])
+            for lora in lora_list:
+                name = lora.get("name", "unknown")
+                strength = lora.get("strength", 1.0)
+                lora_parts.append(f"<lora:{name}:{strength}>")
+        
+        params["loras"] = " ".join(lora_parts)
+        
+        # Set default clip_skip value
+        params["clip_skip"] = "1"  # Common default
+        
+        return params
+    
+    @staticmethod
+    def to_dict(metadata, id=None):
+        """
+        Convert extracted metadata to the ComfyUI output.json format
+        
+        Parameters:
+        - metadata: The workflow metadata
+        - id: Optional ID of a downstream node to help identify the specific primary sampler
+        """              
+        if standalone_mode:
+            # Return empty dictionary in standalone mode
+            return {}
+        
+        params = MetadataProcessor.extract_generation_params(metadata, id)
+        
+        # Convert all values to strings to match output.json format
+        for key in params:
+            if params[key] is not None:
+                params[key] = str(params[key])
+        
+        return params
+    
+    @staticmethod
+    def to_json(metadata, id=None):
+        """Convert metadata to JSON string"""
+        params = MetadataProcessor.to_dict(metadata, id)
+        return json.dumps(params, indent=4)
+    
+    @staticmethod
+    def handle_custom_advanced_sampler(metadata, prompt, primary_sampler_id, params):
+        """
+        Handle parameter extraction for SamplerCustomAdvanced nodes
+        
+        Parameters:
+        - metadata: The workflow metadata
+        - prompt: The prompt object containing node connections
+        - primary_sampler_id: ID of the SamplerCustomAdvanced node
+        - params: Parameters dictionary to update
+        """
+        if not prompt.original_prompt or primary_sampler_id not in prompt.original_prompt:
+            return
+            
+        sampler_inputs = prompt.original_prompt[primary_sampler_id].get("inputs", {})
+        
+        # 1. Trace sigmas input to find BasicScheduler (only if sigmas input exists)
+        if "sigmas" in sampler_inputs:
+            scheduler_node_id = MetadataProcessor.trace_node_input(prompt, primary_sampler_id, "sigmas", None, max_depth=5)
+            if scheduler_node_id and scheduler_node_id in metadata.get(SAMPLING, {}):
+                scheduler_params = metadata[SAMPLING][scheduler_node_id].get("parameters", {})
+                params["steps"] = scheduler_params.get("steps")
+                params["scheduler"] = scheduler_params.get("scheduler")
+                params["denoise"] = scheduler_params.get("denoise")
+        
+        # 2. Trace sampler input to find KSamplerSelect (only if sampler input exists)
+        if "sampler" in sampler_inputs:
+            sampler_node_id = MetadataProcessor.trace_node_input(prompt, primary_sampler_id, "sampler", "KSamplerSelect", max_depth=5)
+            if sampler_node_id and sampler_node_id in metadata.get(SAMPLING, {}):
+                sampler_params = metadata[SAMPLING][sampler_node_id].get("parameters", {})
+                params["sampler"] = sampler_params.get("sampler_name")
+        
+        # 3. Trace guider input for CFGGuider and CLIPTextEncode
+        if "guider" in sampler_inputs:
+            guider_node_id = MetadataProcessor.trace_node_input(prompt, primary_sampler_id, "guider", max_depth=5)
+            if guider_node_id and guider_node_id in prompt.original_prompt:
+                # Check if the guider node is a CFGGuider
+                if prompt.original_prompt[guider_node_id].get("class_type") == "CFGGuider":
+                    # Extract cfg value from the CFGGuider
+                    if guider_node_id in metadata.get(SAMPLING, {}):
+                        cfg_params = metadata[SAMPLING][guider_node_id].get("parameters", {})
+                        params["cfg_scale"] = cfg_params.get("cfg")
+                    
+                    # Find CLIPTextEncode for positive prompt
+                    positive_node_id = MetadataProcessor.trace_node_input(prompt, guider_node_id, "positive", "CLIPTextEncode", max_depth=10)
+                    if positive_node_id and positive_node_id in metadata.get(PROMPTS, {}):
+                        params["prompt"] = metadata[PROMPTS][positive_node_id].get("text", "")
+                    
+                    # Find CLIPTextEncode for negative prompt
+                    negative_node_id = MetadataProcessor.trace_node_input(prompt, guider_node_id, "negative", "CLIPTextEncode", max_depth=10)
+                    if negative_node_id and negative_node_id in metadata.get(PROMPTS, {}):
+                        params["negative_prompt"] = metadata[PROMPTS][negative_node_id].get("text", "")
+                else:
+                    positive_node_id = MetadataProcessor.trace_node_input(prompt, guider_node_id, "conditioning", max_depth=10)
+                    if positive_node_id and positive_node_id in metadata.get(PROMPTS, {}):
+                        params["prompt"] = metadata[PROMPTS][positive_node_id].get("text", "")

py/metadata_collector/metadata_registry.py

@@ -0,0 +1,277 @@
+import time
+from nodes import NODE_CLASS_MAPPINGS
+from .node_extractors import NODE_EXTRACTORS, GenericNodeExtractor
+from .constants import METADATA_CATEGORIES, IMAGES
+
+class MetadataRegistry:
+    """A singleton registry to store and retrieve workflow metadata"""
+    _instance = None
+    
+    def __new__(cls):
+        if cls._instance is None:
+            cls._instance = super().__new__(cls)
+            cls._instance._reset()
+        return cls._instance
+    
+    def _reset(self):
+        self.current_prompt_id = None
+        self.current_prompt = None
+        self.metadata = {}
+        self.prompt_metadata = {}
+        self.executed_nodes = set()
+        
+        # Node-level cache for metadata
+        self.node_cache = {}
+        
+        # Limit the number of stored prompts
+        self.max_prompt_history = 3
+        
+        # Categories we want to track and retrieve from cache
+        self.metadata_categories = METADATA_CATEGORIES
+    
+    def _clean_old_prompts(self):
+        """Clean up old prompt metadata, keeping only recent ones"""
+        if len(self.prompt_metadata) <= self.max_prompt_history:
+            return
+            
+        # Sort all prompt_ids by timestamp
+        sorted_prompts = sorted(
+            self.prompt_metadata.keys(),
+            key=lambda pid: self.prompt_metadata[pid].get("timestamp", 0)
+        )
+        
+        # Remove oldest records
+        prompts_to_remove = sorted_prompts[:len(sorted_prompts) - self.max_prompt_history]
+        for pid in prompts_to_remove:
+            del self.prompt_metadata[pid]
+    
+    def start_collection(self, prompt_id):
+        """Begin metadata collection for a new prompt"""
+        self.current_prompt_id = prompt_id
+        self.executed_nodes = set()
+        self.prompt_metadata[prompt_id] = {
+            category: {} for category in METADATA_CATEGORIES
+        }
+        # Add additional metadata fields
+        self.prompt_metadata[prompt_id].update({
+            "execution_order": [],
+            "current_prompt": None,  # Will store the prompt object
+            "timestamp": time.time()
+        })
+        
+        # Clean up old prompt data
+        self._clean_old_prompts()
+    
+    def set_current_prompt(self, prompt):
+        """Set the current prompt object reference"""
+        self.current_prompt = prompt
+        if self.current_prompt_id and self.current_prompt_id in self.prompt_metadata:
+            # Store the prompt in the metadata for later relationship tracing
+            self.prompt_metadata[self.current_prompt_id]["current_prompt"] = prompt
+        
+    def get_metadata(self, prompt_id=None):
+        """Get collected metadata for a prompt"""
+        key = prompt_id if prompt_id is not None else self.current_prompt_id
+        if key not in self.prompt_metadata:
+            return {}
+            
+        metadata = self.prompt_metadata[key]
+        
+        # If we have a current prompt object, check for non-executed nodes
+        prompt_obj = metadata.get("current_prompt")
+        if prompt_obj and hasattr(prompt_obj, "original_prompt"):
+            original_prompt = prompt_obj.original_prompt
+            
+            # Fill in missing metadata from cache for nodes that weren't executed
+            self._fill_missing_metadata(key, original_prompt)
+            
+        return self.prompt_metadata.get(key, {})
+    
+    def _fill_missing_metadata(self, prompt_id, original_prompt):
+        """Fill missing metadata from cache for non-executed nodes"""
+        if not original_prompt:
+            return
+            
+        executed_nodes = self.executed_nodes
+        metadata = self.prompt_metadata[prompt_id]
+        
+        # Iterate through nodes in the original prompt
+        for node_id, node_data in original_prompt.items():
+            # Skip if already executed in this run
+            if node_id in executed_nodes:
+                continue
+                
+            # Get the node type from the prompt (this is the key in NODE_CLASS_MAPPINGS)
+            prompt_class_type = node_data.get("class_type")
+            if not prompt_class_type:
+                continue
+                
+            # Convert to actual class name (which is what we use in our cache)
+            class_type = prompt_class_type
+            if prompt_class_type in NODE_CLASS_MAPPINGS:
+                class_obj = NODE_CLASS_MAPPINGS[prompt_class_type]
+                class_type = class_obj.__name__
+            
+            # Create cache key using the actual class name
+            cache_key = f"{node_id}:{class_type}"
+            
+            # Check if this node type is relevant for metadata collection
+            if class_type in NODE_EXTRACTORS:
+                # Check if we have cached metadata for this node
+                if cache_key in self.node_cache:
+                    cached_data = self.node_cache[cache_key]
+                    
+                    # Apply cached metadata to the current metadata
+                    for category in self.metadata_categories:
+                        if category in cached_data and node_id in cached_data[category]:
+                            if node_id not in metadata[category]:
+                                metadata[category][node_id] = cached_data[category][node_id]
+    
+    def record_node_execution(self, node_id, class_type, inputs, outputs):
+        """Record information about a node's execution"""
+        if not self.current_prompt_id:
+            return
+            
+        # Add to execution order and mark as executed
+        if node_id not in self.executed_nodes:
+            self.executed_nodes.add(node_id)
+            self.prompt_metadata[self.current_prompt_id]["execution_order"].append(node_id)
+        
+        # Process inputs to simplify working with them
+        processed_inputs = {}
+        for input_name, input_values in inputs.items():
+            if isinstance(input_values, list) and len(input_values) > 0:
+                # For single values, just use the first one (most common case)
+                processed_inputs[input_name] = input_values[0]
+            else:
+                processed_inputs[input_name] = input_values
+            
+        # Extract node-specific metadata
+        extractor = NODE_EXTRACTORS.get(class_type, GenericNodeExtractor)
+        extractor.extract(
+            node_id, 
+            processed_inputs, 
+            outputs, 
+            self.prompt_metadata[self.current_prompt_id]
+        )
+        
+        # Cache this node's metadata
+        self._cache_node_metadata(node_id, class_type)
+    
+    def update_node_execution(self, node_id, class_type, outputs):
+        """Update node metadata with output information"""
+        if not self.current_prompt_id:
+            return
+        
+        # Process outputs to make them more usable
+        processed_outputs = outputs
+            
+        # Use the same extractor to update with outputs
+        extractor = NODE_EXTRACTORS.get(class_type, GenericNodeExtractor)
+        if hasattr(extractor, 'update'):
+            extractor.update(
+                node_id, 
+                processed_outputs, 
+                self.prompt_metadata[self.current_prompt_id]
+            )
+            
+        # Update the cached metadata for this node
+        self._cache_node_metadata(node_id, class_type)
+            
+    def _cache_node_metadata(self, node_id, class_type):
+        """Cache the metadata for a specific node"""
+        if not self.current_prompt_id or not node_id or not class_type:
+            return
+            
+        # Create a cache key combining node_id and class_type
+        cache_key = f"{node_id}:{class_type}"
+        
+        # Create a shallow copy of the node's metadata
+        node_metadata = {}
+        current_metadata = self.prompt_metadata[self.current_prompt_id]
+        
+        for category in self.metadata_categories:
+            if category in current_metadata and node_id in current_metadata[category]:
+                if category not in node_metadata:
+                    node_metadata[category] = {}
+                node_metadata[category][node_id] = current_metadata[category][node_id]
+        
+        # Save new metadata or clear stale cache entries when metadata is empty
+        if any(node_metadata.values()):
+            self.node_cache[cache_key] = node_metadata
+        else:
+            self.node_cache.pop(cache_key, None)
+    
+    def clear_unused_cache(self):
+        """Clean up node_cache entries that are no longer in use"""
+        # Collect all node_ids currently in prompt_metadata
+        active_node_ids = set()
+        for prompt_data in self.prompt_metadata.values():
+            for category in self.metadata_categories:
+                if category in prompt_data:
+                    active_node_ids.update(prompt_data[category].keys())
+        
+        # Find cache keys that are no longer needed
+        keys_to_remove = []
+        for cache_key in self.node_cache:
+            node_id = cache_key.split(':')[0]
+            if node_id not in active_node_ids:
+                keys_to_remove.append(cache_key)
+        
+        # Remove cache entries that are no longer needed
+        for key in keys_to_remove:
+            del self.node_cache[key]
+    
+    def clear_metadata(self, prompt_id=None):
+        """Clear metadata for a specific prompt or reset all data"""
+        if prompt_id is not None:
+            if prompt_id in self.prompt_metadata:
+                del self.prompt_metadata[prompt_id]
+                # Clean up cache after removing prompt
+                self.clear_unused_cache()
+        else:
+            # Reset all data
+            self._reset()
+    
+    def get_first_decoded_image(self, prompt_id=None):
+        """Get the first decoded image result"""
+        key = prompt_id if prompt_id is not None else self.current_prompt_id
+        if key not in self.prompt_metadata:
+            return None
+            
+        metadata = self.prompt_metadata[key]
+        if IMAGES in metadata and "first_decode" in metadata[IMAGES]:
+            image_data = metadata[IMAGES]["first_decode"]["image"]
+            
+            # If it's an image batch or tuple, handle various formats
+            if isinstance(image_data, (list, tuple)) and len(image_data) > 0:
+                # Return first element of list/tuple
+                return image_data[0]
+            
+            # If it's a tensor, return as is for processing in the route handler
+            return image_data
+        
+        # If no image is found in the current metadata, try to find it in the cache
+        # This handles the case where VAEDecode was cached by ComfyUI and not executed
+        prompt_obj = metadata.get("current_prompt")
+        if prompt_obj and hasattr(prompt_obj, "original_prompt"):
+            original_prompt = prompt_obj.original_prompt
+            for node_id, node_data in original_prompt.items():
+                class_type = node_data.get("class_type")
+                if class_type and class_type in NODE_CLASS_MAPPINGS:
+                    class_obj = NODE_CLASS_MAPPINGS[class_type]
+                    class_name = class_obj.__name__
+                    # Check if this is a VAEDecode node
+                    if class_name == "VAEDecode":
+                        # Try to find this node in the cache
+                        cache_key = f"{node_id}:{class_name}"
+                        if cache_key in self.node_cache:
+                            cached_data = self.node_cache[cache_key]
+                            if IMAGES in cached_data and node_id in cached_data[IMAGES]:
+                                image_data = cached_data[IMAGES][node_id]["image"]
+                                # Handle different image formats
+                                if isinstance(image_data, (list, tuple)) and len(image_data) > 0:
+                                    return image_data[0]
+                                return image_data
+        
+        return None

py/metadata_collector/node_extractors.py

@@ -0,0 +1,735 @@
+import os
+
+from .constants import MODELS, PROMPTS, SAMPLING, LORAS, SIZE, IMAGES, IS_SAMPLER
+
+
+def _store_checkpoint_metadata(metadata, node_id, model_name):
+    """Store checkpoint model information when available."""
+    if not model_name:
+        return
+    metadata.setdefault(MODELS, {})
+    metadata[MODELS][node_id] = {
+        "name": model_name,
+        "type": "checkpoint",
+        "node_id": node_id
+    }
+
+
+class NodeMetadataExtractor:
+    """Base class for node-specific metadata extraction"""
+    
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        """Extract metadata from node inputs/outputs"""
+        pass
+        
+    @staticmethod
+    def update(node_id, outputs, metadata):
+        """Update metadata with node outputs after execution"""
+        pass
+        
+class GenericNodeExtractor(NodeMetadataExtractor):
+    """Default extractor for nodes without specific handling"""
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        pass
+        
+class CheckpointLoaderExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "ckpt_name" not in inputs:
+            return
+            
+        model_name = inputs.get("ckpt_name")
+        _store_checkpoint_metadata(metadata, node_id, model_name)
+
+
+class NunchakuFluxDiTLoaderExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "model_path" not in inputs:
+            return
+
+        model_name = inputs.get("model_path")
+        _store_checkpoint_metadata(metadata, node_id, model_name)
+
+
+class NunchakuQwenImageDiTLoaderExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "model_name" not in inputs:
+            return
+
+        model_name = inputs.get("model_name")
+        _store_checkpoint_metadata(metadata, node_id, model_name)
+
+class GGUFLoaderExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "gguf_name" not in inputs:
+            return
+
+        model_name = inputs.get("gguf_name")
+        _store_checkpoint_metadata(metadata, node_id, model_name)
+
+
+class KJNodesModelLoaderExtractor(NodeMetadataExtractor):
+    """Extract metadata from KJNodes loaders that expose `model_name`."""
+
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "model_name" not in inputs:
+            return
+
+        model_name = inputs.get("model_name")
+        _store_checkpoint_metadata(metadata, node_id, model_name)
+
+class TSCCheckpointLoaderExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "ckpt_name" not in inputs:
+            return
+            
+        model_name = inputs.get("ckpt_name")
+        _store_checkpoint_metadata(metadata, node_id, model_name)
+
+        # For loader node has lora_stack input, like Efficient Loader from Efficient Nodes
+        active_loras = []
+        
+        # Process lora_stack if available
+        if "lora_stack" in inputs:
+            lora_stack = inputs.get("lora_stack", [])
+            for lora_path, model_strength, clip_strength in lora_stack:
+                # Extract lora name from path (following the format in lora_loader.py)
+                lora_name = os.path.splitext(os.path.basename(lora_path))[0]
+                active_loras.append({
+                    "name": lora_name,
+                    "strength": model_strength
+                })
+        
+        if active_loras:
+            metadata[LORAS][node_id] = {
+                "lora_list": active_loras,
+                "node_id": node_id
+            }
+        
+        # Extract positive and negative prompt text if available
+        positive_text = inputs.get("positive", "")
+        negative_text = inputs.get("negative", "")
+        
+        if positive_text or negative_text:
+            if node_id not in metadata[PROMPTS]:
+                metadata[PROMPTS][node_id] = {"node_id": node_id}
+            
+            # Store both positive and negative text
+            metadata[PROMPTS][node_id]["positive_text"] = positive_text
+            metadata[PROMPTS][node_id]["negative_text"] = negative_text
+            
+    @staticmethod
+    def update(node_id, outputs, metadata):
+        # Handle conditioning outputs from TSC_EfficientLoader
+        # outputs is a list with [(model, positive_encoded, negative_encoded, {"samples":latent}, vae, clip, dependencies,)]
+        if outputs and isinstance(outputs, list) and len(outputs) > 0:
+            first_output = outputs[0]
+            if isinstance(first_output, tuple) and len(first_output) >= 3:
+                positive_conditioning = first_output[1]
+                negative_conditioning = first_output[2]
+                
+                # Save both conditioning objects in metadata
+                if node_id not in metadata[PROMPTS]:
+                    metadata[PROMPTS][node_id] = {"node_id": node_id}
+                    
+                metadata[PROMPTS][node_id]["positive_encoded"] = positive_conditioning
+                metadata[PROMPTS][node_id]["negative_encoded"] = negative_conditioning
+
+class CLIPTextEncodeExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "text" not in inputs:
+            return
+            
+        text = inputs.get("text", "")
+        metadata[PROMPTS][node_id] = {
+            "text": text,
+            "node_id": node_id
+        }
+
+    @staticmethod
+    def update(node_id, outputs, metadata):
+        if outputs and isinstance(outputs, list) and len(outputs) > 0:
+            if isinstance(outputs[0], tuple) and len(outputs[0]) > 0:
+                conditioning = outputs[0][0]
+                metadata[PROMPTS][node_id]["conditioning"] = conditioning
+
+# Base Sampler Extractor to reduce code redundancy
+class BaseSamplerExtractor(NodeMetadataExtractor):
+    """Base extractor for sampler nodes with common functionality"""
+    @staticmethod
+    def extract_sampling_params(node_id, inputs, metadata, param_keys):
+        """Extract sampling parameters from inputs"""
+        sampling_params = {}
+        for key in param_keys:
+            if key in inputs:
+                sampling_params[key] = inputs[key]
+                
+        metadata[SAMPLING][node_id] = {
+            "parameters": sampling_params,
+            "node_id": node_id,
+            IS_SAMPLER: True  # Add sampler flag
+        }
+    
+    @staticmethod
+    def extract_conditioning(node_id, inputs, metadata):
+        """Extract conditioning objects from inputs"""
+        # Store the conditioning objects directly in metadata for later matching
+        pos_conditioning = inputs.get("positive", None)
+        neg_conditioning = inputs.get("negative", None)
+
+        # Save conditioning objects in metadata for later matching
+        if pos_conditioning is not None or neg_conditioning is not None:
+            if node_id not in metadata[PROMPTS]:
+                metadata[PROMPTS][node_id] = {"node_id": node_id}
+            
+            metadata[PROMPTS][node_id]["pos_conditioning"] = pos_conditioning
+            metadata[PROMPTS][node_id]["neg_conditioning"] = neg_conditioning
+    
+    @staticmethod
+    def extract_latent_dimensions(node_id, inputs, metadata):
+        """Extract dimensions from latent image"""
+        # Extract latent image dimensions if available
+        if "latent_image" in inputs and inputs["latent_image"] is not None:
+            latent = inputs["latent_image"]
+            if isinstance(latent, dict) and "samples" in latent:
+                # Extract dimensions from latent tensor
+                samples = latent["samples"]
+                if hasattr(samples, "shape") and len(samples.shape) >= 3:
+                    # Correct shape interpretation: [batch_size, channels, height/8, width/8]
+                    # Multiply by 8 to get actual pixel dimensions
+                    height = int(samples.shape[2] * 8)
+                    width = int(samples.shape[3] * 8)
+                    
+                    if SIZE not in metadata:
+                        metadata[SIZE] = {}
+                        
+                    metadata[SIZE][node_id] = {
+                        "width": width,
+                        "height": height,
+                        "node_id": node_id
+                    }
+        
+class SamplerExtractor(BaseSamplerExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs:
+            return
+            
+        # Extract common sampling parameters
+        BaseSamplerExtractor.extract_sampling_params(
+            node_id, inputs, metadata, 
+            ["seed", "steps", "cfg", "sampler_name", "scheduler", "denoise"]
+        )
+        
+        # Extract conditioning objects
+        BaseSamplerExtractor.extract_conditioning(node_id, inputs, metadata)
+        
+        # Extract latent dimensions
+        BaseSamplerExtractor.extract_latent_dimensions(node_id, inputs, metadata)
+
+class KSamplerAdvancedExtractor(BaseSamplerExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs:
+            return
+            
+        # Extract common sampling parameters
+        BaseSamplerExtractor.extract_sampling_params(
+            node_id, inputs, metadata, 
+            ["noise_seed", "steps", "cfg", "sampler_name", "scheduler", "add_noise"]
+        )
+        
+        # Extract conditioning objects
+        BaseSamplerExtractor.extract_conditioning(node_id, inputs, metadata)
+        
+        # Extract latent dimensions
+        BaseSamplerExtractor.extract_latent_dimensions(node_id, inputs, metadata)
+
+class KSamplerBasicPipeExtractor(BaseSamplerExtractor):
+    """Extractor for KSamplerBasicPipe and KSampler_inspire_pipe nodes"""
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs:
+            return
+            
+        # Extract common sampling parameters
+        BaseSamplerExtractor.extract_sampling_params(
+            node_id, inputs, metadata, 
+            ["seed", "steps", "cfg", "sampler_name", "scheduler", "denoise"]
+        )
+        
+        # Extract conditioning objects from basic_pipe
+        if "basic_pipe" in inputs and inputs["basic_pipe"] is not None:
+            basic_pipe = inputs["basic_pipe"]
+            # Typically, basic_pipe structure is (model, clip, vae, positive, negative)
+            if isinstance(basic_pipe, tuple) and len(basic_pipe) >= 5:
+                pos_conditioning = basic_pipe[3]  # positive is at index 3
+                neg_conditioning = basic_pipe[4]  # negative is at index 4
+                
+                # Save conditioning objects in metadata
+                if node_id not in metadata[PROMPTS]:
+                    metadata[PROMPTS][node_id] = {"node_id": node_id}
+                
+                metadata[PROMPTS][node_id]["pos_conditioning"] = pos_conditioning
+                metadata[PROMPTS][node_id]["neg_conditioning"] = neg_conditioning
+        
+        # Extract latent dimensions
+        BaseSamplerExtractor.extract_latent_dimensions(node_id, inputs, metadata)
+
+class KSamplerAdvancedBasicPipeExtractor(BaseSamplerExtractor):
+    """Extractor for KSamplerAdvancedBasicPipe nodes"""
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs:
+            return
+            
+        # Extract common sampling parameters
+        BaseSamplerExtractor.extract_sampling_params(
+            node_id, inputs, metadata, 
+            ["noise_seed", "steps", "cfg", "sampler_name", "scheduler", "add_noise"]
+        )
+        
+        # Extract conditioning objects from basic_pipe
+        if "basic_pipe" in inputs and inputs["basic_pipe"] is not None:
+            basic_pipe = inputs["basic_pipe"]
+            # Typically, basic_pipe structure is (model, clip, vae, positive, negative)
+            if isinstance(basic_pipe, tuple) and len(basic_pipe) >= 5:
+                pos_conditioning = basic_pipe[3]  # positive is at index 3
+                neg_conditioning = basic_pipe[4]  # negative is at index 4
+                
+                # Save conditioning objects in metadata
+                if node_id not in metadata[PROMPTS]:
+                    metadata[PROMPTS][node_id] = {"node_id": node_id}
+                
+                metadata[PROMPTS][node_id]["pos_conditioning"] = pos_conditioning
+                metadata[PROMPTS][node_id]["neg_conditioning"] = neg_conditioning
+        
+        # Extract latent dimensions
+        BaseSamplerExtractor.extract_latent_dimensions(node_id, inputs, metadata)
+
+class TSCSamplerBaseExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        # Store vae_decode setting for later use in update
+        if inputs and "vae_decode" in inputs:
+            if SAMPLING not in metadata:
+                metadata[SAMPLING] = {}
+                
+            if node_id not in metadata[SAMPLING]:
+                metadata[SAMPLING][node_id] = {"parameters": {}, "node_id": node_id}
+                
+            # Store the vae_decode setting
+            metadata[SAMPLING][node_id]["vae_decode"] = inputs["vae_decode"]
+
+    @staticmethod
+    def update(node_id, outputs, metadata):
+        # Check if vae_decode was set to "true"
+        should_save_image = True
+        if SAMPLING in metadata and node_id in metadata[SAMPLING]:
+            vae_decode = metadata[SAMPLING][node_id].get("vae_decode")
+            if vae_decode is not None:
+                should_save_image = (vae_decode == "true")
+        
+        # Skip image saving if vae_decode isn't "true"
+        if not should_save_image:
+            return
+        
+        # Ensure IMAGES category exists
+        if IMAGES not in metadata:
+            metadata[IMAGES] = {}
+        
+        # Extract output_images from the TSC sampler format
+        # outputs = [{"ui": {"images": preview_images}, "result": result}]
+        # where result = (original_model, original_positive, original_negative, latent_list, optional_vae, output_images,)
+        if outputs and isinstance(outputs, list) and len(outputs) > 0:
+            # Get the first item in the list
+            output_item = outputs[0]
+            if isinstance(output_item, dict) and "result" in output_item:
+                result = output_item["result"]
+                if isinstance(result, tuple) and len(result) >= 6:
+                    # The output_images is the last element in the result tuple
+                    output_images = (result[5],)
+                    
+                    # Save image data under node ID index to be captured by caching mechanism
+                    metadata[IMAGES][node_id] = {
+                    "node_id": node_id,
+                    "image": output_images
+                    }
+                    
+                    # Only set first_decode if it hasn't been recorded yet
+                    if "first_decode" not in metadata[IMAGES]:
+                        metadata[IMAGES]["first_decode"] = metadata[IMAGES][node_id]
+
+class TSCKSamplerExtractor(SamplerExtractor, TSCSamplerBaseExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        # Call parent extract methods
+        SamplerExtractor.extract(node_id, inputs, outputs, metadata)
+        TSCSamplerBaseExtractor.extract(node_id, inputs, outputs, metadata)
+
+    # Update method is inherited from TSCSamplerBaseExtractor
+
+
+class TSCKSamplerAdvancedExtractor(KSamplerAdvancedExtractor, TSCSamplerBaseExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        # Call parent extract methods
+        KSamplerAdvancedExtractor.extract(node_id, inputs, outputs, metadata)
+        TSCSamplerBaseExtractor.extract(node_id, inputs, outputs, metadata)
+
+    # Update method is inherited from TSCSamplerBaseExtractor
+
+class LoraLoaderExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "lora_name" not in inputs:
+            return
+            
+        lora_name = inputs.get("lora_name")
+        # Extract base filename without extension from path
+        lora_name = os.path.splitext(os.path.basename(lora_name))[0]
+        strength_model = round(float(inputs.get("strength_model", 1.0)), 2)
+        
+        # Use the standardized format with lora_list
+        metadata[LORAS][node_id] = {
+            "lora_list": [
+                {
+                    "name": lora_name,
+                    "strength": strength_model
+                }
+            ],
+            "node_id": node_id
+        }
+
+class ImageSizeExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs:
+            return
+        
+        width = inputs.get("width", 512)
+        height = inputs.get("height", 512)
+        
+        if SIZE not in metadata:
+            metadata[SIZE] = {}
+            
+        metadata[SIZE][node_id] = {
+            "width": width,
+            "height": height,
+            "node_id": node_id
+        }
+
+class LoraLoaderManagerExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs:
+            return
+        
+        active_loras = []
+        
+        # Process lora_stack if available
+        if "lora_stack" in inputs:
+            lora_stack = inputs.get("lora_stack", [])
+            for lora_path, model_strength, clip_strength in lora_stack:
+                # Extract lora name from path (following the format in lora_loader.py)
+                lora_name = os.path.splitext(os.path.basename(lora_path))[0]
+                active_loras.append({
+                    "name": lora_name,
+                    "strength": model_strength
+                })
+        
+        # Process loras from inputs
+        if "loras" in inputs:
+            loras_data = inputs.get("loras", [])
+            
+            # Handle new format: {'loras': {'__value__': [...]}} 
+            if isinstance(loras_data, dict) and '__value__' in loras_data:
+                loras_list = loras_data['__value__']
+            # Handle old format: {'loras': [...]}
+            elif isinstance(loras_data, list):
+                loras_list = loras_data
+            else:
+                loras_list = []
+                
+            # Filter for active loras
+            for lora in loras_list:
+                if isinstance(lora, dict) and lora.get("active", True) and not lora.get("_isDummy", False):
+                    active_loras.append({
+                        "name": lora.get("name", ""),
+                        "strength": float(lora.get("strength", 1.0))
+                    })
+        
+        if active_loras:
+            metadata[LORAS][node_id] = {
+                "lora_list": active_loras,
+                "node_id": node_id
+            }
+
+class FluxGuidanceExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "guidance" not in inputs:
+            return
+            
+        guidance_value = inputs.get("guidance")
+        
+        # Store the guidance value in SAMPLING category
+        if node_id not in metadata[SAMPLING]:
+            metadata[SAMPLING][node_id] = {"parameters": {}, "node_id": node_id}
+            
+        metadata[SAMPLING][node_id]["parameters"]["guidance"] = guidance_value
+
+class UNETLoaderExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "unet_name" not in inputs:
+            return
+            
+        model_name = inputs.get("unet_name")
+        if model_name:
+            metadata[MODELS][node_id] = {
+                "name": model_name,
+                "type": "checkpoint",
+                "node_id": node_id
+            }
+
+class VAEDecodeExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        pass
+        
+    @staticmethod
+    def update(node_id, outputs, metadata):
+        # Ensure IMAGES category exists
+        if IMAGES not in metadata:
+            metadata[IMAGES] = {}
+            
+        # Save image data under node ID index to be captured by caching mechanism
+        metadata[IMAGES][node_id] = {
+            "node_id": node_id,
+            "image": outputs
+        }
+        
+        # Only set first_decode if it hasn't been recorded yet
+        if "first_decode" not in metadata[IMAGES]:
+            metadata[IMAGES]["first_decode"] = metadata[IMAGES][node_id]
+
+class KSamplerSelectExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "sampler_name" not in inputs:
+            return
+            
+        sampling_params = {}
+        if "sampler_name" in inputs:
+            sampling_params["sampler_name"] = inputs["sampler_name"]
+                
+        metadata[SAMPLING][node_id] = {
+            "parameters": sampling_params,
+            "node_id": node_id,
+            IS_SAMPLER: False  # Mark as non-primary sampler
+        }
+
+class BasicSchedulerExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs:
+            return
+            
+        sampling_params = {}
+        for key in ["scheduler", "steps", "denoise"]:
+            if key in inputs:
+                sampling_params[key] = inputs[key]
+                
+        metadata[SAMPLING][node_id] = {
+            "parameters": sampling_params,
+            "node_id": node_id,
+            IS_SAMPLER: False  # Mark as non-primary sampler
+        }
+
+class SamplerCustomAdvancedExtractor(BaseSamplerExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs:
+            return
+            
+        sampling_params = {}
+
+        # Handle noise.seed as seed
+        if "noise" in inputs and inputs["noise"] is not None and hasattr(inputs["noise"], "seed"):
+            noise = inputs["noise"]
+            sampling_params["seed"] = noise.seed
+                
+        metadata[SAMPLING][node_id] = {
+            "parameters": sampling_params,
+            "node_id": node_id,
+            IS_SAMPLER: True  # Add sampler flag
+        }
+        
+        # Extract latent dimensions
+        BaseSamplerExtractor.extract_latent_dimensions(node_id, inputs, metadata)
+
+import json
+
+class CLIPTextEncodeFluxExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "clip_l" not in inputs or "t5xxl" not in inputs:
+            return
+            
+        clip_l_text = inputs.get("clip_l", "")
+        t5xxl_text = inputs.get("t5xxl", "")
+        
+        # If both are empty, use empty string
+        if not clip_l_text and not t5xxl_text:
+            combined_text = ""
+        # If one is empty, use the non-empty one
+        elif not clip_l_text:
+            combined_text = t5xxl_text
+        elif not t5xxl_text:
+            combined_text = clip_l_text
+        # If both have content, use JSON format
+        else:
+            combined_text = json.dumps({
+                "T5": t5xxl_text,
+                "CLIP-L": clip_l_text
+            })
+        
+        metadata[PROMPTS][node_id] = {
+            "text": combined_text,
+            "node_id": node_id
+        }
+        
+        # Extract guidance value if available
+        if "guidance" in inputs:
+            guidance_value = inputs.get("guidance")
+            
+            # Store the guidance value in SAMPLING category
+            if SAMPLING not in metadata:
+                metadata[SAMPLING] = {}
+                
+            if node_id not in metadata[SAMPLING]:
+                metadata[SAMPLING][node_id] = {"parameters": {}, "node_id": node_id}
+                
+            metadata[SAMPLING][node_id]["parameters"]["guidance"] = guidance_value
+
+    @staticmethod
+    def update(node_id, outputs, metadata):
+        if outputs and isinstance(outputs, list) and len(outputs) > 0:
+            if isinstance(outputs[0], tuple) and len(outputs[0]) > 0:
+                conditioning = outputs[0][0]
+                metadata[PROMPTS][node_id]["conditioning"] = conditioning
+
+class CFGGuiderExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs or "cfg" not in inputs:
+            return
+            
+        cfg_value = inputs.get("cfg")
+        
+        # Store the cfg value in SAMPLING category
+        if SAMPLING not in metadata:
+            metadata[SAMPLING] = {}
+            
+        if node_id not in metadata[SAMPLING]:
+            metadata[SAMPLING][node_id] = {"parameters": {}, "node_id": node_id}
+            
+        metadata[SAMPLING][node_id]["parameters"]["cfg"] = cfg_value
+
+class CR_ApplyControlNetStackExtractor(NodeMetadataExtractor):
+    @staticmethod
+    def extract(node_id, inputs, outputs, metadata):
+        if not inputs:
+            return
+            
+        # Save the original conditioning inputs
+        base_positive = inputs.get("base_positive")
+        base_negative = inputs.get("base_negative")
+        
+        if base_positive is not None or base_negative is not None:
+            if node_id not in metadata[PROMPTS]:
+                metadata[PROMPTS][node_id] = {"node_id": node_id}
+            
+            metadata[PROMPTS][node_id]["orig_pos_cond"] = base_positive
+            metadata[PROMPTS][node_id]["orig_neg_cond"] = base_negative
+
+    @staticmethod
+    def update(node_id, outputs, metadata):
+        # Extract transformed conditionings from outputs
+        # outputs structure: [(base_positive, base_negative, show_help, )]
+        if outputs and isinstance(outputs, list) and len(outputs) > 0:
+            first_output = outputs[0]
+            if isinstance(first_output, tuple) and len(first_output) >= 2:
+                transformed_positive = first_output[0]
+                transformed_negative = first_output[1]
+                
+                # Save transformed conditioning objects in metadata
+                if node_id not in metadata[PROMPTS]:
+                    metadata[PROMPTS][node_id] = {"node_id": node_id}
+                    
+                metadata[PROMPTS][node_id]["positive_encoded"] = transformed_positive
+                metadata[PROMPTS][node_id]["negative_encoded"] = transformed_negative
+
+# Registry of node-specific extractors
+# Keys are node class names
+NODE_EXTRACTORS = {
+    # Sampling
+    "KSampler": SamplerExtractor,
+    "KSamplerAdvanced": KSamplerAdvancedExtractor,
+    "SamplerCustom": KSamplerAdvancedExtractor,
+    "SamplerCustomAdvanced": SamplerCustomAdvancedExtractor,
+    "ClownsharKSampler_Beta": SamplerExtractor,
+    "TSC_KSampler": TSCKSamplerExtractor,   # Efficient Nodes
+    "TSC_KSamplerAdvanced": TSCKSamplerAdvancedExtractor,  # Efficient Nodes
+    "KSamplerBasicPipe": KSamplerBasicPipeExtractor,    # comfyui-impact-pack
+    "KSamplerAdvancedBasicPipe": KSamplerAdvancedBasicPipeExtractor,    # comfyui-impact-pack
+    "KSampler_inspire_pipe": KSamplerBasicPipeExtractor,    # comfyui-inspire-pack
+    "KSamplerAdvanced_inspire_pipe": KSamplerAdvancedBasicPipeExtractor,  # comfyui-inspire-pack
+    "KSampler_inspire": SamplerExtractor,  # comfyui-inspire-pack
+    # Sampling Selectors
+    "KSamplerSelect": KSamplerSelectExtractor,  # Add KSamplerSelect
+    "BasicScheduler": BasicSchedulerExtractor,  # Add BasicScheduler
+    "AlignYourStepsScheduler": BasicSchedulerExtractor,  # Add AlignYourStepsScheduler
+    # Loaders
+    "CheckpointLoaderSimple": CheckpointLoaderExtractor,
+    "comfyLoader": CheckpointLoaderExtractor,  # easy comfyLoader
+    "CheckpointLoaderSimpleWithImages": CheckpointLoaderExtractor,  # CheckpointLoader|pysssss
+    "TSC_EfficientLoader": TSCCheckpointLoaderExtractor,  # Efficient Nodes
+    "NunchakuFluxDiTLoader": NunchakuFluxDiTLoaderExtractor,  # ComfyUI-Nunchaku
+    "NunchakuQwenImageDiTLoader": NunchakuQwenImageDiTLoaderExtractor,  # ComfyUI-Nunchaku
+    "LoaderGGUF": GGUFLoaderExtractor,  # calcuis gguf
+    "LoaderGGUFAdvanced": GGUFLoaderExtractor,  # calcuis gguf
+    "GGUFLoaderKJ": KJNodesModelLoaderExtractor,  # KJNodes
+    "DiffusionModelLoaderKJ": KJNodesModelLoaderExtractor,  # KJNodes
+    "CheckpointLoaderKJ": CheckpointLoaderExtractor,  # KJNodes
+    "UNETLoader": UNETLoaderExtractor,          # Updated to use dedicated extractor
+    "UnetLoaderGGUF": UNETLoaderExtractor,  # Updated to use dedicated extractor
+    "LoraLoader": LoraLoaderExtractor,
+    "LoraLoaderLM": LoraLoaderManagerExtractor,
+    # Conditioning
+    "CLIPTextEncode": CLIPTextEncodeExtractor,
+    "PromptLM": CLIPTextEncodeExtractor,
+    "CLIPTextEncodeFlux": CLIPTextEncodeFluxExtractor,  # Add CLIPTextEncodeFlux
+    "WAS_Text_to_Conditioning": CLIPTextEncodeExtractor,
+    "AdvancedCLIPTextEncode": CLIPTextEncodeExtractor,  # From https://github.com/BlenderNeko/ComfyUI_ADV_CLIP_emb
+    "smZ_CLIPTextEncode": CLIPTextEncodeExtractor,  # From https://github.com/shiimizu/ComfyUI_smZNodes
+    "CR_ApplyControlNetStack": CR_ApplyControlNetStackExtractor,  # Add CR_ApplyControlNetStack
+    "PCTextEncode": CLIPTextEncodeExtractor,  # From https://github.com/asagi4/comfyui-prompt-control
+    # Latent
+    "EmptyLatentImage": ImageSizeExtractor,
+    # Flux
+    "FluxGuidance": FluxGuidanceExtractor,      # Add FluxGuidance
+    "CFGGuider": CFGGuiderExtractor,            # Add CFGGuider
+    # Image
+    "VAEDecode": VAEDecodeExtractor,  # Added VAEDecode extractor
+    # Add other nodes as needed
+}

py/middleware/__init__.py

@@ -0,0 +1 @@
+"""Server middleware modules"""

py/middleware/cache_middleware.py

@@ -0,0 +1,53 @@
+"""Cache control middleware for ComfyUI server"""
+
+from aiohttp import web
+from typing import Callable, Awaitable
+
+# Time in seconds
+ONE_HOUR: int = 3600
+ONE_DAY: int = 86400
+IMG_EXTENSIONS = (
+    ".jpg",
+    ".jpeg",
+    ".png",
+    ".ppm",
+    ".bmp",
+    ".pgm",
+    ".tif",
+    ".tiff",
+    ".webp",
+    ".mp4"
+)
+
+
+@web.middleware
+async def cache_control(
+    request: web.Request, handler: Callable[[web.Request], Awaitable[web.Response]]
+) -> web.Response:
+    """Cache control middleware that sets appropriate cache headers based on file type and response status"""
+    response: web.Response = await handler(request)
+
+    if (
+        request.path.endswith(".js")
+        or request.path.endswith(".css")
+        or request.path.endswith("index.json")
+    ):
+        response.headers.setdefault("Cache-Control", "no-cache")
+        return response
+
+    # Early return for non-image files - no cache headers needed
+    if not request.path.lower().endswith(IMG_EXTENSIONS):
+        return response
+
+    # Handle image files
+    if response.status == 404:
+        response.headers.setdefault("Cache-Control", f"public, max-age={ONE_HOUR}")
+    elif response.status in (200, 201, 202, 203, 204, 205, 206, 301, 308):
+        # Success responses and permanent redirects - cache for 1 day
+        response.headers.setdefault("Cache-Control", f"public, max-age={ONE_DAY}")
+    elif response.status in (302, 303, 307):
+        # Temporary redirects - no cache
+        response.headers.setdefault("Cache-Control", "no-cache")
+    # Note: 304 Not Modified falls through - no cache headers set
+
+    return response

py/middleware/csp_middleware.py

@@ -0,0 +1,65 @@
+"""Middleware helpers for adjusting Content Security Policy headers."""
+
+from typing import Awaitable, Callable, Dict, List
+
+from aiohttp import web
+
+REMOTE_MEDIA_SOURCES = (
+    "https://image.civitai.com",
+    "https://img.genur.art",
+)
+
+
+@web.middleware
+async def relax_csp_for_remote_media(
+    request: web.Request, handler: Callable[[web.Request], Awaitable[web.StreamResponse]]
+) -> web.StreamResponse:
+    """Allow LoRA Manager media previews to load from trusted remote domains.
+
+    When ComfyUI is started with ``--disable-api-nodes`` it injects a restrictive
+    ``Content-Security-Policy`` header that blocks remote images and videos. The
+    LoRA Manager UI legitimately needs to fetch previews from Civitai and Genur,
+    so this middleware augments the existing CSP to whitelist those hosts while
+    preserving all other directives.
+    """
+
+    response: web.StreamResponse = await handler(request)
+    header_value = response.headers.get("Content-Security-Policy")
+
+    if not header_value:
+        return response
+
+    directive_order: List[str] = []
+    directives: Dict[str, List[str]] = {}
+
+    for raw_directive in header_value.split(";"):
+        directive = raw_directive.strip()
+        if not directive:
+            continue
+
+        parts = directive.split()
+        name, values = parts[0], parts[1:]
+        if name not in directive_order:
+            directive_order.append(name)
+        directives[name] = values
+
+    def merge_sources(name: str, sources: List[str], defaults: List[str] | None = None) -> None:
+        existing = directives.get(name, list(defaults or []))
+
+        for source in sources:
+            if source not in existing:
+                existing.append(source)
+
+        directives[name] = existing
+        if name not in directive_order:
+            directive_order.append(name)
+
+    merge_sources("img-src", list(REMOTE_MEDIA_SOURCES))
+    merge_sources("media-src", ["'self'", *REMOTE_MEDIA_SOURCES], defaults=["'self'"])
+
+    updated_header = "; ".join(
+        f"{name} {' '.join(directives[name])}".rstrip() for name in directive_order
+    )
+
+    response.headers["Content-Security-Policy"] = f"{updated_header};"
+    return response

py/nodes/debug_metadata.py

@@ -0,0 +1,61 @@
+import logging
+from ..metadata_collector.metadata_processor import MetadataProcessor
+
+logger = logging.getLogger(__name__)
+
+
+class DebugMetadataLM:
+    NAME = "Debug Metadata (LoraManager)"
+    CATEGORY = "Lora Manager/utils"
+    DESCRIPTION = "Debug node to verify metadata_processor functionality"
+    OUTPUT_NODE = True
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "images": ("IMAGE",),
+            },
+            "hidden": {
+                "id": "UNIQUE_ID",
+            },
+        }
+
+    RETURN_TYPES = ()
+    FUNCTION = "process_metadata"
+
+    def process_metadata(self, images, id):
+        """
+        Process metadata from the execution context and return it for UI display.
+
+        The metadata is returned via the 'ui' key in the return dict, which triggers
+        node.onExecuted on the frontend to update the JsonDisplayWidget.
+
+        Args:
+            images: Input images (required for execution flow)
+            id: Node's unique ID (hidden)
+
+        Returns:
+            Dict with 'result' (empty tuple) and 'ui' (metadata dict for widget display)
+        """
+        try:
+            # Get the current execution context's metadata
+            from ..metadata_collector import get_metadata
+
+            metadata = get_metadata()
+
+            # Use the MetadataProcessor to convert it to dict
+            metadata_dict = MetadataProcessor.to_dict(metadata, id)
+
+            return {
+                "result": (),
+                # ComfyUI expects ui values to be lists, wrap the dict in a list
+                "ui": {"metadata": [metadata_dict]},
+            }
+
+        except Exception as e:
+            logger.error(f"Error processing metadata: {e}")
+            return {
+                "result": (),
+                "ui": {"metadata": [{"error": str(e)}]},
+            }

py/nodes/lora_cycler.py

@@ -0,0 +1,134 @@
+"""
+Lora Cycler Node - Sequentially cycles through LoRAs from a pool.
+
+This node accepts optional pool_config input to filter available LoRAs, and outputs
+a LORA_STACK with one LoRA at a time. Returns UI updates with current/next LoRA info
+and tracks the cycle progress which persists across workflow save/load.
+"""
+
+import logging
+import os
+from ..utils.utils import get_lora_info
+
+logger = logging.getLogger(__name__)
+
+
+class LoraCyclerLM:
+    """Node that sequentially cycles through LoRAs from a pool"""
+
+    NAME = "Lora Cycler (LoraManager)"
+    CATEGORY = "Lora Manager/randomizer"
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "cycler_config": ("CYCLER_CONFIG", {}),
+            },
+            "optional": {
+                "pool_config": ("POOL_CONFIG", {}),
+            },
+        }
+
+    RETURN_TYPES = ("LORA_STACK",)
+    RETURN_NAMES = ("LORA_STACK",)
+
+    FUNCTION = "cycle"
+    OUTPUT_NODE = False
+
+    async def cycle(self, cycler_config, pool_config=None):
+        """
+        Cycle through LoRAs based on configuration and pool filters.
+
+        Args:
+            cycler_config: Dict with cycler settings (current_index, model_strength, clip_strength, sort_by)
+            pool_config: Optional config from LoRA Pool node for filtering
+
+        Returns:
+            Dictionary with 'result' (LORA_STACK tuple) and 'ui' (for widget display)
+        """
+        from ..services.service_registry import ServiceRegistry
+        from ..services.lora_service import LoraService
+
+        # Extract settings from cycler_config
+        current_index = cycler_config.get("current_index", 1)  # 1-based
+        model_strength = float(cycler_config.get("model_strength", 1.0))
+        clip_strength = float(cycler_config.get("clip_strength", 1.0))
+        sort_by = "filename"
+
+        # Dual-index mechanism for batch queue synchronization
+        execution_index = cycler_config.get("execution_index")  # Can be None
+        # next_index_from_config = cycler_config.get("next_index")  # Not used on backend
+
+        # Get scanner and service
+        scanner = await ServiceRegistry.get_lora_scanner()
+        lora_service = LoraService(scanner)
+
+        # Get filtered and sorted LoRA list
+        lora_list = await lora_service.get_cycler_list(
+            pool_config=pool_config, sort_by=sort_by
+        )
+
+        total_count = len(lora_list)
+
+        if total_count == 0:
+            logger.warning("[LoraCyclerLM] No LoRAs available in pool")
+            return {
+                "result": ([],),
+                "ui": {
+                    "current_index": [1],
+                    "next_index": [1],
+                    "total_count": [0],
+                    "current_lora_name": [""],
+                    "current_lora_filename": [""],
+                    "error": ["No LoRAs available in pool"],
+                },
+            }
+
+        # Determine which index to use for this execution
+        # If execution_index is provided (batch queue case), use it
+        # Otherwise use current_index (first execution or non-batch case)
+        if execution_index is not None:
+            actual_index = execution_index
+        else:
+            actual_index = current_index
+
+        # Clamp index to valid range (1-based)
+        clamped_index = max(1, min(actual_index, total_count))
+
+        # Get LoRA at current index (convert to 0-based for list access)
+        current_lora = lora_list[clamped_index - 1]
+
+        # Build LORA_STACK with single LoRA
+        lora_path, _ = get_lora_info(current_lora["file_name"])
+        if not lora_path:
+            logger.warning(
+                f"[LoraCyclerLM] Could not find path for LoRA: {current_lora['file_name']}"
+            )
+            lora_stack = []
+        else:
+            # Normalize path separators
+            lora_path = lora_path.replace("/", os.sep)
+            lora_stack = [(lora_path, model_strength, clip_strength)]
+
+        # Calculate next index (wrap to 1 if at end)
+        next_index = clamped_index + 1
+        if next_index > total_count:
+            next_index = 1
+
+        # Get next LoRA for UI display (what will be used next generation)
+        next_lora = lora_list[next_index - 1]
+        next_display_name = next_lora["file_name"]
+
+        return {
+            "result": (lora_stack,),
+            "ui": {
+                "current_index": [clamped_index],
+                "next_index": [next_index],
+                "total_count": [total_count],
+                "current_lora_name": [current_lora["file_name"]],
+                "current_lora_filename": [current_lora["file_name"]],
+                "next_lora_name": [next_display_name],
+                "next_lora_filename": [next_lora["file_name"]],
+            },
+        }

py/nodes/lora_loader.py

@@ -1,99 +1,266 @@
+import logging
+import re
 from nodes import LoraLoader
-from comfy.comfy_types import IO # type: ignore
-from ..services.lora_scanner import LoraScanner
-from ..config import config
-import asyncio
-import os
-from .utils import FlexibleOptionalInputType, any_type
+from ..utils.utils import get_lora_info
+from .utils import FlexibleOptionalInputType, any_type, extract_lora_name, get_loras_list, nunchaku_load_lora
 
-class LoraManagerLoader:
+logger = logging.getLogger(__name__)
+
+class LoraLoaderLM:
     NAME = "Lora Loader (LoraManager)"
     CATEGORY = "Lora Manager/loaders"
-
+    
     @classmethod
     def INPUT_TYPES(cls):
         return {
             "required": {
                 "model": ("MODEL",),
-                "clip": ("CLIP",),
-                "text": (IO.STRING, {
-                    "multiline": True, 
-                    "dynamicPrompts": True, 
+                # "clip": ("CLIP",),
+                "text": ("AUTOCOMPLETE_TEXT_LORAS", {
+                    "placeholder": "Search LoRAs to add...",
                     "tooltip": "Format: <lora:lora_name:strength> separated by spaces or punctuation",
-                    "placeholder": "LoRA syntax input: <lora:name:strength>"
                 }),
             },
             "optional": FlexibleOptionalInputType(any_type),
         }
 
-    RETURN_TYPES = ("MODEL", "CLIP", IO.STRING)
-    RETURN_NAMES = ("MODEL", "CLIP", "trigger_words")
+    RETURN_TYPES = ("MODEL", "CLIP", "STRING", "STRING")
+    RETURN_NAMES = ("MODEL", "CLIP", "trigger_words", "loaded_loras")
     FUNCTION = "load_loras"
-
-    async def get_lora_info(self, lora_name):
-        """Get the lora path and trigger words from cache"""
-        scanner = await LoraScanner.get_instance()
-        cache = await scanner.get_cached_data()
-        
-        for item in cache.raw_data:
-            if item.get('file_name') == lora_name:
-                file_path = item.get('file_path')
-                if file_path:
-                    for root in config.loras_roots:
-                        root = root.replace(os.sep, '/')
-                        if file_path.startswith(root):
-                            relative_path = os.path.relpath(file_path, root).replace(os.sep, '/')
-                            # Get trigger words from civitai metadata
-                            civitai = item.get('civitai', {})
-                            trigger_words = civitai.get('trainedWords', []) if civitai else []
-                            return relative_path, trigger_words
-        return lora_name, []  # Fallback if not found
-
-    def extract_lora_name(self, lora_path):
-        """Extract the lora name from a lora path (e.g., 'IL\\aorunIllstrious.safetensors' -> 'aorunIllstrious')"""
-        # Get the basename without extension
-        basename = os.path.basename(lora_path)
-        return os.path.splitext(basename)[0]
     
-    def load_loras(self, model, clip, text, **kwargs):
+    def load_loras(self, model, text, **kwargs):
         """Loads multiple LoRAs based on the kwargs input and lora_stack."""
         loaded_loras = []
         all_trigger_words = []
         
+        clip = kwargs.get('clip', None)
         lora_stack = kwargs.get('lora_stack', None)
+        
+        # Check if model is a Nunchaku Flux model - simplified approach
+        is_nunchaku_model = False
+        
+        try:
+            model_wrapper = model.model.diffusion_model
+            # Check if model is a Nunchaku Flux model using only class name
+            if model_wrapper.__class__.__name__ == "ComfyFluxWrapper":
+                is_nunchaku_model = True
+                logger.info("Detected Nunchaku Flux model")
+        except (AttributeError, TypeError):
+            # Not a model with the expected structure
+            pass
+        
         # First process lora_stack if available
         if lora_stack:
             for lora_path, model_strength, clip_strength in lora_stack:
-                # Apply the LoRA using the provided path and strengths
-                model, clip = LoraLoader().load_lora(model, clip, lora_path, model_strength, clip_strength)
+                # Apply the LoRA using the appropriate loader
+                if is_nunchaku_model:
+                    # Use our custom function for Flux models
+                    model = nunchaku_load_lora(model, lora_path, model_strength)
+                    # clip remains unchanged for Nunchaku models
+                else:
+                    # Use default loader for standard models
+                    model, clip = LoraLoader().load_lora(model, clip, lora_path, model_strength, clip_strength)
                 
                 # Extract lora name for trigger words lookup
-                lora_name = self.extract_lora_name(lora_path)
-                _, trigger_words = asyncio.run(self.get_lora_info(lora_name))
+                lora_name = extract_lora_name(lora_path)
+                _, trigger_words = get_lora_info(lora_name)
                 
                 all_trigger_words.extend(trigger_words)
-                loaded_loras.append(f"{lora_name}: {model_strength}")
+                # Add clip strength to output if different from model strength (except for Nunchaku models)
+                if not is_nunchaku_model and abs(model_strength - clip_strength) > 0.001:
+                    loaded_loras.append(f"{lora_name}: {model_strength},{clip_strength}")
+                else:
+                    loaded_loras.append(f"{lora_name}: {model_strength}")
         
-        # Then process loras from kwargs
-        if 'loras' in kwargs:
-            for lora in kwargs['loras']:
-                if not lora.get('active', False):
-                    continue
-                    
-                lora_name = lora['name']
-                strength = float(lora['strength'])
+        # Then process loras from kwargs with support for both old and new formats
+        loras_list = get_loras_list(kwargs)
+        for lora in loras_list:
+            if not lora.get('active', False):
+                continue
                 
-                # Get lora path and trigger words
-                lora_path, trigger_words = asyncio.run(self.get_lora_info(lora_name))
+            lora_name = lora['name']
+            model_strength = float(lora['strength'])
+            # Get clip strength - use model strength as default if not specified
+            clip_strength = float(lora.get('clipStrength', model_strength))
+            
+            # Get lora path and trigger words
+            lora_path, trigger_words = get_lora_info(lora_name)
+            
+            # Apply the LoRA using the appropriate loader
+            if is_nunchaku_model:
+                # For Nunchaku models, use our custom function
+                model = nunchaku_load_lora(model, lora_path, model_strength)
+                # clip remains unchanged
+            else:
+                # Use default loader for standard models
+                model, clip = LoraLoader().load_lora(model, clip, lora_path, model_strength, clip_strength)
+            
+            # Include clip strength in output if different from model strength and not a Nunchaku model
+            if not is_nunchaku_model and abs(model_strength - clip_strength) > 0.001:
+                loaded_loras.append(f"{lora_name}: {model_strength},{clip_strength}")
+            else:
+                loaded_loras.append(f"{lora_name}: {model_strength}")
+            
+            # Add trigger words to collection
+            all_trigger_words.extend(trigger_words)
+        
+        # use ',, ' to separate trigger words for group mode
+        trigger_words_text = ",, ".join(all_trigger_words) if all_trigger_words else ""
+        
+        # Format loaded_loras with support for both formats
+        formatted_loras = []
+        for item in loaded_loras:
+            parts = item.split(":")
+            lora_name = parts[0]
+            strength_parts = parts[1].strip().split(",")
+            
+            if len(strength_parts) > 1:
+                # Different model and clip strengths
+                model_str = strength_parts[0].strip()
+                clip_str = strength_parts[1].strip()
+                formatted_loras.append(f"<lora:{lora_name}:{model_str}:{clip_str}>")
+            else:
+                # Same strength for both
+                model_str = strength_parts[0].strip()
+                formatted_loras.append(f"<lora:{lora_name}:{model_str}>")
                 
-                # Apply the LoRA using the resolved path
-                model, clip = LoraLoader().load_lora(model, clip, lora_path, strength, strength)
-                loaded_loras.append(f"{lora_name}: {strength}")
+        formatted_loras_text = " ".join(formatted_loras)
+
+        return (model, clip, trigger_words_text, formatted_loras_text)
+
+class LoraTextLoaderLM:
+    NAME = "LoRA Text Loader (LoraManager)"
+    CATEGORY = "Lora Manager/loaders"
+    
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "model": ("MODEL",),
+                "lora_syntax": ("STRING", {
+                    "forceInput": True,
+                    "tooltip": "Format: <lora:lora_name:strength> separated by spaces or punctuation"
+                }),
+            },
+            "optional": {
+                "clip": ("CLIP",),
+                "lora_stack": ("LORA_STACK",),
+            }
+        }
+
+    RETURN_TYPES = ("MODEL", "CLIP", "STRING", "STRING")
+    RETURN_NAMES = ("MODEL", "CLIP", "trigger_words", "loaded_loras")
+    FUNCTION = "load_loras_from_text"
+    
+    def parse_lora_syntax(self, text):
+        """Parse LoRA syntax from text input."""
+        # Pattern to match <lora:name:strength> or <lora:name:model_strength:clip_strength>
+        pattern = r'<lora:([^:>]+):([^:>]+)(?::([^:>]+))?>'
+        matches = re.findall(pattern, text, re.IGNORECASE)
+        
+        loras = []
+        for match in matches:
+            lora_name = match[0]
+            model_strength = float(match[1])
+            clip_strength = float(match[2]) if match[2] else model_strength
+            
+            loras.append({
+                'name': lora_name,
+                'model_strength': model_strength,
+                'clip_strength': clip_strength
+            })
+        
+        return loras
+    
+    def load_loras_from_text(self, model, lora_syntax, clip=None, lora_stack=None):
+        """Load LoRAs based on text syntax input."""
+        loaded_loras = []
+        all_trigger_words = []
+        
+        # Check if model is a Nunchaku Flux model - simplified approach
+        is_nunchaku_model = False
+        
+        try:
+            model_wrapper = model.model.diffusion_model
+            # Check if model is a Nunchaku Flux model using only class name
+            if model_wrapper.__class__.__name__ == "ComfyFluxWrapper":
+                is_nunchaku_model = True
+                logger.info("Detected Nunchaku Flux model")
+        except (AttributeError, TypeError):
+            # Not a model with the expected structure
+            pass
+        
+        # First process lora_stack if available
+        if lora_stack:
+            for lora_path, model_strength, clip_strength in lora_stack:
+                # Apply the LoRA using the appropriate loader
+                if is_nunchaku_model:
+                    # Use our custom function for Flux models
+                    model = nunchaku_load_lora(model, lora_path, model_strength)
+                    # clip remains unchanged for Nunchaku models
+                else:
+                    # Use default loader for standard models
+                    model, clip = LoraLoader().load_lora(model, clip, lora_path, model_strength, clip_strength)
+                
+                # Extract lora name for trigger words lookup
+                lora_name = extract_lora_name(lora_path)
+                _, trigger_words = get_lora_info(lora_name)
                 
-                # Add trigger words to collection
                 all_trigger_words.extend(trigger_words)
+                # Add clip strength to output if different from model strength (except for Nunchaku models)
+                if not is_nunchaku_model and abs(model_strength - clip_strength) > 0.001:
+                    loaded_loras.append(f"{lora_name}: {model_strength},{clip_strength}")
+                else:
+                    loaded_loras.append(f"{lora_name}: {model_strength}")
+        
+        # Parse and process LoRAs from text syntax
+        parsed_loras = self.parse_lora_syntax(lora_syntax)
+        for lora in parsed_loras:
+            lora_name = lora['name']
+            model_strength = lora['model_strength']
+            clip_strength = lora['clip_strength']
+            
+            # Get lora path and trigger words
+            lora_path, trigger_words = get_lora_info(lora_name)
+            
+            # Apply the LoRA using the appropriate loader
+            if is_nunchaku_model:
+                # For Nunchaku models, use our custom function
+                model = nunchaku_load_lora(model, lora_path, model_strength)
+                # clip remains unchanged
+            else:
+                # Use default loader for standard models
+                model, clip = LoraLoader().load_lora(model, clip, lora_path, model_strength, clip_strength)
+            
+            # Include clip strength in output if different from model strength and not a Nunchaku model
+            if not is_nunchaku_model and abs(model_strength - clip_strength) > 0.001:
+                loaded_loras.append(f"{lora_name}: {model_strength},{clip_strength}")
+            else:
+                loaded_loras.append(f"{lora_name}: {model_strength}")
+            
+            # Add trigger words to collection
+            all_trigger_words.extend(trigger_words)
         
         # use ',, ' to separate trigger words for group mode
         trigger_words_text = ",, ".join(all_trigger_words) if all_trigger_words else ""
 
-        return (model, clip, trigger_words_text)
+        # Format loaded_loras with support for both formats
+        formatted_loras = []
+        for item in loaded_loras:
+            parts = item.split(":")
+            lora_name = parts[0].strip()
+            strength_parts = parts[1].strip().split(",")
+            
+            if len(strength_parts) > 1:
+                # Different model and clip strengths
+                model_str = strength_parts[0].strip()
+                clip_str = strength_parts[1].strip()
+                formatted_loras.append(f"<lora:{lora_name}:{model_str}:{clip_str}>")
+            else:
+                # Same strength for both
+                model_str = strength_parts[0].strip()
+                formatted_loras.append(f"<lora:{lora_name}:{model_str}>")
+                
+        formatted_loras_text = " ".join(formatted_loras)
+
+        return (model, clip, trigger_words_text, formatted_loras_text)

py/nodes/lora_pool.py

@@ -0,0 +1,87 @@
+"""
+LoRA Pool Node - Defines filter configuration for LoRA selection.
+
+This node provides a visual filter editor that generates a LORA_POOL_CONFIG
+object for use by downstream nodes (like LoRA Randomizer).
+"""
+
+import logging
+
+logger = logging.getLogger(__name__)
+
+
+class LoraPoolLM:
+    """
+    A node that defines LoRA filter criteria through a Vue-based widget.
+
+    Outputs a LORA_POOL_CONFIG that can be consumed by:
+    - Frontend: LoRA Randomizer widget reads connected pool's widget value
+    - Backend: LoRA Randomizer receives config during workflow execution
+    """
+
+    NAME = "Lora Pool (LoraManager)"
+    CATEGORY = "Lora Manager/randomizer"
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "pool_config": ("LORA_POOL_CONFIG", {}),
+            },
+            "hidden": {
+                # Hidden input to pass through unique node ID for frontend
+                "unique_id": "UNIQUE_ID",
+            },
+        }
+
+    RETURN_TYPES = ("POOL_CONFIG",)
+    RETURN_NAMES = ("POOL_CONFIG",)
+
+    FUNCTION = "process"
+    OUTPUT_NODE = False
+
+    def process(self, pool_config, unique_id=None):
+        """
+        Pass through the pool configuration filters.
+
+        The config is generated entirely by the frontend widget.
+        This function validates and returns only the filters field.
+
+        Args:
+            pool_config: Dict containing filter criteria from widget
+            unique_id: Node's unique ID (hidden)
+
+        Returns:
+            Tuple containing the filters dict from pool_config
+        """
+        # Validate required structure
+        if not isinstance(pool_config, dict):
+            logger.warning("Invalid pool_config type, using empty config")
+            pool_config = self._default_config()
+
+        # Ensure version field exists
+        if "version" not in pool_config:
+            pool_config["version"] = 1
+
+        # Extract filters field
+        filters = pool_config.get("filters", self._default_config()["filters"])
+
+        # Log for debugging
+        logger.debug(f"[LoraPoolLM] Processing filters: {filters}")
+
+        return (filters,)
+
+    @staticmethod
+    def _default_config():
+        """Return default empty configuration."""
+        return {
+            "version": 1,
+            "filters": {
+                "baseModels": [],
+                "tags": {"include": [], "exclude": []},
+                "folders": {"include": [], "exclude": []},
+                "favoritesOnly": False,
+                "license": {"noCreditRequired": False, "allowSelling": False},
+            },
+            "preview": {"matchCount": 0, "lastUpdated": 0},
+        }

py/nodes/lora_randomizer.py

@@ -0,0 +1,206 @@
+"""
+Lora Randomizer Node - Randomly selects LoRAs from a pool with configurable settings.
+
+This node accepts optional pool_config input to filter available LoRAs, and outputs
+a LORA_STACK with randomly selected LoRAs. Returns UI updates with new random LoRAs
+and tracks the last used combination for reuse.
+"""
+
+import logging
+import random
+import os
+from ..utils.utils import get_lora_info
+from .utils import extract_lora_name
+
+logger = logging.getLogger(__name__)
+
+
+class LoraRandomizerLM:
+    """Node that randomly selects LoRAs from a pool"""
+
+    NAME = "Lora Randomizer (LoraManager)"
+    CATEGORY = "Lora Manager/randomizer"
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "randomizer_config": ("RANDOMIZER_CONFIG", {}),
+                "loras": ("LORAS", {}),
+            },
+            "optional": {
+                "pool_config": ("POOL_CONFIG", {}),
+            },
+        }
+
+    RETURN_TYPES = ("LORA_STACK",)
+    RETURN_NAMES = ("LORA_STACK",)
+
+    FUNCTION = "randomize"
+    OUTPUT_NODE = False
+
+    def _preprocess_loras_input(self, loras):
+        """
+        Preprocess loras input to handle different widget formats.
+
+        Args:
+            loras: Input from widget, either:
+                  - List of LoRA dicts (expected format)
+                  - Dict with '__value__' key containing the list
+
+        Returns:
+            List of LoRA dicts
+        """
+        if isinstance(loras, dict) and "__value__" in loras:
+            return loras["__value__"]
+        return loras
+
+    async def randomize(self, randomizer_config, loras, pool_config=None):
+        """
+        Randomize LoRAs based on configuration and pool filters.
+
+        Args:
+            randomizer_config: Dict with randomizer settings (count, strength ranges, roll_mode)
+            loras: List of LoRA dicts from LORAS widget (includes locked state)
+            pool_config: Optional config from LoRA Pool node for filtering
+
+        Returns:
+            Dictionary with 'result' (LORA_STACK tuple) and 'ui' (for widget display)
+        """
+        from ..services.service_registry import ServiceRegistry
+
+        loras = self._preprocess_loras_input(loras)
+
+        roll_mode = randomizer_config.get("roll_mode", "always")
+        logger.debug(f"[LoraRandomizerLM] roll_mode: {roll_mode}")
+
+        # Dual seed mechanism for batch queue synchronization
+        # execution_seed: seed for generating execution_stack (= previous next_seed)
+        # next_seed: seed for generating ui_loras (= what will be displayed after execution)
+        execution_seed = randomizer_config.get("execution_seed", None)
+        next_seed = randomizer_config.get("next_seed", None)
+
+        if roll_mode == "fixed":
+            ui_loras = loras
+            execution_loras = loras
+        else:
+            scanner = await ServiceRegistry.get_lora_scanner()
+
+            # Generate execution_loras from execution_seed (if available)
+            if execution_seed is not None:
+                # Use execution_seed to regenerate the same loras that were shown to user
+                execution_loras = await self._generate_random_loras_for_ui(
+                    scanner, randomizer_config, loras, pool_config, seed=execution_seed
+                )
+            else:
+                # First execution: use loras input (what user sees in the widget)
+                execution_loras = loras
+
+            # Generate ui_loras from next_seed (for display after execution)
+            ui_loras = await self._generate_random_loras_for_ui(
+                scanner, randomizer_config, loras, pool_config, seed=next_seed
+            )
+
+        execution_stack = self._build_execution_stack_from_input(execution_loras)
+
+        return {
+            "result": (execution_stack,),
+            "ui": {"loras": ui_loras, "last_used": execution_loras},
+        }
+
+    def _build_execution_stack_from_input(self, loras):
+        """
+        Build LORA_STACK tuple from input loras list for execution.
+
+        Args:
+            loras: List of LoRA dicts with name, strength, clipStrength, active
+
+        Returns:
+            List of tuples (lora_path, model_strength, clip_strength)
+        """
+        lora_stack = []
+        for lora in loras:
+            if not lora.get("active", False):
+                continue
+
+            # Get file path
+            lora_path, trigger_words = get_lora_info(lora["name"])
+            if not lora_path:
+                logger.warning(
+                    f"[LoraRandomizerLM] Could not find path for LoRA: {lora['name']}"
+                )
+                continue
+
+            # Normalize path separators
+            lora_path = lora_path.replace("/", os.sep)
+
+            # Extract strengths (convert to float to prevent string subtraction errors)
+            model_strength = float(lora.get("strength", 1.0))
+            clip_strength = float(lora.get("clipStrength", model_strength))
+
+            lora_stack.append((lora_path, model_strength, clip_strength))
+
+        return lora_stack
+
+    async def _generate_random_loras_for_ui(
+        self, scanner, randomizer_config, input_loras, pool_config=None, seed=None
+    ):
+        """
+        Generate new random loras for UI display.
+
+        Args:
+            scanner: LoraScanner instance
+            randomizer_config: Dict with randomizer settings
+            input_loras: Current input loras (for extracting locked loras)
+            pool_config: Optional pool filters
+            seed: Optional seed for deterministic randomization
+
+        Returns:
+            List of LoRA dicts for UI display
+        """
+        from ..services.lora_service import LoraService
+
+        # Parse randomizer settings (convert numeric values to float to prevent type errors)
+        count_mode = randomizer_config.get("count_mode", "range")
+        count_fixed = int(randomizer_config.get("count_fixed", 5))
+        count_min = int(randomizer_config.get("count_min", 3))
+        count_max = int(randomizer_config.get("count_max", 7))
+        model_strength_min = float(randomizer_config.get("model_strength_min", 0.0))
+        model_strength_max = float(randomizer_config.get("model_strength_max", 1.0))
+        use_same_clip_strength = randomizer_config.get("use_same_clip_strength", True)
+        clip_strength_min = float(randomizer_config.get("clip_strength_min", 0.0))
+        clip_strength_max = float(randomizer_config.get("clip_strength_max", 1.0))
+        use_recommended_strength = randomizer_config.get(
+            "use_recommended_strength", False
+        )
+        recommended_strength_scale_min = float(
+            randomizer_config.get("recommended_strength_scale_min", 0.5)
+        )
+        recommended_strength_scale_max = float(
+            randomizer_config.get("recommended_strength_scale_max", 1.0)
+        )
+
+        # Extract locked LoRAs from input
+        locked_loras = [lora for lora in input_loras if lora.get("locked", False)]
+
+        # Use LoraService to generate random LoRAs
+        lora_service = LoraService(scanner)
+        result_loras = await lora_service.get_random_loras(
+            count=count_fixed,
+            model_strength_min=model_strength_min,
+            model_strength_max=model_strength_max,
+            use_same_clip_strength=use_same_clip_strength,
+            clip_strength_min=clip_strength_min,
+            clip_strength_max=clip_strength_max,
+            locked_loras=locked_loras,
+            pool_config=pool_config,
+            count_mode=count_mode,
+            count_min=count_min,
+            count_max=count_max,
+            use_recommended_strength=use_recommended_strength,
+            recommended_strength_scale_min=recommended_strength_scale_min,
+            recommended_strength_scale_max=recommended_strength_scale_max,
+            seed=seed,
+        )
+
+        return result_loras

py/nodes/lora_stacker.py

@@ -1,11 +1,12 @@
-from comfy.comfy_types import IO # type: ignore
-from ..services.lora_scanner import LoraScanner
-from ..config import config
-import asyncio
 import os
-from .utils import FlexibleOptionalInputType, any_type
+from ..utils.utils import get_lora_info
+from .utils import FlexibleOptionalInputType, any_type, extract_lora_name, get_loras_list
 
-class LoraStacker:
+import logging
+
+logger = logging.getLogger(__name__)
+
+class LoraStackerLM:
     NAME = "Lora Stacker (LoraManager)"
     CATEGORY = "Lora Manager/stackers"
 
@@ -13,79 +14,69 @@ class LoraStacker:
     def INPUT_TYPES(cls):
         return {
             "required": {
-                "text": (IO.STRING, {
-                    "multiline": True, 
-                    "dynamicPrompts": True, 
+                "text": ("AUTOCOMPLETE_TEXT_LORAS", {
+                    "placeholder": "Search LoRAs to add...",
                     "tooltip": "Format: <lora:lora_name:strength> separated by spaces or punctuation",
-                    "placeholder": "LoRA syntax input: <lora:name:strength>"
                 }),
             },
             "optional": FlexibleOptionalInputType(any_type),
         }
 
-    RETURN_TYPES = ("LORA_STACK", IO.STRING)
-    RETURN_NAMES = ("LORA_STACK", "trigger_words")
+    RETURN_TYPES = ("LORA_STACK", "STRING", "STRING")
+    RETURN_NAMES = ("LORA_STACK", "trigger_words", "active_loras")
     FUNCTION = "stack_loras"
-
-    async def get_lora_info(self, lora_name):
-        """Get the lora path and trigger words from cache"""
-        scanner = await LoraScanner.get_instance()
-        cache = await scanner.get_cached_data()
-        
-        for item in cache.raw_data:
-            if item.get('file_name') == lora_name:
-                file_path = item.get('file_path')
-                if file_path:
-                    for root in config.loras_roots:
-                        root = root.replace(os.sep, '/')
-                        if file_path.startswith(root):
-                            relative_path = os.path.relpath(file_path, root).replace(os.sep, '/')
-                            # Get trigger words from civitai metadata
-                            civitai = item.get('civitai', {})
-                            trigger_words = civitai.get('trainedWords', []) if civitai else []
-                            return relative_path, trigger_words
-        return lora_name, []  # Fallback if not found
-
-    def extract_lora_name(self, lora_path):
-        """Extract the lora name from a lora path (e.g., 'IL\\aorunIllstrious.safetensors' -> 'aorunIllstrious')"""
-        # Get the basename without extension
-        basename = os.path.basename(lora_path)
-        return os.path.splitext(basename)[0]
     
     def stack_loras(self, text, **kwargs):
         """Stacks multiple LoRAs based on the kwargs input without loading them."""
         stack = []
+        active_loras = []
         all_trigger_words = []
         
         # Process existing lora_stack if available
         lora_stack = kwargs.get('lora_stack', None)
-        if lora_stack:
+        if (lora_stack):
             stack.extend(lora_stack)
             # Get trigger words from existing stack entries
             for lora_path, _, _ in lora_stack:
-                lora_name = self.extract_lora_name(lora_path)
-                _, trigger_words = asyncio.run(self.get_lora_info(lora_name))
+                lora_name = extract_lora_name(lora_path)
+                _, trigger_words = get_lora_info(lora_name)
                 all_trigger_words.extend(trigger_words)
         
-        if 'loras' in kwargs:
-            for lora in kwargs['loras']:
-                if not lora.get('active', False):
-                    continue
-                    
-                lora_name = lora['name']
-                model_strength = float(lora['strength'])
-                clip_strength = model_strength  # Using same strength for both as in the original loader
+        # Process loras from kwargs with support for both old and new formats
+        loras_list = get_loras_list(kwargs)
+        for lora in loras_list:
+            if not lora.get('active', False):
+                continue
                 
-                # Get lora path and trigger words
-                lora_path, trigger_words = asyncio.run(self.get_lora_info(lora_name))
-                
-                # Add to stack without loading
-                stack.append((lora_path, model_strength, clip_strength))
-                
-                # Add trigger words to collection
-                all_trigger_words.extend(trigger_words)
+            lora_name = lora['name']
+            model_strength = float(lora['strength'])
+            # Get clip strength - use model strength as default if not specified
+            clip_strength = float(lora.get('clipStrength', model_strength))
+            
+            # Get lora path and trigger words
+            lora_path, trigger_words = get_lora_info(lora_name)
+            
+            # Add to stack without loading
+            # replace '/' with os.sep to avoid different OS path format
+            stack.append((lora_path.replace('/', os.sep), model_strength, clip_strength))
+            active_loras.append((lora_name, model_strength, clip_strength))
+            
+            # Add trigger words to collection
+            all_trigger_words.extend(trigger_words)
         
         # use ',, ' to separate trigger words for group mode
         trigger_words_text = ",, ".join(all_trigger_words) if all_trigger_words else ""
+        
+        # Format active_loras with support for both formats
+        formatted_loras = []
+        for name, model_strength, clip_strength in active_loras:
+            if abs(model_strength - clip_strength) > 0.001:
+                # Different model and clip strengths
+                formatted_loras.append(f"<lora:{name}:{str(model_strength).strip()}:{str(clip_strength).strip()}>")
+            else:
+                # Same strength for both
+                formatted_loras.append(f"<lora:{name}:{str(model_strength).strip()}>")
+                
+        active_loras_text = " ".join(formatted_loras)
 
-        return (stack, trigger_words_text)
+        return (stack, trigger_words_text, active_loras_text)

py/nodes/prompt.py

@@ -0,0 +1,58 @@
+from typing import Any, Optional
+
+class PromptLM:
+    """Encodes text (and optional trigger words) into CLIP conditioning."""
+
+    NAME = "Prompt (LoraManager)"
+    CATEGORY = "Lora Manager/conditioning"
+    DESCRIPTION = (
+        "Encodes a text prompt using a CLIP model into an embedding that can be used "
+        "to guide the diffusion model towards generating specific images."
+    )
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "text": (
+                    "AUTOCOMPLETE_TEXT_PROMPT,STRING",
+                    {
+                        "widgetType": "AUTOCOMPLETE_TEXT_PROMPT",
+                        "placeholder": "Enter prompt... /char, /artist for quick tag search",
+                        "tooltip": "The text to be encoded.",
+                    },
+                ),
+                "clip": (
+                    'CLIP',
+                    {"tooltip": "The CLIP model used for encoding the text."},
+                ),
+            },
+            "optional": {
+                "trigger_words": (
+                    'STRING',
+                    {
+                        "forceInput": True,
+                        "tooltip": (
+                            "Optional trigger words to prepend to the text before "
+                            "encoding."
+                        )
+                    },
+                )
+            },
+        }
+
+    RETURN_TYPES = ('CONDITIONING', 'STRING',)
+    RETURN_NAMES = ('CONDITIONING', 'PROMPT',)
+    OUTPUT_TOOLTIPS = (
+        "A conditioning containing the embedded text used to guide the diffusion model.",
+    )
+    FUNCTION = "encode"
+
+    def encode(self, text: str, clip: Any, trigger_words: Optional[str] = None):
+        prompt = text
+        if trigger_words:
+            prompt = ", ".join([trigger_words, text])
+
+        from nodes import CLIPTextEncode  # type: ignore
+        conditioning = CLIPTextEncode().encode(clip, prompt)[0]
+        return (conditioning, prompt,)

py/nodes/save_image.py

@@ -0,0 +1,454 @@
+import json
+import os
+import re
+import numpy as np
+import folder_paths # type: ignore
+from ..services.service_registry import ServiceRegistry
+from ..metadata_collector.metadata_processor import MetadataProcessor
+from ..metadata_collector import get_metadata
+from PIL import Image, PngImagePlugin
+import piexif
+import logging
+
+logger = logging.getLogger(__name__)
+
+class SaveImageLM:
+    NAME = "Save Image (LoraManager)"
+    CATEGORY = "Lora Manager/utils"
+    DESCRIPTION = "Save images with embedded generation metadata in compatible format"
+
+    def __init__(self):
+        self.output_dir = folder_paths.get_output_directory()
+        self.type = "output"
+        self.prefix_append = ""
+        self.compress_level = 4
+        self.counter = 0
+    
+    # Add pattern format regex for filename substitution
+    pattern_format = re.compile(r"(%[^%]+%)")
+    
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "images": ("IMAGE",),
+                "filename_prefix": ("STRING", {
+                    "default": "ComfyUI", 
+                    "tooltip": "Base filename for saved images. Supports format patterns like %seed%, %width%, %height%, %model%, etc."
+                }),
+                "file_format": (["png", "jpeg", "webp"], {
+                    "tooltip": "Image format to save as. PNG preserves quality, JPEG is smaller, WebP balances size and quality."
+                }),
+            },
+            "optional": {
+                "lossless_webp": ("BOOLEAN", {
+                    "default": False,
+                    "tooltip": "When enabled, saves WebP images with lossless compression. Results in larger files but no quality loss."
+                }),
+                "quality": ("INT", {
+                    "default": 100, 
+                    "min": 1, 
+                    "max": 100,
+                    "tooltip": "Compression quality for JPEG and lossy WebP formats (1-100). Higher values mean better quality but larger files."
+                }),
+                "embed_workflow": ("BOOLEAN", {
+                    "default": False,
+                    "tooltip": "Embeds the complete workflow data into the image metadata. Only works with PNG and WebP formats."
+                }),
+                "add_counter_to_filename": ("BOOLEAN", {
+                    "default": True,
+                    "tooltip": "Adds an incremental counter to filenames to prevent overwriting previous images."
+                }),
+            },
+            "hidden": {
+                "id": "UNIQUE_ID",
+                "prompt": "PROMPT",
+                "extra_pnginfo": "EXTRA_PNGINFO",
+            },
+        }
+
+    RETURN_TYPES = ("IMAGE",)
+    RETURN_NAMES = ("images",)
+    FUNCTION = "process_image"
+    OUTPUT_NODE = True
+
+    def get_lora_hash(self, lora_name):
+        """Get the lora hash from cache"""
+        scanner = ServiceRegistry.get_service_sync("lora_scanner")
+        
+        # Use the new direct filename lookup method
+        hash_value = scanner.get_hash_by_filename(lora_name)
+        if hash_value:
+            return hash_value
+            
+        return None
+
+    def get_checkpoint_hash(self, checkpoint_path):
+        """Get the checkpoint hash from cache"""
+        scanner = ServiceRegistry.get_service_sync("checkpoint_scanner")
+        
+        if not checkpoint_path:
+            return None
+            
+        # Extract basename without extension
+        checkpoint_name = os.path.basename(checkpoint_path)
+        checkpoint_name = os.path.splitext(checkpoint_name)[0]
+        
+        # Try direct filename lookup first
+        hash_value = scanner.get_hash_by_filename(checkpoint_name)
+        if hash_value:
+            return hash_value
+                
+        return None
+
+    def format_metadata(self, metadata_dict):
+        """Format metadata in the requested format similar to userComment example"""
+        if not metadata_dict:
+            return ""
+        
+        # Helper function to only add parameter if value is not None
+        def add_param_if_not_none(param_list, label, value):
+            if value is not None:
+                param_list.append(f"{label}: {value}")
+        
+        # Extract the prompt and negative prompt
+        prompt = metadata_dict.get('prompt', '')
+        negative_prompt = metadata_dict.get('negative_prompt', '')
+        
+        # Extract loras from the prompt if present
+        loras_text = metadata_dict.get('loras', '')
+        lora_hashes = {}
+        
+        # If loras are found, add them on a new line after the prompt
+        if loras_text:
+            prompt_with_loras = f"{prompt}\n{loras_text}"
+            
+            # Extract lora names from the format <lora:name:strength>
+            lora_matches = re.findall(r'<lora:([^:]+):([^>]+)>', loras_text)
+            
+            # Get hash for each lora
+            for lora_name, strength in lora_matches:
+                hash_value = self.get_lora_hash(lora_name)
+                if hash_value:
+                    lora_hashes[lora_name] = hash_value
+        else:
+            prompt_with_loras = prompt
+        
+        # Format the first part (prompt and loras)
+        metadata_parts = [prompt_with_loras]
+        
+        # Add negative prompt
+        if negative_prompt:
+            metadata_parts.append(f"Negative prompt: {negative_prompt}")
+        
+        # Format the second part (generation parameters)
+        params = []
+        
+        # Add standard parameters in the correct order
+        if 'steps' in metadata_dict:
+            add_param_if_not_none(params, "Steps", metadata_dict.get('steps'))
+        
+        # Combine sampler and scheduler information
+        sampler_name = None
+        scheduler_name = None
+        
+        if 'sampler' in metadata_dict:
+            sampler = metadata_dict.get('sampler')
+            # Convert ComfyUI sampler names to user-friendly names
+            sampler_mapping = {
+                'euler': 'Euler',
+                'euler_ancestral': 'Euler a',
+                'dpm_2': 'DPM2',
+                'dpm_2_ancestral': 'DPM2 a',
+                'heun': 'Heun',
+                'dpm_fast': 'DPM fast',
+                'dpm_adaptive': 'DPM adaptive',
+                'lms': 'LMS',
+                'dpmpp_2s_ancestral': 'DPM++ 2S a',
+                'dpmpp_sde': 'DPM++ SDE',
+                'dpmpp_sde_gpu': 'DPM++ SDE',
+                'dpmpp_2m': 'DPM++ 2M',
+                'dpmpp_2m_sde': 'DPM++ 2M SDE',
+                'dpmpp_2m_sde_gpu': 'DPM++ 2M SDE',
+                'ddim': 'DDIM'
+            }
+            sampler_name = sampler_mapping.get(sampler, sampler)
+        
+        if 'scheduler' in metadata_dict:
+            scheduler = metadata_dict.get('scheduler')
+            scheduler_mapping = {
+                'normal': 'Simple',
+                'karras': 'Karras',
+                'exponential': 'Exponential',
+                'sgm_uniform': 'SGM Uniform',
+                'sgm_quadratic': 'SGM Quadratic'
+            }
+            scheduler_name = scheduler_mapping.get(scheduler, scheduler)
+        
+        # Add combined sampler and scheduler information
+        if sampler_name:
+            if scheduler_name:
+                params.append(f"Sampler: {sampler_name} {scheduler_name}")
+            else:
+                params.append(f"Sampler: {sampler_name}")
+        
+        # CFG scale (Use guidance if available, otherwise fall back to cfg_scale or cfg)
+        if 'guidance' in metadata_dict:
+            add_param_if_not_none(params, "CFG scale", metadata_dict.get('guidance'))
+        elif 'cfg_scale' in metadata_dict:
+            add_param_if_not_none(params, "CFG scale", metadata_dict.get('cfg_scale'))
+        elif 'cfg' in metadata_dict:
+            add_param_if_not_none(params, "CFG scale", metadata_dict.get('cfg'))
+        
+        # Seed
+        if 'seed' in metadata_dict:
+            add_param_if_not_none(params, "Seed", metadata_dict.get('seed'))
+        
+        # Size
+        if 'size' in metadata_dict:
+            add_param_if_not_none(params, "Size", metadata_dict.get('size'))
+        
+        # Model info
+        if 'checkpoint' in metadata_dict:
+            # Ensure checkpoint is a string before processing
+            checkpoint = metadata_dict.get('checkpoint')
+            if checkpoint is not None:
+                # Get model hash
+                model_hash = self.get_checkpoint_hash(checkpoint)
+                
+                # Extract basename without path
+                checkpoint_name = os.path.basename(checkpoint)
+                # Remove extension if present
+                checkpoint_name = os.path.splitext(checkpoint_name)[0]
+                
+                # Add model hash if available
+                if model_hash:
+                    params.append(f"Model hash: {model_hash[:10]}, Model: {checkpoint_name}")
+                else:
+                    params.append(f"Model: {checkpoint_name}")
+        
+        # Add LoRA hashes if available
+        if lora_hashes:
+            lora_hash_parts = []
+            for lora_name, hash_value in lora_hashes.items():
+                lora_hash_parts.append(f"{lora_name}: {hash_value[:10]}")
+            
+            if lora_hash_parts:
+                params.append(f"Lora hashes: \"{', '.join(lora_hash_parts)}\"")
+        
+        # Combine all parameters with commas
+        metadata_parts.append(", ".join(params))
+        
+        # Join all parts with a new line
+        return "\n".join(metadata_parts)
+
+    # credit to nkchocoai
+    # Add format_filename method to handle pattern substitution
+    def format_filename(self, filename, metadata_dict):
+        """Format filename with metadata values"""
+        if not metadata_dict:
+            return filename
+            
+        result = re.findall(self.pattern_format, filename)
+        for segment in result:
+            parts = segment.replace("%", "").split(":")
+            key = parts[0]
+            
+            if key == "seed" and 'seed' in metadata_dict:
+                filename = filename.replace(segment, str(metadata_dict.get('seed', '')))
+            elif key == "width" and 'size' in metadata_dict:
+                size = metadata_dict.get('size', 'x')
+                w = size.split('x')[0] if isinstance(size, str) else size[0]
+                filename = filename.replace(segment, str(w))
+            elif key == "height" and 'size' in metadata_dict:
+                size = metadata_dict.get('size', 'x')
+                h = size.split('x')[1] if isinstance(size, str) else size[1]
+                filename = filename.replace(segment, str(h))
+            elif key == "pprompt" and 'prompt' in metadata_dict:
+                prompt = metadata_dict.get('prompt', '').replace("\n", " ")
+                if len(parts) >= 2:
+                    length = int(parts[1])
+                    prompt = prompt[:length]
+                filename = filename.replace(segment, prompt.strip())
+            elif key == "nprompt" and 'negative_prompt' in metadata_dict:
+                prompt = metadata_dict.get('negative_prompt', '').replace("\n", " ")
+                if len(parts) >= 2:
+                    length = int(parts[1])
+                    prompt = prompt[:length]
+                filename = filename.replace(segment, prompt.strip())
+            elif key == "model":
+                model_value = metadata_dict.get('checkpoint')
+                if isinstance(model_value, (bytes, os.PathLike)):
+                    model_value = str(model_value)
+
+                if not isinstance(model_value, str) or not model_value:
+                    model = "model_unavailable"
+                else:
+                    model = os.path.splitext(os.path.basename(model_value))[0]
+                if len(parts) >= 2:
+                    length = int(parts[1])
+                    model = model[:length]
+                filename = filename.replace(segment, model)
+            elif key == "date":
+                from datetime import datetime
+                now = datetime.now()
+                date_table = {
+                    "yyyy": f"{now.year:04d}",
+                    "yy": f"{now.year % 100:02d}",
+                    "MM": f"{now.month:02d}",
+                    "dd": f"{now.day:02d}",
+                    "hh": f"{now.hour:02d}",
+                    "mm": f"{now.minute:02d}",
+                    "ss": f"{now.second:02d}",
+                }
+                if len(parts) >= 2:
+                    date_format = parts[1]
+                    for k, v in date_table.items():
+                        date_format = date_format.replace(k, v)
+                    filename = filename.replace(segment, date_format)
+                else:
+                    date_format = "yyyyMMddhhmmss"
+                    for k, v in date_table.items():
+                        date_format = date_format.replace(k, v)
+                    filename = filename.replace(segment, date_format)
+                    
+        return filename
+
+    def save_images(self, images, filename_prefix, file_format, id, prompt=None, extra_pnginfo=None, 
+                   lossless_webp=True, quality=100, embed_workflow=False, add_counter_to_filename=True):
+        """Save images with metadata"""
+        results = []
+
+        # Get metadata using the metadata collector
+        raw_metadata = get_metadata()
+        metadata_dict = MetadataProcessor.to_dict(raw_metadata, id)
+            
+        metadata = self.format_metadata(metadata_dict)
+        
+        # Process filename_prefix with pattern substitution
+        filename_prefix = self.format_filename(filename_prefix, metadata_dict)
+        
+        # Get initial save path info once for the batch
+        full_output_folder, filename, counter, subfolder, processed_prefix = folder_paths.get_save_image_path(
+            filename_prefix, self.output_dir, images[0].shape[1], images[0].shape[0]
+        )
+        
+        # Create directory if it doesn't exist
+        if not os.path.exists(full_output_folder):
+            os.makedirs(full_output_folder, exist_ok=True)
+        
+        # Process each image with incrementing counter
+        for i, image in enumerate(images):
+            # Convert the tensor image to numpy array
+            img = 255. * image.cpu().numpy()
+            img = Image.fromarray(np.clip(img, 0, 255).astype(np.uint8))
+            
+            # Generate filename with counter if needed
+            base_filename = filename
+            if add_counter_to_filename:
+                # Use counter + i to ensure unique filenames for all images in batch
+                current_counter = counter + i
+                base_filename += f"_{current_counter:05}_"
+                
+            # Set file extension and prepare saving parameters
+            if file_format == "png":
+                file = base_filename + ".png"
+                file_extension = ".png"
+                # Remove "optimize": True to match built-in node behavior
+                save_kwargs = {"compress_level": self.compress_level}
+                pnginfo = PngImagePlugin.PngInfo()
+            elif file_format == "jpeg":
+                file = base_filename + ".jpg"
+                file_extension = ".jpg"
+                save_kwargs = {"quality": quality, "optimize": True}
+            elif file_format == "webp":
+                file = base_filename + ".webp" 
+                file_extension = ".webp"
+                # Add optimization param to control performance
+                save_kwargs = {"quality": quality, "lossless": lossless_webp, "method": 0}
+            
+            # Full save path
+            file_path = os.path.join(full_output_folder, file)
+            
+            # Save the image with metadata
+            try:
+                if file_format == "png":
+                    if metadata:
+                        pnginfo.add_text("parameters", metadata)
+                    if embed_workflow and extra_pnginfo is not None:
+                        workflow_json = json.dumps(extra_pnginfo["workflow"])
+                        pnginfo.add_text("workflow", workflow_json)
+                    save_kwargs["pnginfo"] = pnginfo
+                    img.save(file_path, format="PNG", **save_kwargs)
+                elif file_format == "jpeg":
+                    # For JPEG, use piexif
+                    if metadata:
+                        try:
+                            exif_dict = {'Exif': {piexif.ExifIFD.UserComment: b'UNICODE\0' + metadata.encode('utf-16be')}}
+                            exif_bytes = piexif.dump(exif_dict)
+                            save_kwargs["exif"] = exif_bytes
+                        except Exception as e:
+                            logger.error(f"Error adding EXIF data: {e}")
+                    img.save(file_path, format="JPEG", **save_kwargs)
+                elif file_format == "webp":
+                    try:
+                        # For WebP, use piexif for metadata
+                        exif_dict = {}
+
+                        if metadata:
+                            exif_dict['Exif'] = {piexif.ExifIFD.UserComment: b'UNICODE\0' + metadata.encode('utf-16be')}
+                        
+                        # Add workflow if needed
+                        if embed_workflow and extra_pnginfo is not None:
+                            workflow_json = json.dumps(extra_pnginfo["workflow"]) 
+                            exif_dict['0th'] = {piexif.ImageIFD.ImageDescription: "Workflow:" + workflow_json}
+                            
+                        exif_bytes = piexif.dump(exif_dict)
+                        save_kwargs["exif"] = exif_bytes
+                    except Exception as e:
+                        logger.error(f"Error adding EXIF data: {e}")
+                    
+                    img.save(file_path, format="WEBP", **save_kwargs)
+                
+                results.append({
+                    "filename": file,
+                    "subfolder": subfolder,
+                    "type": self.type
+                })
+                
+            except Exception as e:
+                logger.error(f"Error saving image: {e}")
+        
+        return results
+
+    def process_image(self, images, id, filename_prefix="ComfyUI", file_format="png", prompt=None, extra_pnginfo=None,
+                     lossless_webp=True, quality=100, embed_workflow=False, add_counter_to_filename=True):
+        """Process and save image with metadata"""
+        # Make sure the output directory exists
+        os.makedirs(self.output_dir, exist_ok=True)
+        
+        # If images is already a list or array of images, do nothing; otherwise, convert to list
+        if isinstance(images, (list, np.ndarray)):
+            pass
+        else:
+            # Ensure images is always a list of images
+            if len(images.shape) == 3:  # Single image (height, width, channels)
+                images = [images]
+            else:  # Multiple images (batch, height, width, channels)
+                images = [img for img in images]
+        
+        # Save all images
+        results = self.save_images(
+            images, 
+            filename_prefix, 
+            file_format, 
+            id,
+            prompt, 
+            extra_pnginfo,
+            lossless_webp,
+            quality,
+            embed_workflow,
+            add_counter_to_filename
+        )
+        
+        return (images,)

py/nodes/text.py

@@ -0,0 +1,33 @@
+class TextLM:
+    """A simple text node with autocomplete support."""
+
+    NAME = "Text (LoraManager)"
+    CATEGORY = "Lora Manager/utils"
+    DESCRIPTION = (
+        "A simple text input node with autocomplete support for tags and styles."
+    )
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "text": (
+                    "AUTOCOMPLETE_TEXT_PROMPT,STRING",
+                    {
+                        "widgetType": "AUTOCOMPLETE_TEXT_PROMPT",
+                        "placeholder": "Enter text... /char, /artist for quick tag search",
+                        "tooltip": "The text output.",
+                    },
+                ),
+            },
+        }
+
+    RETURN_TYPES = ("STRING",)
+    RETURN_NAMES = ("STRING",)
+    OUTPUT_TOOLTIPS = (
+        "The text output.",
+    )
+    FUNCTION = "process"
+
+    def process(self, text: str):
+        return (text,)

py/nodes/trigger_word_toggle.py

@@ -1,22 +1,45 @@
 import json
 import re
-from server import PromptServer # type: ignore
 from .utils import FlexibleOptionalInputType, any_type
+import logging
 
-class TriggerWordToggle:
+logger = logging.getLogger(__name__)
+
+
+class TriggerWordToggleLM:
     NAME = "TriggerWord Toggle (LoraManager)"
     CATEGORY = "Lora Manager/utils"
     DESCRIPTION = "Toggle trigger words on/off"
-    
+
     @classmethod
     def INPUT_TYPES(cls):
         return {
             "required": {
-                "group_mode": ("BOOLEAN", {"default": True}),
+                "group_mode": (
+                    "BOOLEAN",
+                    {
+                        "default": True,
+                        "tooltip": "When enabled, treats each group of trigger words as a single toggleable unit.",
+                    },
+                ),
+                "default_active": (
+                    "BOOLEAN",
+                    {
+                        "default": True,
+                        "tooltip": "Sets the default initial state (active or inactive) when trigger words are added.",
+                    },
+                ),
+                "allow_strength_adjustment": (
+                    "BOOLEAN",
+                    {
+                        "default": False,
+                        "tooltip": "Enable mouse wheel adjustment of each trigger word's strength.",
+                    },
+                ),
             },
             "optional": FlexibleOptionalInputType(any_type),
             "hidden": {
-                "id": "UNIQUE_ID",  # 会被 ComfyUI 自动替换为唯一ID
+                "id": "UNIQUE_ID",
             },
         }
 
@@ -24,55 +47,154 @@ class TriggerWordToggle:
     RETURN_NAMES = ("filtered_trigger_words",)
     FUNCTION = "process_trigger_words"
 
-    def process_trigger_words(self, id, group_mode, **kwargs):
-        print("process_trigger_words kwargs: ", kwargs)
-        trigger_words = kwargs.get("trigger_words", "")
-        # Send trigger words to frontend
-        PromptServer.instance.send_sync("trigger_word_update", {
-            "id": id,
-            "message": trigger_words
-        })
+    def _get_toggle_data(self, kwargs, key="toggle_trigger_words"):
+        """Helper to extract data from either old or new kwargs format"""
+        if key not in kwargs:
+            return None
+
+        data = kwargs[key]
+        # Handle new format: {'key': {'__value__': ...}}
+        if isinstance(data, dict) and "__value__" in data:
+            return data["__value__"]
+        # Handle old format: {'key': ...}
+        else:
+            return data
+
+    def _normalize_trigger_words(self, trigger_words):
+        """Normalize trigger words by splitting by both single and double commas, stripping whitespace, and filtering empty strings"""
+        if not trigger_words or not isinstance(trigger_words, str):
+            return set()
         
+        # Split by double commas first to preserve groups, then by single commas
+        groups = re.split(r",{2,}", trigger_words)
+        words = []
+        for group in groups:
+            # Split each group by single comma
+            group_words = [word.strip() for word in group.split(",")]
+            words.extend(group_words)
+        
+        # Filter out empty strings and return as set
+        return set(word for word in words if word)
+
+    def process_trigger_words(
+        self,
+        id,
+        group_mode,
+        default_active,
+        allow_strength_adjustment=False,
+        **kwargs,
+    ):
+        # Handle both old and new formats for trigger_words
+        trigger_words_data = self._get_toggle_data(kwargs, "orinalMessage")
+        trigger_words = (
+            trigger_words_data if isinstance(trigger_words_data, str) else ""
+        )
+
         filtered_triggers = trigger_words
-        
-        if 'toggle_trigger_words' in kwargs:
+
+        # Check if trigger_words is provided and different from orinalMessage
+        trigger_words_override = self._get_toggle_data(kwargs, "trigger_words")
+        if (
+            trigger_words_override
+            and isinstance(trigger_words_override, str)
+            and self._normalize_trigger_words(trigger_words_override) != self._normalize_trigger_words(trigger_words)
+        ):
+            filtered_triggers = trigger_words_override
+            return (filtered_triggers,)
+
+        # Get toggle data with support for both formats
+        trigger_data = self._get_toggle_data(kwargs, "toggle_trigger_words")
+        if trigger_data:
             try:
-                # Get trigger word toggle data
-                trigger_data = kwargs['toggle_trigger_words']
-                
                 # Convert to list if it's a JSON string
                 if isinstance(trigger_data, str):
                     trigger_data = json.loads(trigger_data)
-                
-                # Create dictionaries to track active state of words or groups
-                active_state = {item['text']: item.get('active', False) for item in trigger_data}
-                
-                if group_mode:
-                    # Split by two or more consecutive commas to get groups
-                    groups = re.split(r',{2,}', trigger_words)
-                    # Remove leading/trailing whitespace from each group
-                    groups = [group.strip() for group in groups]
-                    
-                    # Filter groups: keep those not in toggle_trigger_words or those that are active
-                    filtered_groups = [group for group in groups if group not in active_state or active_state[group]]
-                    
-                    if filtered_groups:
-                        filtered_triggers = ', '.join(filtered_groups)
+
+                if isinstance(trigger_data, list):
+                    if group_mode:
+                        if allow_strength_adjustment:
+                            parsed_items = [
+                                self._parse_trigger_item(
+                                    item, allow_strength_adjustment
+                                )
+                                for item in trigger_data
+                            ]
+                            filtered_groups = [
+                                self._format_word_output(
+                                    item["text"],
+                                    item["strength"],
+                                    allow_strength_adjustment,
+                                )
+                                for item in parsed_items
+                                if item["text"] and item["active"]
+                            ]
+                        else:
+                            filtered_groups = [
+                                (item.get("text") or "").strip()
+                                for item in trigger_data
+                                if (item.get("text") or "").strip()
+                                and item.get("active", False)
+                            ]
+                        filtered_triggers = (
+                            ", ".join(filtered_groups) if filtered_groups else ""
+                        )
                     else:
-                        filtered_triggers = ""
+                        parsed_items = [
+                            self._parse_trigger_item(item, allow_strength_adjustment)
+                            for item in trigger_data
+                        ]
+                        filtered_words = [
+                            self._format_word_output(
+                                item["text"],
+                                item["strength"],
+                                allow_strength_adjustment,
+                            )
+                            for item in parsed_items
+                            if item["text"] and item["active"]
+                        ]
+                        filtered_triggers = (
+                            ", ".join(filtered_words) if filtered_words else ""
+                        )
                 else:
-                    # Original behavior for individual words mode
-                    original_words = [word.strip() for word in trigger_words.split(',')]
-                    # Filter out empty strings
-                    original_words = [word for word in original_words if word]
-                    filtered_words = [word for word in original_words if word not in active_state or active_state[word]]
-                    
-                    if filtered_words:
-                        filtered_triggers = ', '.join(filtered_words)
+                    # Fallback to original message parsing if data is not in the expected list format
+                    if group_mode:
+                        groups = re.split(r",{2,}", trigger_words)
+                        groups = [group.strip() for group in groups if group.strip()]
+                        filtered_triggers = ", ".join(groups)
                     else:
-                        filtered_triggers = ""
-                    
+                        words = [
+                            word.strip()
+                            for word in trigger_words.split(",")
+                            if word.strip()
+                        ]
+                        filtered_triggers = ", ".join(words)
+
             except Exception as e:
-                print(f"Error processing trigger words: {e}")
-            
-        return (filtered_triggers,)
+                logger.error(f"Error processing trigger words: {e}")
+
+        return (filtered_triggers,)
+
+    def _parse_trigger_item(self, item, allow_strength_adjustment):
+        text = (item.get("text") or "").strip()
+        active = bool(item.get("active", False))
+        strength = item.get("strength")
+
+        strength_match = re.match(r"^\((.+):([\d.]+)\)$", text)
+        if strength_match:
+            text = strength_match.group(1).strip()
+            if strength is None:
+                try:
+                    strength = float(strength_match.group(2))
+                except ValueError:
+                    strength = None
+
+        return {
+            "text": text,
+            "active": active,
+            "strength": strength if allow_strength_adjustment else None,
+        }
+
+    def _format_word_output(self, base_word, strength, allow_strength_adjustment):
+        if allow_strength_adjustment and strength is not None:
+            return f"({base_word}:{strength:.2f})"
+        return base_word

py/nodes/utils.py

@@ -30,4 +30,118 @@ class FlexibleOptionalInputType(dict):
     return True
 
 
-any_type = AnyType("*")
+any_type = AnyType("*")
+
+# Common methods extracted from lora_loader.py and lora_stacker.py
+import os
+import logging
+import copy
+import sys
+import folder_paths
+
+logger = logging.getLogger(__name__)
+
+def extract_lora_name(lora_path):
+    """Extract the lora name from a lora path (e.g., 'IL\\aorunIllstrious.safetensors' -> 'aorunIllstrious')"""
+    # Get the basename without extension
+    basename = os.path.basename(lora_path)
+    return os.path.splitext(basename)[0]
+
+def get_loras_list(kwargs):
+    """Helper to extract loras list from either old or new kwargs format"""
+    if 'loras' not in kwargs:
+        return []
+        
+    loras_data = kwargs['loras']
+    # Handle new format: {'loras': {'__value__': [...]}}
+    if isinstance(loras_data, dict) and '__value__' in loras_data:
+        return loras_data['__value__']
+    # Handle old format: {'loras': [...]}
+    elif isinstance(loras_data, list):
+        return loras_data
+    # Unexpected format
+    else:
+        logger.warning(f"Unexpected loras format: {type(loras_data)}")
+        return []
+
+def load_state_dict_in_safetensors(path, device="cpu", filter_prefix=""):
+    """Simplified version of load_state_dict_in_safetensors that just loads from a local path"""  
+    import safetensors.torch
+    
+    state_dict = {}
+    with safetensors.torch.safe_open(path, framework="pt", device=device) as f:
+        for k in f.keys():
+            if filter_prefix and not k.startswith(filter_prefix):
+                continue
+            state_dict[k.removeprefix(filter_prefix)] = f.get_tensor(k)
+    return state_dict
+
+def to_diffusers(input_lora):
+    """Simplified version of to_diffusers for Flux LoRA conversion"""
+    import torch
+    from diffusers.utils.state_dict_utils import convert_unet_state_dict_to_peft
+    from diffusers.loaders import FluxLoraLoaderMixin
+    
+    if isinstance(input_lora, str):
+        tensors = load_state_dict_in_safetensors(input_lora, device="cpu")
+    else:
+        tensors = {k: v for k, v in input_lora.items()}
+
+    # Convert FP8 tensors to BF16
+    for k, v in tensors.items():
+        if v.dtype not in [torch.float64, torch.float32, torch.bfloat16, torch.float16]:
+            tensors[k] = v.to(torch.bfloat16)
+    
+    new_tensors = FluxLoraLoaderMixin.lora_state_dict(tensors)
+    new_tensors = convert_unet_state_dict_to_peft(new_tensors)
+
+    return new_tensors
+
+def nunchaku_load_lora(model, lora_name, lora_strength):
+    """Load a Flux LoRA for Nunchaku model"""   
+    # Get full path to the LoRA file. Allow both direct paths and registered LoRA names.
+    lora_path = lora_name if os.path.isfile(lora_name) else folder_paths.get_full_path("loras", lora_name)
+    if not lora_path or not os.path.isfile(lora_path):
+        logger.warning("Skipping LoRA '%s' because it could not be found", lora_name)
+        return model
+
+    model_wrapper = model.model.diffusion_model
+    
+    # Try to find copy_with_ctx in the same module as ComfyFluxWrapper
+    module_name = model_wrapper.__class__.__module__
+    module = sys.modules.get(module_name)
+    copy_with_ctx = getattr(module, "copy_with_ctx", None)
+
+    if copy_with_ctx is not None:
+        # New logic using copy_with_ctx from ComfyUI-nunchaku 1.1.0+
+        ret_model_wrapper, ret_model = copy_with_ctx(model_wrapper)
+        ret_model_wrapper.loras = [*model_wrapper.loras, (lora_path, lora_strength)]
+    else:
+        # Fallback to legacy logic
+        logger.warning("Please upgrade ComfyUI-nunchaku to 1.1.0 or above for better LoRA support. Falling back to legacy loading logic.")
+        transformer = model_wrapper.model
+        
+        # Save the transformer temporarily
+        model_wrapper.model = None
+        ret_model = copy.deepcopy(model)  # copy everything except the model
+        ret_model_wrapper = ret_model.model.diffusion_model
+        
+        # Restore the model and set it for the copy
+        model_wrapper.model = transformer
+        ret_model_wrapper.model = transformer
+        ret_model_wrapper.loras.append((lora_path, lora_strength))
+
+    # Convert the LoRA to diffusers format
+    sd = to_diffusers(lora_path)
+    
+    # Handle embedding adjustment if needed
+    if "transformer.x_embedder.lora_A.weight" in sd:
+        new_in_channels = sd["transformer.x_embedder.lora_A.weight"].shape[1]
+        assert new_in_channels % 4 == 0
+        new_in_channels = new_in_channels // 4
+        
+        old_in_channels = ret_model.model.model_config.unet_config["in_channels"]
+        if old_in_channels < new_in_channels:
+            ret_model.model.model_config.unet_config["in_channels"] = new_in_channels
+    
+    return ret_model

py/nodes/wanvideo_lora_select.py

@@ -0,0 +1,94 @@
+import folder_paths # type: ignore
+from ..utils.utils import get_lora_info
+from .utils import FlexibleOptionalInputType, any_type, get_loras_list
+import logging
+
+logger = logging.getLogger(__name__)
+
+class WanVideoLoraSelectLM:
+    NAME = "WanVideo Lora Select (LoraManager)"
+    CATEGORY = "Lora Manager/stackers"
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "low_mem_load": ("BOOLEAN", {"default": False, "tooltip": "Load LORA models with less VRAM usage, slower loading. This affects ALL LoRAs, not just the current ones. No effect if merge_loras is False"}),
+                "merge_loras": ("BOOLEAN", {"default": True, "tooltip": "Merge LoRAs into the model, otherwise they are loaded on the fly. Always disabled for GGUF and scaled fp8 models. This affects ALL LoRAs, not just the current one"}),
+                "text": ("AUTOCOMPLETE_TEXT_LORAS", {
+                    "placeholder": "Search LoRAs to add...",
+                    "tooltip": "Format: <lora:lora_name:strength> separated by spaces or punctuation",
+                }),
+            },
+            "optional": FlexibleOptionalInputType(any_type),
+        }
+
+    RETURN_TYPES = ("WANVIDLORA", "STRING", "STRING")
+    RETURN_NAMES = ("lora", "trigger_words", "active_loras")
+    FUNCTION = "process_loras"
+    
+    def process_loras(self, text, low_mem_load=False, merge_loras=True, **kwargs):
+        loras_list = []
+        all_trigger_words = []
+        active_loras = []
+        
+        # Process existing prev_lora if available
+        prev_lora = kwargs.get('prev_lora', None)
+        if prev_lora is not None:
+            loras_list.extend(prev_lora)
+
+        if not merge_loras:
+            low_mem_load = False  # Unmerged LoRAs don't need low_mem_load
+        
+        # Get blocks if available
+        blocks = kwargs.get('blocks', {})
+        selected_blocks = blocks.get("selected_blocks", {})
+        layer_filter = blocks.get("layer_filter", "")
+        
+        # Process loras from kwargs with support for both old and new formats
+        loras_from_widget = get_loras_list(kwargs)
+        for lora in loras_from_widget:
+            if not lora.get('active', False):
+                continue
+                
+            lora_name = lora['name']
+            model_strength = float(lora['strength'])
+            clip_strength = float(lora.get('clipStrength', model_strength))
+            
+            # Get lora path and trigger words
+            lora_path, trigger_words = get_lora_info(lora_name)
+            
+            # Create lora item for WanVideo format
+            lora_item = {
+                "path": folder_paths.get_full_path("loras", lora_path),
+                "strength": model_strength,
+                "name": lora_path.split(".")[0],
+                "blocks": selected_blocks,
+                "layer_filter": layer_filter,
+                "low_mem_load": low_mem_load,
+                "merge_loras": merge_loras,
+            }
+            
+            # Add to list and collect active loras
+            loras_list.append(lora_item)
+            active_loras.append((lora_name, model_strength, clip_strength))
+            
+            # Add trigger words to collection
+            all_trigger_words.extend(trigger_words)
+        
+        # Format trigger_words for output
+        trigger_words_text = ",, ".join(all_trigger_words) if all_trigger_words else ""
+        
+        # Format active_loras for output
+        formatted_loras = []
+        for name, model_strength, clip_strength in active_loras:
+            if abs(model_strength - clip_strength) > 0.001:
+                # Different model and clip strengths
+                formatted_loras.append(f"<lora:{name}:{str(model_strength).strip()}:{str(clip_strength).strip()}>")
+            else:
+                # Same strength for both
+                formatted_loras.append(f"<lora:{name}:{str(model_strength).strip()}>")
+                
+        active_loras_text = " ".join(formatted_loras)
+
+        return (loras_list, trigger_words_text, active_loras_text)

py/nodes/wanvideo_lora_select_from_text.py

@@ -0,0 +1,117 @@
+import folder_paths # type: ignore
+from ..utils.utils import get_lora_info
+from .utils import any_type 
+import logging
+
+# 初始化日志记录器
+logger = logging.getLogger(__name__)
+
+# 定义新节点的类
+class WanVideoLoraTextSelectLM:
+    # 节点在UI中显示的名称
+    NAME = "WanVideo Lora Select From Text (LoraManager)"
+    # 节点所属的分类
+    CATEGORY = "Lora Manager/stackers"
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "low_mem_load": ("BOOLEAN", {"default": False, "tooltip": "Load LORA models with less VRAM usage, slower loading. This affects ALL LoRAs, not just the current ones. No effect if merge_loras is False"}),
+                "merge_lora": ("BOOLEAN", {"default": True, "tooltip": "Merge LoRAs into the model, otherwise they are loaded on the fly. Always disabled for GGUF and scaled fp8 models. This affects ALL LoRAs, not just the current one"}),
+                "lora_syntax": ("STRING", {
+                    "multiline": True,
+                    "forceInput": True,
+                    "tooltip": "Connect a TEXT output for LoRA syntax: <lora:name:strength>"
+                }),
+            },
+
+            "optional": {
+                "prev_lora": ("WANVIDLORA",), 
+                "blocks": ("BLOCKS",)        
+            }
+        }
+
+    RETURN_TYPES = ("WANVIDLORA", "STRING", "STRING")
+    RETURN_NAMES = ("lora", "trigger_words", "active_loras")
+
+    FUNCTION = "process_loras_from_syntax"
+    
+    def process_loras_from_syntax(self, lora_syntax, low_mem_load=False, merge_lora=True, **kwargs):
+        text_to_process = lora_syntax
+
+        blocks = kwargs.get('blocks', {})
+        selected_blocks = blocks.get("selected_blocks", {})
+        layer_filter = blocks.get("layer_filter", "")
+
+        loras_list = []
+        all_trigger_words = []
+        active_loras = []
+        
+        prev_lora = kwargs.get('prev_lora', None)
+        if prev_lora is not None:
+            loras_list.extend(prev_lora)
+
+        if not merge_lora:
+            low_mem_load = False
+
+        parts = text_to_process.split('<lora:')
+        for part in parts[1:]:
+            end_index = part.find('>')
+            if end_index == -1:
+                continue
+
+            content = part[:end_index]
+            lora_parts = content.split(':')
+            
+            lora_name_raw = ""
+            model_strength = 1.0
+            clip_strength = 1.0
+
+            if len(lora_parts) == 2:
+                lora_name_raw = lora_parts[0].strip()
+                try:
+                    model_strength = float(lora_parts[1])
+                    clip_strength = model_strength
+                except (ValueError, IndexError):
+                    logger.warning(f"Invalid strength for LoRA '{lora_name_raw}'. Skipping.")
+                    continue
+            elif len(lora_parts) >= 3:
+                lora_name_raw = lora_parts[0].strip()
+                try:
+                    model_strength = float(lora_parts[1])
+                    clip_strength = float(lora_parts[2])
+                except (ValueError, IndexError):
+                    logger.warning(f"Invalid strengths for LoRA '{lora_name_raw}'. Skipping.")
+                    continue
+            else:
+                continue
+
+            lora_path, trigger_words = get_lora_info(lora_name_raw)
+
+            lora_item = {
+                "path": folder_paths.get_full_path("loras", lora_path),
+                "strength": model_strength,
+                "name": lora_path.split(".")[0],
+                "blocks": selected_blocks,
+                "layer_filter": layer_filter,
+                "low_mem_load": low_mem_load,
+                "merge_loras": merge_lora,
+            }
+
+            loras_list.append(lora_item)
+            active_loras.append((lora_name_raw, model_strength, clip_strength))
+            all_trigger_words.extend(trigger_words)
+
+        trigger_words_text = ",, ".join(all_trigger_words) if all_trigger_words else ""
+        
+        formatted_loras = []
+        for name, model_strength, clip_strength in active_loras:
+            if abs(model_strength - clip_strength) > 0.001:
+                formatted_loras.append(f"<lora:{name}:{str(model_strength).strip()}:{str(clip_strength).strip()}>")
+            else:
+                formatted_loras.append(f"<lora:{name}:{str(model_strength).strip()}>")
+                
+        active_loras_text = " ".join(formatted_loras)
+
+        return (loras_list, trigger_words_text, active_loras_text)

py/recipes/__init__.py

@@ -0,0 +1,24 @@
+"""Recipe metadata parser package for ComfyUI-Lora-Manager."""
+
+from .base import RecipeMetadataParser
+from .factory import RecipeParserFactory
+from .constants import GEN_PARAM_KEYS, VALID_LORA_TYPES
+from .parsers import (
+    RecipeFormatParser,
+    ComfyMetadataParser,
+    MetaFormatParser,
+    AutomaticMetadataParser,
+    CivitaiApiMetadataParser
+)
+
+__all__ = [
+    'RecipeMetadataParser',
+    'RecipeParserFactory',
+    'GEN_PARAM_KEYS',
+    'VALID_LORA_TYPES',
+    'RecipeFormatParser',
+    'ComfyMetadataParser',
+    'MetaFormatParser',
+    'AutomaticMetadataParser',
+    'CivitaiApiMetadataParser'
+]

py/recipes/base.py

@@ -0,0 +1,217 @@
+"""Base classes for recipe parsers."""
+
+import json
+import logging
+import os
+import re
+from typing import Dict, List, Any, Optional, Tuple
+from abc import ABC, abstractmethod
+from ..config import config
+from ..utils.constants import VALID_LORA_TYPES
+from ..utils.civitai_utils import rewrite_preview_url
+
+logger = logging.getLogger(__name__)
+
+class RecipeMetadataParser(ABC):
+    """Interface for parsing recipe metadata from image user comments"""
+
+    METADATA_MARKER = None
+
+    @abstractmethod
+    def is_metadata_matching(self, user_comment: str) -> bool:
+        """Check if the user comment matches the metadata format"""
+        pass
+    
+    @abstractmethod
+    async def parse_metadata(self, user_comment: str, recipe_scanner=None, civitai_client=None) -> Dict[str, Any]:
+        """
+        Parse metadata from user comment and return structured recipe data
+        
+        Args:
+            user_comment: The EXIF UserComment string from the image
+            recipe_scanner: Optional recipe scanner instance for local LoRA lookup
+            civitai_client: Optional Civitai client for fetching model information
+            
+        Returns:
+            Dict containing parsed recipe data with standardized format
+        """
+        pass
+    
+    @staticmethod
+    async def populate_lora_from_civitai(lora_entry: Dict[str, Any], civitai_info_tuple: Tuple[Dict[str, Any], Optional[str]], 
+                                         recipe_scanner=None, base_model_counts=None, hash_value=None) -> Optional[Dict[str, Any]]:
+        """
+        Populate a lora entry with information from Civitai API response
+        
+        Args:
+            lora_entry: The lora entry to populate
+            civitai_info_tuple: The response tuple from Civitai API (data, error_msg)
+            recipe_scanner: Optional recipe scanner for local file lookup
+            base_model_counts: Optional dict to track base model counts
+            hash_value: Optional hash value to use if not available in civitai_info
+            
+        Returns:
+            The populated lora_entry dict if type is valid, None otherwise
+        """
+        try:
+            # Unpack the tuple to get the actual data
+            civitai_info, error_msg = civitai_info_tuple if isinstance(civitai_info_tuple, tuple) else (civitai_info_tuple, None)
+            
+            if not civitai_info or error_msg == "Model not found":
+                # Model not found or deleted
+                lora_entry['isDeleted'] = True
+                lora_entry['thumbnailUrl'] = '/loras_static/images/no-preview.png'
+                return lora_entry
+                
+            # Get model type and validate
+            model_type = civitai_info.get('model', {}).get('type', '').lower()
+            lora_entry['type'] = model_type
+            if model_type not in VALID_LORA_TYPES:
+                logger.debug(f"Skipping non-LoRA model type: {model_type}")
+                return None
+
+            # Check if this is an early access lora
+            if civitai_info.get('earlyAccessEndsAt'):
+                # Convert earlyAccessEndsAt to a human-readable date
+                early_access_date = civitai_info.get('earlyAccessEndsAt', '')
+                lora_entry['isEarlyAccess'] = True
+                lora_entry['earlyAccessEndsAt'] = early_access_date
+                
+            # Update model name if available
+            if 'model' in civitai_info and 'name' in civitai_info['model']:
+                lora_entry['name'] = civitai_info['model']['name']
+
+            lora_entry['id'] = civitai_info.get('id')
+            lora_entry['modelId'] = civitai_info.get('modelId')
+            
+            # Update version if available
+            if 'name' in civitai_info:
+                lora_entry['version'] = civitai_info.get('name', '')
+            
+            # Get thumbnail URL from first image
+            if 'images' in civitai_info and civitai_info['images']:
+                image_url = civitai_info['images'][0].get('url')
+                if image_url:
+                    rewritten_image_url, _ = rewrite_preview_url(image_url, media_type='image')
+                    lora_entry['thumbnailUrl'] = rewritten_image_url or image_url
+            
+            # Get base model
+            current_base_model = civitai_info.get('baseModel', '')
+            lora_entry['baseModel'] = current_base_model
+            
+            # Update base model counts if tracking them
+            if base_model_counts is not None and current_base_model:
+                base_model_counts[current_base_model] = base_model_counts.get(current_base_model, 0) + 1
+            
+            # Get download URL
+            lora_entry['downloadUrl'] = civitai_info.get('downloadUrl', '')
+            
+            # Process file information if available
+            if 'files' in civitai_info:
+                # Find the primary model file (type="Model" and primary=true) in the files list
+                model_file = next((file for file in civitai_info.get('files', []) 
+                                    if file.get('type') == 'Model' and file.get('primary') == True), None)
+                
+                if model_file:
+                    # Get size
+                    lora_entry['size'] = model_file.get('sizeKB', 0) * 1024
+                    
+                    # Get SHA256 hash
+                    sha256 = model_file.get('hashes', {}).get('SHA256', hash_value)
+                    if sha256:
+                        lora_entry['hash'] = sha256.lower()
+                    
+                    # Check if exists locally
+                    if recipe_scanner and lora_entry['hash']:
+                        lora_scanner = recipe_scanner._lora_scanner
+                        exists_locally = lora_scanner.has_hash(lora_entry['hash'])
+                        if exists_locally:
+                            try:
+                                local_path = lora_scanner.get_path_by_hash(lora_entry['hash'])
+                                lora_entry['existsLocally'] = True
+                                lora_entry['localPath'] = local_path
+                                lora_entry['file_name'] = os.path.splitext(os.path.basename(local_path))[0]
+                                
+                                # Get thumbnail from local preview if available
+                                lora_cache = await lora_scanner.get_cached_data()
+                                lora_item = next((item for item in lora_cache.raw_data 
+                                                    if item['sha256'].lower() == lora_entry['hash'].lower()), None)
+                                if lora_item and 'preview_url' in lora_item:
+                                    lora_entry['thumbnailUrl'] = config.get_preview_static_url(lora_item['preview_url'])
+                            except Exception as e:
+                                logger.error(f"Error getting local lora path: {e}")
+                        else:
+                            # For missing LoRAs, get file_name from model_file.name
+                            file_name = model_file.get('name', '')
+                            lora_entry['file_name'] = os.path.splitext(file_name)[0] if file_name else ''
+                
+        except Exception as e:
+            logger.error(f"Error populating lora from Civitai info: {e}")
+            
+        return lora_entry
+    
+    @staticmethod
+    async def populate_checkpoint_from_civitai(checkpoint: Dict[str, Any], civitai_info: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        Populate checkpoint information from Civitai API response
+        
+        Args:
+            checkpoint: The checkpoint entry to populate
+            civitai_info: The response from Civitai API or a (data, error_msg) tuple
+            
+        Returns:
+            The populated checkpoint dict
+        """
+        try:
+            civitai_data, error_msg = (
+                (civitai_info, None)
+                if not isinstance(civitai_info, tuple)
+                else civitai_info
+            )
+
+            if not civitai_data or error_msg == "Model not found":
+                checkpoint['isDeleted'] = True
+                return checkpoint
+
+            if 'model' in civitai_data and 'name' in civitai_data['model']:
+                checkpoint['name'] = civitai_data['model']['name']
+
+            if 'name' in civitai_data:
+                checkpoint['version'] = civitai_data.get('name', '')
+
+            if 'images' in civitai_data and civitai_data['images']:
+                image_url = civitai_data['images'][0].get('url')
+                if image_url:
+                    rewritten_image_url, _ = rewrite_preview_url(image_url, media_type='image')
+                    checkpoint['thumbnailUrl'] = rewritten_image_url or image_url
+
+            checkpoint['baseModel'] = civitai_data.get('baseModel', '')
+            checkpoint['downloadUrl'] = civitai_data.get('downloadUrl', '')
+
+            checkpoint['modelId'] = civitai_data.get('modelId', checkpoint.get('modelId', 0))
+            checkpoint['id'] = civitai_data.get('id', 0)
+
+            if 'files' in civitai_data:
+                model_file = next(
+                    (
+                        file
+                        for file in civitai_data.get('files', [])
+                        if file.get('type') == 'Model'
+                    ),
+                    None,
+                )
+
+                if model_file:
+                    checkpoint['size'] = model_file.get('sizeKB', 0) * 1024
+
+                    sha256 = model_file.get('hashes', {}).get('SHA256')
+                    if sha256:
+                        checkpoint['hash'] = sha256.lower()
+
+                    file_name = model_file.get('name', '')
+                    if file_name:
+                        checkpoint['file_name'] = os.path.splitext(file_name)[0]
+        except Exception as e:
+            logger.error(f"Error populating checkpoint from Civitai info: {e}")
+            
+        return checkpoint

py/recipes/constants.py

@@ -0,0 +1,16 @@
+"""Constants used across recipe parsers."""
+
+# Import VALID_LORA_TYPES from utils.constants
+from ..utils.constants import VALID_LORA_TYPES
+
+# Constants for generation parameters
+GEN_PARAM_KEYS = [
+    'prompt',
+    'negative_prompt', 
+    'steps',
+    'sampler',
+    'cfg_scale',
+    'seed',
+    'size',
+    'clip_skip',
+]

py/recipes/enrichment.py

@@ -0,0 +1,216 @@
+import logging
+import json
+import re
+import os
+from typing import Any, Dict, Optional
+from .merger import GenParamsMerger
+from .base import RecipeMetadataParser
+from ..services.metadata_service import get_default_metadata_provider
+
+logger = logging.getLogger(__name__)
+
+class RecipeEnricher:
+    """Service to enrich recipe metadata from multiple sources (Civitai, Embedded, User)."""
+
+    @staticmethod
+    async def enrich_recipe(
+        recipe: Dict[str, Any],
+        civitai_client: Any,
+        request_params: Optional[Dict[str, Any]] = None
+    ) -> bool:
+        """
+        Enrich a recipe dictionary in-place with metadata from Civitai and embedded params.
+        
+        Args:
+            recipe: The recipe dictionary to enrich. Must have 'gen_params' initialized.
+            civitai_client: Authenticated Civitai client instance.
+            request_params: (Optional) Parameters from a user request (e.g. import).
+            
+        Returns:
+            bool: True if the recipe was modified, False otherwise.
+        """
+        updated = False
+        gen_params = recipe.get("gen_params", {})
+        
+        # 1. Fetch Civitai Info if available
+        civitai_meta = None
+        model_version_id = None
+        
+        source_url = recipe.get("source_url") or recipe.get("source_path", "")
+        
+        # Check if it's a Civitai image URL
+        image_id_match = re.search(r'civitai\.com/images/(\d+)', str(source_url))
+        if image_id_match:
+            image_id = image_id_match.group(1)
+            try:
+                image_info = await civitai_client.get_image_info(image_id)
+                if image_info:
+                    # Handle nested meta often found in Civitai API responses
+                    raw_meta = image_info.get("meta")
+                    if isinstance(raw_meta, dict):
+                        if "meta" in raw_meta and isinstance(raw_meta["meta"], dict):
+                            civitai_meta = raw_meta["meta"]
+                        else:
+                            civitai_meta = raw_meta
+                    
+                    model_version_id = image_info.get("modelVersionId")
+                    
+                    # If not at top level, check resources in meta
+                    if not model_version_id and civitai_meta:
+                        resources = civitai_meta.get("civitaiResources", [])
+                        for res in resources:
+                            if res.get("type") == "checkpoint":
+                                model_version_id = res.get("modelVersionId")
+                                break
+            except Exception as e:
+                logger.warning(f"Failed to fetch Civitai image info: {e}")
+
+        # 2. Merge Parameters
+        # Priority: request_params > civitai_meta > embedded (existing gen_params)
+        new_gen_params = GenParamsMerger.merge(
+            request_params=request_params,
+            civitai_meta=civitai_meta,
+            embedded_metadata=gen_params
+        )
+        
+        if new_gen_params != gen_params:
+            recipe["gen_params"] = new_gen_params
+            updated = True
+        
+        # 3. Checkpoint Enrichment
+        # If we have a checkpoint entry, or we can find one
+        # Use 'id' (from Civitai version) as a marker that it's been enriched
+        checkpoint_entry = recipe.get("checkpoint")
+        has_full_checkpoint = checkpoint_entry and checkpoint_entry.get("name") and checkpoint_entry.get("id")
+        
+        if not has_full_checkpoint:
+            # Helper to look up values in priority order
+            def start_lookup(keys):
+                for source in [request_params, civitai_meta, gen_params]:
+                    if source:
+                        if isinstance(keys, list):
+                            for k in keys:
+                                if k in source: return source[k]
+                        else:
+                            if keys in source: return source[keys]
+                return None
+
+            target_version_id = model_version_id or start_lookup("modelVersionId")
+            
+            # Also check existing checkpoint entry
+            if not target_version_id and checkpoint_entry:
+                target_version_id = checkpoint_entry.get("modelVersionId") or checkpoint_entry.get("id")
+            
+            # Check for version ID in resources (which might be a string in gen_params)
+            if not target_version_id:
+                # Look in all sources for "Civitai resources"
+                resources_val = start_lookup(["Civitai resources", "civitai_resources", "resources"])
+                if resources_val:
+                    target_version_id = RecipeEnricher._extract_version_id_from_resources({"Civitai resources": resources_val})
+            
+            target_hash = start_lookup(["Model hash", "checkpoint_hash", "hashes"])
+            if not target_hash and checkpoint_entry:
+                target_hash = checkpoint_entry.get("hash") or checkpoint_entry.get("model_hash")
+            
+            # Look for 'Model' which sometimes is the hash or name
+            model_val = start_lookup("Model")
+
+            # Look for Checkpoint name fallback
+            checkpoint_val = checkpoint_entry.get("name") if checkpoint_entry else None
+            if not checkpoint_val:
+                checkpoint_val = start_lookup(["Checkpoint", "checkpoint"])
+
+            checkpoint_updated = await RecipeEnricher._resolve_and_populate_checkpoint(
+                recipe, target_version_id, target_hash, model_val, checkpoint_val
+            )
+            if checkpoint_updated:
+                updated = True
+        else:
+            # Checkpoint exists, no need to sync to gen_params anymore.
+            pass
+        # base_model resolution moved to _resolve_and_populate_checkpoint to support strict formatting
+        return updated
+        
+    @staticmethod
+    def _extract_version_id_from_resources(gen_params: Dict[str, Any]) -> Optional[Any]:
+        """Try to find modelVersionId in Civitai resources parameter."""
+        civitai_resources_raw = gen_params.get("Civitai resources")
+        if not civitai_resources_raw:
+            return None
+            
+        resources_list = None
+        if isinstance(civitai_resources_raw, str):
+            try:
+                resources_list = json.loads(civitai_resources_raw)
+            except Exception:
+                pass
+        elif isinstance(civitai_resources_raw, list):
+            resources_list = civitai_resources_raw
+            
+        if isinstance(resources_list, list):
+            for res in resources_list:
+                if res.get("type") == "checkpoint":
+                    return res.get("modelVersionId")
+        return None
+
+    @staticmethod
+    async def _resolve_and_populate_checkpoint(
+        recipe: Dict[str, Any], 
+        target_version_id: Optional[Any], 
+        target_hash: Optional[str],
+        model_val: Optional[str],
+        checkpoint_val: Optional[str]
+    ) -> bool:
+        """Find checkpoint metadata and populate it in the recipe."""
+        metadata_provider = await get_default_metadata_provider()
+        civitai_info = None
+        
+        if target_version_id:
+            civitai_info = await metadata_provider.get_model_version_info(str(target_version_id))
+        elif target_hash:
+            civitai_info = await metadata_provider.get_model_by_hash(target_hash)
+        else:
+            # Look for 'Model' which sometimes is the hash or name
+            if model_val and len(model_val) == 10: # Likely a short hash
+                civitai_info = await metadata_provider.get_model_by_hash(model_val)
+                
+        if civitai_info and not (isinstance(civitai_info, tuple) and civitai_info[1] == "Model not found"):
+            # If we already have a partial checkpoint, use it as base
+            existing_cp = recipe.get("checkpoint")
+            if existing_cp is None:
+                existing_cp = {}
+            checkpoint_data = await RecipeMetadataParser.populate_checkpoint_from_civitai(existing_cp, civitai_info)
+            # 1. First, resolve base_model using full data before we format it away
+            current_base_model = recipe.get("base_model")
+            resolved_base_model = checkpoint_data.get("baseModel")
+            if resolved_base_model:
+                # Update if empty OR if it matches our generic prefix but is less specific
+                is_generic = not current_base_model or current_base_model.lower() in ["flux", "sdxl", "sd15"]
+                if is_generic and resolved_base_model != current_base_model:
+                    recipe["base_model"] = resolved_base_model
+            
+            # 2. Format according to requirements: type, modelId, modelVersionId, modelName, modelVersionName
+            formatted_checkpoint = {
+                "type": "checkpoint",
+                "modelId": checkpoint_data.get("modelId"),
+                "modelVersionId": checkpoint_data.get("id") or checkpoint_data.get("modelVersionId"),
+                "modelName": checkpoint_data.get("name"), # In base.py, 'name' is populated from civitai_data['model']['name']
+                "modelVersionName": checkpoint_data.get("version") # In base.py, 'version' is populated from civitai_data['name']
+            }
+            # Remove None values
+            recipe["checkpoint"] = {k: v for k, v in formatted_checkpoint.items() if v is not None}
+            
+            return True
+        else:
+            # Fallback to name extraction if we don't already have one
+            existing_cp = recipe.get("checkpoint")
+            if not existing_cp or not existing_cp.get("modelName"):
+                cp_name = checkpoint_val
+                if cp_name:
+                    recipe["checkpoint"] = {
+                        "type": "checkpoint",
+                        "modelName": cp_name
+                    }
+                    return True
+                
+        return False

py/recipes/factory.py

@@ -0,0 +1,64 @@
+"""Factory for creating recipe metadata parsers."""
+
+import logging
+from .parsers import (
+    RecipeFormatParser,
+    ComfyMetadataParser,
+    MetaFormatParser,
+    AutomaticMetadataParser,
+    CivitaiApiMetadataParser
+)
+from .base import RecipeMetadataParser
+
+logger = logging.getLogger(__name__)
+
+class RecipeParserFactory:
+    """Factory for creating recipe metadata parsers"""
+    
+    @staticmethod
+    def create_parser(metadata) -> RecipeMetadataParser:
+        """
+        Create appropriate parser based on the metadata content
+        
+        Args:
+            metadata: The metadata from the image (dict or str)
+            
+        Returns:
+            Appropriate RecipeMetadataParser implementation
+        """
+        # First, try CivitaiApiMetadataParser for dict input
+        if isinstance(metadata, dict):
+            try:
+                if CivitaiApiMetadataParser().is_metadata_matching(metadata):
+                    return CivitaiApiMetadataParser()
+            except Exception as e:
+                logger.debug(f"CivitaiApiMetadataParser check failed: {e}")
+                pass
+            
+            # Convert dict to string for other parsers that expect string input
+            try:
+                import json
+                metadata_str = json.dumps(metadata)
+            except Exception as e:
+                logger.debug(f"Failed to convert dict to JSON string: {e}")
+                return None
+        else:
+            metadata_str = metadata
+        
+        # Try ComfyMetadataParser which requires valid JSON
+        try:
+            if ComfyMetadataParser().is_metadata_matching(metadata_str):
+                return ComfyMetadataParser()
+        except Exception:
+            # If JSON parsing fails, move on to other parsers
+            pass
+        
+        # Check other parsers that expect string input
+        if RecipeFormatParser().is_metadata_matching(metadata_str):
+            return RecipeFormatParser()
+        elif AutomaticMetadataParser().is_metadata_matching(metadata_str):
+            return AutomaticMetadataParser()
+        elif MetaFormatParser().is_metadata_matching(metadata_str):
+            return MetaFormatParser()
+        else:
+            return None

py/recipes/merger.py

@@ -0,0 +1,98 @@
+from typing import Any, Dict, Optional
+import logging
+
+logger = logging.getLogger(__name__)
+
+class GenParamsMerger:
+    """Utility to merge generation parameters from multiple sources with priority."""
+
+    BLACKLISTED_KEYS = {
+        "id", "url", "userId", "username", "createdAt", "updatedAt", "hash", "meta",
+        "draft", "extra", "width", "height", "process", "quantity", "workflow",
+        "baseModel", "resources", "disablePoi", "aspectRatio", "Created Date",
+        "experimental", "civitaiResources", "civitai_resources", "Civitai resources",
+        "modelVersionId", "modelId", "hashes", "Model", "Model hash", "checkpoint_hash",
+        "checkpoint", "checksum", "model_checksum"
+    }
+    
+    NORMALIZATION_MAPPING = {
+        # Civitai specific
+        "cfgScale": "cfg_scale",
+        "clipSkip": "clip_skip",
+        "negativePrompt": "negative_prompt",
+        # Case variations
+        "Sampler": "sampler",
+        "Steps": "steps",
+        "Seed": "seed",
+        "Size": "size",
+        "Prompt": "prompt",
+        "Negative prompt": "negative_prompt",
+        "Cfg scale": "cfg_scale",
+        "Clip skip": "clip_skip",
+        "Denoising strength": "denoising_strength",
+    }
+
+    @staticmethod
+    def merge(
+        request_params: Optional[Dict[str, Any]] = None,
+        civitai_meta: Optional[Dict[str, Any]] = None,
+        embedded_metadata: Optional[Dict[str, Any]] = None
+    ) -> Dict[str, Any]:
+        """
+        Merge generation parameters from three sources.
+        
+        Priority: request_params > civitai_meta > embedded_metadata
+        
+        Args:
+            request_params: Params provided directly in the import request
+            civitai_meta: Params from Civitai Image API 'meta' field
+            embedded_metadata: Params extracted from image EXIF/embedded metadata
+            
+        Returns:
+            Merged parameters dictionary
+        """
+        result = {}
+
+        # 1. Start with embedded metadata (lowest priority)
+        if embedded_metadata:
+            # If it's a full recipe metadata, we use its gen_params
+            if "gen_params" in embedded_metadata and isinstance(embedded_metadata["gen_params"], dict):
+                GenParamsMerger._update_normalized(result, embedded_metadata["gen_params"])
+            else:
+                # Otherwise assume the dict itself contains gen_params
+                GenParamsMerger._update_normalized(result, embedded_metadata)
+
+        # 2. Layer Civitai meta (medium priority)
+        if civitai_meta:
+            GenParamsMerger._update_normalized(result, civitai_meta)
+
+        # 3. Layer request params (highest priority)
+        if request_params:
+            GenParamsMerger._update_normalized(result, request_params)
+
+        # Filter out blacklisted keys and also the original camelCase keys if they were normalized
+        final_result = {}
+        for k, v in result.items():
+            if k in GenParamsMerger.BLACKLISTED_KEYS:
+                continue
+            if k in GenParamsMerger.NORMALIZATION_MAPPING:
+                continue
+            final_result[k] = v
+            
+        return final_result
+
+    @staticmethod
+    def _update_normalized(target: Dict[str, Any], source: Dict[str, Any]) -> None:
+        """Update target dict with normalized keys from source."""
+        for k, v in source.items():
+            normalized_key = GenParamsMerger.NORMALIZATION_MAPPING.get(k, k)
+            target[normalized_key] = v
+            # Also keep the original key for now if it's not the same, 
+            # so we can filter at the end or avoid losing it if it wasn't supposed to be renamed?
+            # Actually, if we rename it, we should probably NOT keep both in 'target' 
+            # because we want to filter them out at the end anyway.
+            if normalized_key != k:
+                # If we are overwriting an existing snake_case key with a camelCase one's value,
+                # that's fine because of the priority order of calls to _update_normalized.
+                pass
+            target[k] = v

py/recipes/parsers/__init__.py

@@ -0,0 +1,15 @@
+"""Recipe parsers package."""
+
+from .recipe_format import RecipeFormatParser
+from .comfy import ComfyMetadataParser
+from .meta_format import MetaFormatParser
+from .automatic import AutomaticMetadataParser
+from .civitai_image import CivitaiApiMetadataParser
+
+__all__ = [
+    'RecipeFormatParser',
+    'ComfyMetadataParser',
+    'MetaFormatParser',
+    'AutomaticMetadataParser',
+    'CivitaiApiMetadataParser',
+]

py/recipes/parsers/automatic.py

@@ -0,0 +1,441 @@
+"""Parser for Automatic1111 metadata format."""
+
+import re
+import os
+import json
+import logging
+from typing import Dict, Any
+from ..base import RecipeMetadataParser
+from ..constants import GEN_PARAM_KEYS
+from ...services.metadata_service import get_default_metadata_provider
+
+logger = logging.getLogger(__name__)
+
+class AutomaticMetadataParser(RecipeMetadataParser):
+    """Parser for Automatic1111 metadata format"""
+    
+    METADATA_MARKER = r"Steps: \d+"
+    
+    # Regular expressions for extracting specific metadata
+    HASHES_REGEX = r', Hashes:\s*({[^}]+})'
+    LORA_HASHES_REGEX = r', Lora hashes:\s*"([^"]+)"'
+    CIVITAI_RESOURCES_REGEX = r', Civitai resources:\s*(\[\{.*?\}\])'
+    CIVITAI_METADATA_REGEX = r', Civitai metadata:\s*(\{.*?\})'
+    EXTRANETS_REGEX = r'<(lora|hypernet):([^:]+):(-?[0-9.]+)>'
+    MODEL_HASH_PATTERN = r'Model hash: ([a-zA-Z0-9]+)'
+    MODEL_NAME_PATTERN = r'Model: ([^,]+)'
+    VAE_HASH_PATTERN = r'VAE hash: ([a-zA-Z0-9]+)'
+    
+    def is_metadata_matching(self, user_comment: str) -> bool:
+        """Check if the user comment matches the Automatic1111 format"""
+        return re.search(self.METADATA_MARKER, user_comment) is not None
+    
+    async def parse_metadata(self, user_comment: str, recipe_scanner=None, civitai_client=None) -> Dict[str, Any]:
+        """Parse metadata from Automatic1111 format"""
+        try:
+            # Get metadata provider instead of using civitai_client directly
+            metadata_provider = await get_default_metadata_provider()
+            
+            # Split on Negative prompt if it exists
+            if "Negative prompt:" in user_comment:
+                parts = user_comment.split('Negative prompt:', 1)
+                prompt = parts[0].strip()
+                negative_and_params = parts[1] if len(parts) > 1 else ""
+            else:
+                # No negative prompt section
+                param_start = re.search(self.METADATA_MARKER, user_comment)
+                if param_start:
+                    prompt = user_comment[:param_start.start()].strip()
+                    negative_and_params = user_comment[param_start.start():]
+                else:
+                    prompt = user_comment.strip()
+                    negative_and_params = ""
+            
+            # Initialize metadata
+            metadata = {
+                "prompt": prompt,
+                "loras": []
+            }
+            
+            # Extract negative prompt and parameters
+            if negative_and_params:
+                # If we split on "Negative prompt:", check for params section
+                if "Negative prompt:" in user_comment:
+                    param_start = re.search(r'Steps: ', negative_and_params)
+                    if param_start:
+                        neg_prompt = negative_and_params[:param_start.start()].strip()
+                        metadata["negative_prompt"] = neg_prompt
+                        params_section = negative_and_params[param_start.start():]
+                    else:
+                        metadata["negative_prompt"] = negative_and_params.strip()
+                        params_section = ""
+                else:
+                    # No negative prompt, entire section is params
+                    params_section = negative_and_params
+                
+                # Extract generation parameters
+                if params_section:
+                    # Extract Civitai resources
+                    civitai_resources_match = re.search(self.CIVITAI_RESOURCES_REGEX, params_section)
+                    if civitai_resources_match:
+                        try:
+                            civitai_resources = json.loads(civitai_resources_match.group(1))
+                            metadata["civitai_resources"] = civitai_resources
+                            params_section = params_section.replace(civitai_resources_match.group(0), '')
+                        except json.JSONDecodeError:
+                            logger.error("Error parsing Civitai resources JSON")
+                    
+                    # Extract Hashes
+                    hashes_match = re.search(self.HASHES_REGEX, params_section)
+                    if hashes_match:
+                        try:
+                            hashes = json.loads(hashes_match.group(1))
+                            # Process hash keys
+                            processed_hashes = {}
+                            for key, value in hashes.items():
+                                # Convert Model: or LORA: prefix to lowercase if present
+                                if ':' in key:
+                                    prefix, name = key.split(':', 1)
+                                    prefix = prefix.lower()
+                                else:
+                                    prefix = ''
+                                    name = key
+
+                                # Clean up the name part
+                                if '/' in name:
+                                    name = name.split('/')[-1]  # Get last part after /
+                                if '.safetensors' in name:
+                                    name = name.split('.safetensors')[0]  # Remove .safetensors
+                                
+                                # Reconstruct the key
+                                new_key = f"{prefix}:{name}" if prefix else name
+                                processed_hashes[new_key] = value
+
+                            metadata["hashes"] = processed_hashes
+                            # Remove hashes from params section to not interfere with other parsing
+                            params_section = params_section.replace(hashes_match.group(0), '')
+                        except json.JSONDecodeError:
+                            logger.error("Error parsing hashes JSON")
+                    
+                    # Pick up model hash from parsed hashes if available
+                    if "hashes" in metadata and not metadata.get("model_hash"):
+                        model_hash_from_hashes = metadata["hashes"].get("model")
+                        if model_hash_from_hashes:
+                            metadata["model_hash"] = model_hash_from_hashes
+                    
+                    # Extract Lora hashes in alternative format
+                    lora_hashes_match = re.search(self.LORA_HASHES_REGEX, params_section)
+                    if not hashes_match and lora_hashes_match:
+                        try:
+                            lora_hashes_str = lora_hashes_match.group(1)
+                            lora_hash_entries = lora_hashes_str.split(', ')
+                            
+                            # Initialize hashes dict if it doesn't exist
+                            if "hashes" not in metadata:
+                                metadata["hashes"] = {}
+                                
+                            # Parse each lora hash entry (format: "name: hash")
+                            for entry in lora_hash_entries:
+                                if ': ' in entry:
+                                    lora_name, lora_hash = entry.split(': ', 1)
+                                    # Add as lora type in the same format as regular hashes
+                                    metadata["hashes"][f"lora:{lora_name}"] = lora_hash.strip()
+                            
+                            # Remove lora hashes from params section
+                            params_section = params_section.replace(lora_hashes_match.group(0), '')
+                        except Exception as e:
+                            logger.error(f"Error parsing Lora hashes: {e}")
+
+                    # Extract checkpoint model hash/name when provided outside Civitai resources
+                    model_hash_match = re.search(self.MODEL_HASH_PATTERN, params_section)
+                    if model_hash_match:
+                        metadata["model_hash"] = model_hash_match.group(1).strip()
+                        params_section = params_section.replace(model_hash_match.group(0), '')
+
+                    model_name_match = re.search(self.MODEL_NAME_PATTERN, params_section)
+                    if model_name_match:
+                        metadata["model_name"] = model_name_match.group(1).strip()
+                        params_section = params_section.replace(model_name_match.group(0), '')
+                    
+                    # Extract basic parameters
+                    param_pattern = r'([A-Za-z\s]+): ([^,]+)'
+                    params = re.findall(param_pattern, params_section)
+                    gen_params = {}
+                    
+                    for key, value in params:
+                        clean_key = key.strip().lower().replace(' ', '_')
+                        
+                        # Skip if not in recognized gen param keys
+                        if clean_key not in GEN_PARAM_KEYS:
+                            continue
+                            
+                        # Convert numeric values
+                        if clean_key in ['steps', 'seed']:
+                            try:
+                                gen_params[clean_key] = int(value.strip())
+                            except ValueError:
+                                gen_params[clean_key] = value.strip()
+                        elif clean_key in ['cfg_scale']:
+                            try:
+                                gen_params[clean_key] = float(value.strip())
+                            except ValueError:
+                                gen_params[clean_key] = value.strip()
+                        else:
+                            gen_params[clean_key] = value.strip()
+                    
+                    # Extract size if available and add to gen_params if a recognized key
+                    size_match = re.search(r'Size: (\d+)x(\d+)', params_section)
+                    if size_match and 'size' in GEN_PARAM_KEYS:
+                        width, height = size_match.groups()
+                        gen_params['size'] = f"{width}x{height}"
+                    
+                    # Add prompt and negative_prompt to gen_params if they're in GEN_PARAM_KEYS
+                    if 'prompt' in GEN_PARAM_KEYS and 'prompt' in metadata:
+                        gen_params['prompt'] = metadata['prompt']
+                    if 'negative_prompt' in GEN_PARAM_KEYS and 'negative_prompt' in metadata:
+                        gen_params['negative_prompt'] = metadata['negative_prompt']
+                    
+                    metadata["gen_params"] = gen_params
+            
+            # Extract LoRA and checkpoint information 
+            loras = []
+            base_model_counts = {}
+            checkpoint = None
+            
+            # First use Civitai resources if available (more reliable source)
+            if metadata.get("civitai_resources"):
+                for resource in metadata.get("civitai_resources", []):
+                    # --- Added: Parse 'air' field if present ---
+                    air = resource.get("air")
+                    if air:
+                        # Format: urn:air:sdxl:lora:civitai:1221007@1375651
+                        # Or: urn:air:sdxl:checkpoint:civitai:623891@2019115
+                        air_pattern = r"urn:air:[^:]+:(?P<type>[^:]+):civitai:(?P<modelId>\d+)@(?P<modelVersionId>\d+)"
+                        air_match = re.match(air_pattern, air)
+                        if air_match:
+                            air_type = air_match.group("type")
+                            air_modelId = int(air_match.group("modelId"))
+                            air_modelVersionId = int(air_match.group("modelVersionId"))
+                            # checkpoint/lycoris/lora/hypernet
+                            resource["type"] = air_type
+                            resource["modelId"] = air_modelId
+                            resource["modelVersionId"] = air_modelVersionId
+                    # --- End added ---
+
+                    if resource.get("type") == "checkpoint" and resource.get("modelVersionId"):
+                        version_id = resource.get("modelVersionId")
+                        version_id_str = str(version_id)
+                        checkpoint_entry = {
+                            'id': version_id,
+                            'modelId': resource.get("modelId", 0),
+                            'name': resource.get("modelName", "Unknown Checkpoint"),
+                            'version': resource.get("modelVersionName", resource.get("versionName", "")),
+                            'type': resource.get("type", "checkpoint"),
+                            'existsLocally': False,
+                            'localPath': None,
+                            'file_name': resource.get("modelName", ""),
+                            'hash': resource.get("hash", "") or "",
+                            'thumbnailUrl': '/loras_static/images/no-preview.png',
+                            'baseModel': '',
+                            'size': 0,
+                            'downloadUrl': '',
+                            'isDeleted': False
+                        }
+
+                        if metadata_provider:
+                            try:
+                                civitai_info = await metadata_provider.get_model_version_info(version_id_str)
+                                checkpoint_entry = await self.populate_checkpoint_from_civitai(
+                                    checkpoint_entry,
+                                    civitai_info
+                                )
+                            except Exception as e:
+                                logger.error(
+                                    "Error fetching Civitai info for checkpoint version %s: %s",
+                                    version_id,
+                                    e,
+                                )
+
+                        # Prefer the first checkpoint found
+                        if checkpoint_entry.get("baseModel"):
+                            base_model_value = checkpoint_entry["baseModel"]
+                            base_model_counts[base_model_value] = base_model_counts.get(base_model_value, 0) + 1
+
+                        if checkpoint is None:
+                            checkpoint = checkpoint_entry
+
+                        continue
+
+                    if resource.get("type") in ["lora", "lycoris", "hypernet"] and resource.get("modelVersionId"):
+                        # Initialize lora entry
+                        lora_entry = {
+                            'id': resource.get("modelVersionId", 0),
+                            'modelId': resource.get("modelId", 0),
+                            'name': resource.get("modelName", "Unknown LoRA"),
+                            'version': resource.get("modelVersionName", resource.get("versionName", "")),
+                            'type': resource.get("type", "lora"),
+                            'weight': round(float(resource.get("weight", 1.0)), 2),
+                            'existsLocally': False,
+                            'thumbnailUrl': '/loras_static/images/no-preview.png',
+                            'baseModel': '',
+                            'size': 0,
+                            'downloadUrl': '',
+                            'isDeleted': False
+                        }
+                        
+                        # Get additional info from Civitai
+                        if metadata_provider:
+                            try:
+                                civitai_info = await metadata_provider.get_model_version_info(resource.get("modelVersionId"))
+                                populated_entry = await self.populate_lora_from_civitai(
+                                    lora_entry,
+                                    civitai_info,
+                                    recipe_scanner,
+                                    base_model_counts
+                                )
+                                if populated_entry is None:
+                                    continue  # Skip invalid LoRA types
+                                lora_entry = populated_entry
+                            except Exception as e:
+                                logger.error(f"Error fetching Civitai info for LoRA {lora_entry['name']}: {e}")
+                        
+                        loras.append(lora_entry)
+            
+            # Fallback checkpoint parsing from generic "Model" and "Model hash" fields
+            if checkpoint is None:
+                model_hash = metadata.get("model_hash")
+                if not model_hash and metadata.get("hashes"):
+                    model_hash = metadata["hashes"].get("model")
+
+                model_name = metadata.get("model_name")
+                file_name = ""
+                if model_name:
+                    cleaned_name = re.split(r"[\\\\/]", model_name)[-1]
+                    file_name = os.path.splitext(cleaned_name)[0]
+
+                if model_hash or model_name:
+                    checkpoint_entry = {
+                        'id': 0,
+                        'modelId': 0,
+                        'name': model_name or "Unknown Checkpoint",
+                        'version': '',
+                        'type': 'checkpoint',
+                        'hash': model_hash or "",
+                        'existsLocally': False,
+                        'localPath': None,
+                        'file_name': file_name,
+                        'thumbnailUrl': '/loras_static/images/no-preview.png',
+                        'baseModel': '',
+                        'size': 0,
+                        'downloadUrl': '',
+                        'isDeleted': False
+                    }
+
+                    if metadata_provider and model_hash:
+                        try:
+                            civitai_info = await metadata_provider.get_model_by_hash(model_hash)
+                            checkpoint_entry = await self.populate_checkpoint_from_civitai(
+                                checkpoint_entry,
+                                civitai_info
+                            )
+                        except Exception as e:
+                            logger.error(f"Error fetching Civitai info for checkpoint hash {model_hash}: {e}")
+
+                    if checkpoint_entry.get("baseModel"):
+                        base_model_value = checkpoint_entry["baseModel"]
+                        base_model_counts[base_model_value] = base_model_counts.get(base_model_value, 0) + 1
+
+                    checkpoint = checkpoint_entry
+
+            # If no LoRAs from Civitai resources or to supplement, extract from metadata["hashes"]
+            if not loras or len(loras) == 0:
+                # Extract lora weights from extranet tags in prompt (for later use)
+                lora_weights = {}
+                lora_matches = re.findall(self.EXTRANETS_REGEX, prompt)
+                for lora_type, lora_name, lora_weight in lora_matches:
+                    key = f"{lora_type}:{lora_name}"
+                    lora_weights[key] = round(float(lora_weight), 2)
+                
+                # Use hashes from metadata as the primary source
+                if metadata.get("hashes"):
+                    for hash_key, lora_hash in metadata.get("hashes", {}).items():
+                        # Only process lora or hypernet types
+                        if not hash_key.startswith(("lora:", "hypernet:")):
+                            continue
+                            
+                        lora_type, lora_name = hash_key.split(':', 1)
+                        
+                        # Get weight from extranet tags if available, else default to 1.0
+                        weight = lora_weights.get(hash_key, 1.0)
+                        
+                        # Initialize lora entry
+                        lora_entry = {
+                            'name': lora_name,
+                            'type': lora_type,  # 'lora' or 'hypernet'
+                            'weight': weight,
+                            'hash': lora_hash,
+                            'existsLocally': False,
+                            'localPath': None,
+                            'file_name': lora_name,
+                            'thumbnailUrl': '/loras_static/images/no-preview.png',
+                            'baseModel': '',
+                            'size': 0,
+                            'downloadUrl': '',
+                            'isDeleted': False
+                        }
+                        
+                        # Try to get info from Civitai
+                        if metadata_provider:
+                            try:
+                                if lora_hash:
+                                    # If we have hash, use it for lookup
+                                    civitai_info = await metadata_provider.get_model_by_hash(lora_hash)
+                                else:
+                                    civitai_info = None
+                                
+                                populated_entry = await self.populate_lora_from_civitai(
+                                    lora_entry, 
+                                    civitai_info, 
+                                    recipe_scanner,
+                                    base_model_counts,
+                                    lora_hash
+                                )
+                                if populated_entry is None:
+                                    continue  # Skip invalid LoRA types
+                                lora_entry = populated_entry
+                            except Exception as e:
+                                logger.error(f"Error fetching Civitai info for LoRA {lora_name}: {e}")
+                        
+                        loras.append(lora_entry)
+                
+            # Try to get base model from resources or make educated guess
+            base_model = None
+            if checkpoint and checkpoint.get("baseModel"):
+                base_model = checkpoint.get("baseModel")
+            elif base_model_counts:
+                # Use the most common base model from the loras
+                base_model = max(base_model_counts.items(), key=lambda x: x[1])[0]
+            
+            # Prepare final result structure
+            # Make sure gen_params only contains recognized keys
+            filtered_gen_params = {}
+            for key in GEN_PARAM_KEYS:
+                if key in metadata.get("gen_params", {}):
+                    filtered_gen_params[key] = metadata["gen_params"][key]
+            
+            result = {
+                'base_model': base_model,
+                'loras': loras,
+                'gen_params': filtered_gen_params,
+                'from_automatic_metadata': True
+            }
+
+            if checkpoint:
+                result['checkpoint'] = checkpoint
+                result['model'] = checkpoint
+            
+            return result
+            
+        except Exception as e:
+            logger.error(f"Error parsing Automatic1111 metadata: {e}", exc_info=True)
+            return {"error": str(e), "loras": []}

py/recipes/parsers/civitai_image.py

@@ -0,0 +1,500 @@
+"""Parser for Civitai image metadata format."""
+
+import json
+import logging
+from typing import Dict, Any, Union
+from ..base import RecipeMetadataParser
+from ..constants import GEN_PARAM_KEYS
+from ...services.metadata_service import get_default_metadata_provider
+
+logger = logging.getLogger(__name__)
+
+class CivitaiApiMetadataParser(RecipeMetadataParser):
+    """Parser for Civitai image metadata format"""
+    
+    def is_metadata_matching(self, metadata) -> bool:
+        """Check if the metadata matches the Civitai image metadata format
+        
+        Args:
+            metadata: The metadata from the image (dict)
+            
+        Returns:
+            bool: True if this parser can handle the metadata
+        """
+        if not metadata or not isinstance(metadata, dict):
+            return False
+
+        def has_markers(payload: Dict[str, Any]) -> bool:
+            # Check for common CivitAI image metadata fields
+            civitai_image_fields = (
+                "resources",
+                "civitaiResources", 
+                "additionalResources",
+                "hashes",
+                "prompt",
+                "negativePrompt",
+                "steps",
+                "sampler",
+                "cfgScale",
+                "seed",
+                "width",
+                "height",
+                "Model",
+                "Model hash"
+            )
+            return any(key in payload for key in civitai_image_fields)
+
+        # Check the main metadata object
+        if has_markers(metadata):
+            return True
+
+        # Check for LoRA hash patterns
+        hashes = metadata.get("hashes")
+        if isinstance(hashes, dict) and any(str(key).lower().startswith("lora:") for key in hashes):
+            return True
+
+        # Check nested meta object (common in CivitAI image responses)
+        nested_meta = metadata.get("meta")
+        if isinstance(nested_meta, dict):
+            if has_markers(nested_meta):
+                return True
+
+            # Also check for LoRA hash patterns in nested meta
+            hashes = nested_meta.get("hashes")
+            if isinstance(hashes, dict) and any(str(key).lower().startswith("lora:") for key in hashes):
+                return True
+
+        return False
+    
+    async def parse_metadata(self, metadata, recipe_scanner=None, civitai_client=None) -> Dict[str, Any]:
+        """Parse metadata from Civitai image format
+        
+        Args:
+            metadata: The metadata from the image (dict)
+            recipe_scanner: Optional recipe scanner service
+            civitai_client: Optional Civitai API client (deprecated, use metadata_provider instead)
+            
+        Returns:
+            Dict containing parsed recipe data
+        """
+        try:
+            # Get metadata provider instead of using civitai_client directly
+            metadata_provider = await get_default_metadata_provider()
+
+            # Civitai image responses may wrap the actual metadata inside a "meta" key
+            if (
+                isinstance(metadata, dict)
+                and "meta" in metadata
+                and isinstance(metadata["meta"], dict)
+            ):
+                inner_meta = metadata["meta"]
+                if any(
+                    key in inner_meta
+                    for key in (
+                        "resources",
+                        "civitaiResources",
+                        "additionalResources",
+                        "hashes",
+                        "prompt",
+                        "negativePrompt",
+                    )
+                ):
+                    metadata = inner_meta
+            
+            # Initialize result structure
+            result = {
+                'base_model': None,
+                'loras': [],
+                'model': None,
+                'gen_params': {},
+                'from_civitai_image': True
+            }
+            
+            # Track already added LoRAs to prevent duplicates
+            added_loras = {}  # key: model_version_id or hash, value: index in result["loras"]
+            
+            # Extract hash information from hashes field for LoRA matching
+            lora_hashes = {}
+            if "hashes" in metadata and isinstance(metadata["hashes"], dict):
+                for key, hash_value in metadata["hashes"].items():
+                    key_str = str(key)
+                    if key_str.lower().startswith("lora:"):
+                        lora_name = key_str.split(":", 1)[1]
+                        lora_hashes[lora_name] = hash_value
+            
+            # Extract prompt and negative prompt
+            if "prompt" in metadata:
+                result["gen_params"]["prompt"] = metadata["prompt"]
+            
+            if "negativePrompt" in metadata:
+                result["gen_params"]["negative_prompt"] = metadata["negativePrompt"]
+            
+            # Extract other generation parameters
+            param_mapping = {
+                "steps": "steps",
+                "sampler": "sampler",
+                "cfgScale": "cfg_scale",
+                "seed": "seed",
+                "Size": "size",
+                "clipSkip": "clip_skip",
+            }
+            
+            for civitai_key, our_key in param_mapping.items():
+                if civitai_key in metadata and our_key in GEN_PARAM_KEYS:
+                    result["gen_params"][our_key] = metadata[civitai_key]
+            
+            # Extract base model information - directly if available
+            if "baseModel" in metadata:
+                result["base_model"] = metadata["baseModel"]
+            elif "Model hash" in metadata and metadata_provider:
+                model_hash = metadata["Model hash"]
+                model_info, error = await metadata_provider.get_model_by_hash(model_hash)
+                if model_info:
+                    result["base_model"] = model_info.get("baseModel", "")
+            elif "Model" in metadata and isinstance(metadata.get("resources"), list):
+                # Try to find base model in resources
+                for resource in metadata.get("resources", []):
+                    if resource.get("type") == "model" and resource.get("name") == metadata.get("Model"):
+                        # This is likely the checkpoint model
+                        if metadata_provider and resource.get("hash"):
+                            model_info, error = await metadata_provider.get_model_by_hash(resource.get("hash"))
+                            if model_info:
+                                result["base_model"] = model_info.get("baseModel", "")
+            
+            base_model_counts = {}
+            
+            # Process standard resources array
+            if "resources" in metadata and isinstance(metadata["resources"], list):
+                for resource in metadata["resources"]:
+                    # Modified to process resources without a type field as potential LoRAs
+                    if resource.get("type", "lora") == "lora":
+                        lora_hash = resource.get("hash", "")
+                        
+                        # Try to get hash from the hashes field if not present in resource
+                        if not lora_hash and resource.get("name"):
+                            lora_hash = lora_hashes.get(resource["name"], "")
+                        
+                        # Skip LoRAs without proper identification (hash or modelVersionId)
+                        if not lora_hash and not resource.get("modelVersionId"):
+                            logger.debug(f"Skipping LoRA resource '{resource.get('name', 'Unknown')}' - no hash or modelVersionId")
+                            continue
+                        
+                        # Skip if we've already added this LoRA by hash
+                        if lora_hash and lora_hash in added_loras:
+                            continue
+                        
+                        lora_entry = {
+                            'name': resource.get("name", "Unknown LoRA"),
+                            'type': "lora",
+                            'weight': float(resource.get("weight", 1.0)),
+                            'hash': lora_hash,
+                            'existsLocally': False,
+                            'localPath': None,
+                            'file_name': resource.get("name", "Unknown"),
+                            'thumbnailUrl': '/loras_static/images/no-preview.png',
+                            'baseModel': '',
+                            'size': 0,
+                            'downloadUrl': '',
+                            'isDeleted': False
+                        }
+                        
+                        # Try to get info from Civitai if hash is available
+                        if lora_entry['hash'] and metadata_provider:
+                            try:
+                                civitai_info = await metadata_provider.get_model_by_hash(lora_hash)
+                                
+                                populated_entry = await self.populate_lora_from_civitai(
+                                    lora_entry,
+                                    civitai_info,
+                                    recipe_scanner,
+                                    base_model_counts,
+                                    lora_hash
+                                )
+                                
+                                if populated_entry is None:
+                                    continue  # Skip invalid LoRA types
+                                    
+                                lora_entry = populated_entry
+                                
+                                # If we have a version ID from Civitai, track it for deduplication
+                                if 'id' in lora_entry and lora_entry['id']:
+                                    added_loras[str(lora_entry['id'])] = len(result["loras"])
+                            except Exception as e:
+                                logger.error(f"Error fetching Civitai info for LoRA hash {lora_entry['hash']}: {e}")
+                        
+                        # Track by hash if we have it
+                        if lora_hash:
+                            added_loras[lora_hash] = len(result["loras"])
+                            
+                        result["loras"].append(lora_entry)
+            
+            # Process civitaiResources array
+            if "civitaiResources" in metadata and isinstance(metadata["civitaiResources"], list):
+                for resource in metadata["civitaiResources"]:
+                    # Get resource type and identifier
+                    resource_type = str(resource.get("type") or "").lower()
+                    version_id = str(resource.get("modelVersionId", ""))
+
+                    if resource_type == "checkpoint":
+                        checkpoint_entry = {
+                            'id': resource.get("modelVersionId", 0),
+                            'modelId': resource.get("modelId", 0),
+                            'name': resource.get("modelName", "Unknown Checkpoint"),
+                            'version': resource.get("modelVersionName", ""),
+                            'type': resource.get("type", "checkpoint"),
+                            'existsLocally': False,
+                            'localPath': None,
+                            'file_name': resource.get("modelName", ""),
+                            'hash': resource.get("hash", "") or "",
+                            'thumbnailUrl': '/loras_static/images/no-preview.png',
+                            'baseModel': '',
+                            'size': 0,
+                            'downloadUrl': '',
+                            'isDeleted': False
+                        }
+
+                        if version_id and metadata_provider:
+                            try:
+                                civitai_info = await metadata_provider.get_model_version_info(version_id)
+
+                                checkpoint_entry = await self.populate_checkpoint_from_civitai(
+                                    checkpoint_entry,
+                                    civitai_info
+                                )
+                            except Exception as e:
+                                logger.error(f"Error fetching Civitai info for checkpoint version {version_id}: {e}")
+
+                        if result["model"] is None:
+                            result["model"] = checkpoint_entry
+
+                        continue
+
+                    # Skip if we've already added this LoRA
+                    if version_id and version_id in added_loras:
+                        continue
+
+                    # Initialize lora entry
+                    lora_entry = {
+                        'id': resource.get("modelVersionId", 0),
+                        'modelId': resource.get("modelId", 0),
+                        'name': resource.get("modelName", "Unknown LoRA"),
+                        'version': resource.get("modelVersionName", ""),
+                        'type': resource.get("type", "lora"),
+                        'weight': round(float(resource.get("weight", 1.0)), 2),
+                        'existsLocally': False,
+                        'thumbnailUrl': '/loras_static/images/no-preview.png',
+                        'baseModel': '',
+                        'size': 0,
+                        'downloadUrl': '',
+                        'isDeleted': False
+                    }
+
+                    # Try to get info from Civitai if modelVersionId is available
+                    if version_id and metadata_provider:
+                        try:
+                            # Use get_model_version_info instead of get_model_version
+                            civitai_info = await metadata_provider.get_model_version_info(version_id)
+
+                            populated_entry = await self.populate_lora_from_civitai(
+                                lora_entry,
+                                civitai_info,
+                                recipe_scanner,
+                                base_model_counts
+                            )
+
+                            if populated_entry is None:
+                                continue  # Skip invalid LoRA types
+
+                            lora_entry = populated_entry
+                        except Exception as e:
+                            logger.error(f"Error fetching Civitai info for model version {version_id}: {e}")
+
+                    # Track this LoRA in our deduplication dict
+                    if version_id:
+                        added_loras[version_id] = len(result["loras"])
+
+                    result["loras"].append(lora_entry)
+            
+            # Process additionalResources array
+            if "additionalResources" in metadata and isinstance(metadata["additionalResources"], list):
+                for resource in metadata["additionalResources"]:
+                    # Skip resources that aren't LoRAs or LyCORIS
+                    if resource.get("type") not in ["lora", "lycoris"] and "type" not in resource:
+                        continue
+                        
+                    lora_type = resource.get("type", "lora")
+                    name = resource.get("name", "")
+                    
+                    # Extract ID from URN format if available
+                    version_id = None
+                    if name and "civitai:" in name:
+                        parts = name.split("@")
+                        if len(parts) > 1:
+                            version_id = parts[1]
+                            
+                            # Skip if we've already added this LoRA
+                            if version_id in added_loras:
+                                continue
+                    
+                    lora_entry = {
+                        'name': name,
+                        'type': lora_type,
+                        'weight': float(resource.get("strength", 1.0)),
+                        'hash': "",
+                        'existsLocally': False,
+                        'localPath': None,
+                        'file_name': name,
+                        'thumbnailUrl': '/loras_static/images/no-preview.png',
+                        'baseModel': '',
+                        'size': 0,
+                        'downloadUrl': '',
+                        'isDeleted': False
+                    }
+                    
+                    # If we have a version ID and metadata provider, try to get more info
+                    if version_id and metadata_provider:
+                        try:
+                            # Use get_model_version_info with the version ID
+                            civitai_info = await metadata_provider.get_model_version_info(version_id)
+                            
+                            populated_entry = await self.populate_lora_from_civitai(
+                                lora_entry,
+                                civitai_info,
+                                recipe_scanner,
+                                base_model_counts
+                            )
+                            
+                            if populated_entry is None:
+                                continue  # Skip invalid LoRA types
+                                
+                            lora_entry = populated_entry
+                            
+                            # Track this LoRA for deduplication
+                            if version_id:
+                                added_loras[version_id] = len(result["loras"])
+                        except Exception as e:
+                            logger.error(f"Error fetching Civitai info for model ID {version_id}: {e}")
+                    
+                        result["loras"].append(lora_entry)
+
+            # If we found LoRA hashes in the metadata but haven't already
+            # populated entries for them, fall back to creating LoRAs from
+            # the hashes section. Some Civitai image responses only include
+            # LoRA information here without explicit resources entries.
+            for lora_name, lora_hash in lora_hashes.items():
+                if not lora_hash:
+                    continue
+
+                # Skip LoRAs we've already added via resources or other fields
+                if lora_hash in added_loras:
+                    continue
+
+                lora_entry = {
+                    'name': lora_name,
+                    'type': "lora",
+                    'weight': 1.0,
+                    'hash': lora_hash,
+                    'existsLocally': False,
+                    'localPath': None,
+                    'file_name': lora_name,
+                    'thumbnailUrl': '/loras_static/images/no-preview.png',
+                    'baseModel': '',
+                    'size': 0,
+                    'downloadUrl': '',
+                    'isDeleted': False
+                }
+
+                if metadata_provider:
+                    try:
+                        civitai_info = await metadata_provider.get_model_by_hash(lora_hash)
+
+                        populated_entry = await self.populate_lora_from_civitai(
+                            lora_entry,
+                            civitai_info,
+                            recipe_scanner,
+                            base_model_counts,
+                            lora_hash
+                        )
+
+                        if populated_entry is None:
+                            continue
+
+                        lora_entry = populated_entry
+
+                        if 'id' in lora_entry and lora_entry['id']:
+                            added_loras[str(lora_entry['id'])] = len(result["loras"])
+                    except Exception as e:
+                        logger.error(f"Error fetching Civitai info for LoRA hash {lora_hash}: {e}")
+
+                added_loras[lora_hash] = len(result["loras"])
+                result["loras"].append(lora_entry)
+
+            # Check for LoRA info in the format "Lora_0 Model hash", "Lora_0 Model name", etc.
+            lora_index = 0
+            while f"Lora_{lora_index} Model hash" in metadata and f"Lora_{lora_index} Model name" in metadata:
+                lora_hash = metadata[f"Lora_{lora_index} Model hash"]
+                lora_name = metadata[f"Lora_{lora_index} Model name"]
+                lora_strength_model = float(metadata.get(f"Lora_{lora_index} Strength model", 1.0))
+                
+                # Skip if we've already added this LoRA by hash
+                if lora_hash and lora_hash in added_loras:
+                    lora_index += 1
+                    continue
+                
+                lora_entry = {
+                    'name': lora_name,
+                    'type': "lora",
+                    'weight': lora_strength_model,
+                    'hash': lora_hash,
+                    'existsLocally': False,
+                    'localPath': None,
+                    'file_name': lora_name,
+                    'thumbnailUrl': '/loras_static/images/no-preview.png',
+                    'baseModel': '',
+                    'size': 0,
+                    'downloadUrl': '',
+                    'isDeleted': False
+                }
+                
+                # Try to get info from Civitai if hash is available
+                if lora_entry['hash'] and metadata_provider:
+                    try:
+                        civitai_info = await metadata_provider.get_model_by_hash(lora_hash)
+                        
+                        populated_entry = await self.populate_lora_from_civitai(
+                            lora_entry,
+                            civitai_info,
+                            recipe_scanner,
+                            base_model_counts,
+                            lora_hash
+                        )
+                        
+                        if populated_entry is None:
+                            lora_index += 1
+                            continue  # Skip invalid LoRA types
+                            
+                        lora_entry = populated_entry
+                        
+                        # If we have a version ID from Civitai, track it for deduplication
+                        if 'id' in lora_entry and lora_entry['id']:
+                            added_loras[str(lora_entry['id'])] = len(result["loras"])
+                    except Exception as e:
+                        logger.error(f"Error fetching Civitai info for LoRA hash {lora_entry['hash']}: {e}")
+                
+                # Track by hash if we have it
+                if lora_hash:
+                    added_loras[lora_hash] = len(result["loras"])
+                    
+                result["loras"].append(lora_entry)
+                
+                lora_index += 1
+            
+            # If base model wasn't found earlier, use the most common one from LoRAs
+            if not result["base_model"] and base_model_counts:
+                result["base_model"] = max(base_model_counts.items(), key=lambda x: x[1])[0]
+            
+            return result
+            
+        except Exception as e:
+            logger.error(f"Error parsing Civitai image metadata: {e}", exc_info=True)
+            return {"error": str(e), "loras": []}

py/recipes/parsers/comfy.py

@@ -0,0 +1,217 @@
+"""Parser for ComfyUI metadata format."""
+
+import re
+import json
+import logging
+from typing import Dict, Any
+from ..base import RecipeMetadataParser
+from ..constants import GEN_PARAM_KEYS
+from ...services.metadata_service import get_default_metadata_provider
+
+logger = logging.getLogger(__name__)
+
+class ComfyMetadataParser(RecipeMetadataParser):
+    """Parser for Civitai ComfyUI metadata JSON format"""
+    
+    METADATA_MARKER = r"class_type"
+    
+    def is_metadata_matching(self, user_comment: str) -> bool:
+        """Check if the user comment matches the ComfyUI metadata format"""
+        try:
+            data = json.loads(user_comment)
+            # Check if it contains class_type nodes typical of ComfyUI workflow
+            return isinstance(data, dict) and any(isinstance(v, dict) and 'class_type' in v for v in data.values())
+        except (json.JSONDecodeError, TypeError):
+            return False
+    
+    async def parse_metadata(self, user_comment: str, recipe_scanner=None, civitai_client=None) -> Dict[str, Any]:
+        """Parse metadata from Civitai ComfyUI metadata format"""
+        try:
+            # Get metadata provider instead of using civitai_client directly
+            metadata_provider = await get_default_metadata_provider()
+            
+            data = json.loads(user_comment)
+            loras = []
+            
+            # Find all LoraLoader nodes
+            lora_nodes = {k: v for k, v in data.items() if isinstance(v, dict) and v.get('class_type') == 'LoraLoader'}
+            
+            # Process each LoraLoader node
+            for node_id, node in lora_nodes.items():
+                if 'inputs' not in node or 'lora_name' not in node['inputs']:
+                    continue
+                    
+                lora_name = node['inputs'].get('lora_name', '')
+                
+                # Parse the URN to extract model ID and version ID
+                # Format: "urn:air:sdxl:lora:civitai:1107767@1253442"
+                lora_id_match = re.search(r'civitai:(\d+)@(\d+)', lora_name)
+                if not lora_id_match:
+                    continue
+                    
+                model_id = lora_id_match.group(1)
+                model_version_id = lora_id_match.group(2)
+                
+                # Get strength from node inputs
+                weight = node['inputs'].get('strength_model', 1.0)
+                
+                # Initialize lora entry with default values
+                lora_entry = {
+                    'id': model_version_id,
+                    'modelId': model_id,
+                    'name': f"Lora {model_id}",  # Default name
+                    'version': '',
+                    'type': 'lora',
+                    'weight': weight,
+                    'existsLocally': False,
+                    'localPath': None,
+                    'file_name': '',
+                    'hash': '',
+                    'thumbnailUrl': '/loras_static/images/no-preview.png',
+                    'baseModel': '',
+                    'size': 0,
+                    'downloadUrl': '',
+                    'isDeleted': False
+                }
+                
+                # Get additional info from Civitai if metadata provider is available
+                if metadata_provider:
+                    try:
+                        civitai_info_tuple = await metadata_provider.get_model_version_info(model_version_id)
+                        # Populate lora entry with Civitai info
+                        populated_entry = await self.populate_lora_from_civitai(
+                            lora_entry, 
+                            civitai_info_tuple, 
+                            recipe_scanner
+                        )
+                        if populated_entry is None:
+                            continue  # Skip invalid LoRA types
+                        lora_entry = populated_entry
+                    except Exception as e:
+                        logger.error(f"Error fetching Civitai info for LoRA: {e}")
+                
+                loras.append(lora_entry)
+            
+            # Find checkpoint info
+            checkpoint_nodes = {k: v for k, v in data.items() if isinstance(v, dict) and v.get('class_type') == 'CheckpointLoaderSimple'}
+            checkpoint = None
+            checkpoint_id = None
+            checkpoint_version_id = None
+            
+            if checkpoint_nodes:
+                # Get the first checkpoint node
+                checkpoint_node = next(iter(checkpoint_nodes.values()))
+                if 'inputs' in checkpoint_node and 'ckpt_name' in checkpoint_node['inputs']:
+                    checkpoint_name = checkpoint_node['inputs']['ckpt_name']
+                    # Parse checkpoint URN
+                    checkpoint_match = re.search(r'civitai:(\d+)@(\d+)', checkpoint_name)
+                    if checkpoint_match:
+                        checkpoint_id = checkpoint_match.group(1)
+                        checkpoint_version_id = checkpoint_match.group(2)
+                        checkpoint = {
+                            'id': checkpoint_version_id,
+                            'modelId': checkpoint_id,
+                            'name': f"Checkpoint {checkpoint_id}",
+                            'version': '',
+                            'type': 'checkpoint'
+                        }
+                        
+                        # Get additional checkpoint info from Civitai
+                        if metadata_provider:
+                            try:
+                                civitai_info_tuple = await metadata_provider.get_model_version_info(checkpoint_version_id)
+                                civitai_info, _ = civitai_info_tuple if isinstance(civitai_info_tuple, tuple) else (civitai_info_tuple, None)
+                                # Populate checkpoint with Civitai info
+                                checkpoint = await self.populate_checkpoint_from_civitai(checkpoint, civitai_info)
+                            except Exception as e:
+                                logger.error(f"Error fetching Civitai info for checkpoint: {e}")
+            
+            # Extract generation parameters
+            gen_params = {}
+            
+            # First try to get from extraMetadata
+            if 'extraMetadata' in data:
+                try:
+                    # extraMetadata is a JSON string that needs to be parsed
+                    extra_metadata = json.loads(data['extraMetadata'])
+                    
+                    # Map fields from extraMetadata to our standard format
+                    mapping = {
+                        'prompt': 'prompt',
+                        'negativePrompt': 'negative_prompt',
+                        'steps': 'steps',
+                        'sampler': 'sampler',
+                        'cfgScale': 'cfg_scale',
+                        'seed': 'seed'
+                    }
+                    
+                    for src_key, dest_key in mapping.items():
+                        if src_key in extra_metadata:
+                            gen_params[dest_key] = extra_metadata[src_key]
+                    
+                    # If size info is available, format as "width x height"
+                    if 'width' in extra_metadata and 'height' in extra_metadata:
+                        gen_params['size'] = f"{extra_metadata['width']}x{extra_metadata['height']}"
+                    
+                except Exception as e:
+                    logger.error(f"Error parsing extraMetadata: {e}")
+            
+            # If extraMetadata doesn't have all the info, try to get from nodes
+            if not gen_params or len(gen_params) < 3:  # At least we want prompt, negative_prompt, and steps
+                # Find positive prompt node
+                positive_nodes = {k: v for k, v in data.items() if isinstance(v, dict) and 
+                                v.get('class_type', '').endswith('CLIPTextEncode') and 
+                                v.get('_meta', {}).get('title') == 'Positive'}
+                
+                if positive_nodes:
+                    positive_node = next(iter(positive_nodes.values()))
+                    if 'inputs' in positive_node and 'text' in positive_node['inputs']:
+                        gen_params['prompt'] = positive_node['inputs']['text']
+                
+                # Find negative prompt node
+                negative_nodes = {k: v for k, v in data.items() if isinstance(v, dict) and 
+                                v.get('class_type', '').endswith('CLIPTextEncode') and 
+                                v.get('_meta', {}).get('title') == 'Negative'}
+                
+                if negative_nodes:
+                    negative_node = next(iter(negative_nodes.values()))
+                    if 'inputs' in negative_node and 'text' in negative_node['inputs']:
+                        gen_params['negative_prompt'] = negative_node['inputs']['text']
+                
+                # Find KSampler node for other parameters
+                ksampler_nodes = {k: v for k, v in data.items() if isinstance(v, dict) and v.get('class_type') == 'KSampler'}
+                
+                if ksampler_nodes:
+                    ksampler_node = next(iter(ksampler_nodes.values()))
+                    if 'inputs' in ksampler_node:
+                        inputs = ksampler_node['inputs']
+                        if 'sampler_name' in inputs:
+                            gen_params['sampler'] = inputs['sampler_name']
+                        if 'steps' in inputs:
+                            gen_params['steps'] = inputs['steps']
+                        if 'cfg' in inputs:
+                            gen_params['cfg_scale'] = inputs['cfg']
+                        if 'seed' in inputs:
+                            gen_params['seed'] = inputs['seed']
+            
+            # Determine base model from loras info
+            base_model = None
+            if loras:
+                # Use the most common base model from loras
+                base_models = [lora['baseModel'] for lora in loras if lora.get('baseModel')]
+                if base_models:
+                    from collections import Counter
+                    base_model_counts = Counter(base_models)
+                    base_model = base_model_counts.most_common(1)[0][0]
+            
+            return {
+                'base_model': base_model,
+                'loras': loras,
+                'checkpoint': checkpoint,
+                'gen_params': gen_params,
+                'from_comfy_metadata': True
+            }
+            
+        except Exception as e:
+            logger.error(f"Error parsing ComfyUI metadata: {e}", exc_info=True)
+            return {"error": str(e), "loras": []}

py/recipes/parsers/meta_format.py

@@ -0,0 +1,219 @@
+"""Parser for meta format (Lora_N Model hash) metadata."""
+
+import os
+import re
+import logging
+from typing import Dict, Any
+from ..base import RecipeMetadataParser
+from ..constants import GEN_PARAM_KEYS
+from ...services.metadata_service import get_default_metadata_provider
+
+logger = logging.getLogger(__name__)
+
+class MetaFormatParser(RecipeMetadataParser):
+    """Parser for images with meta format metadata (Lora_N Model hash format)"""
+    
+    METADATA_MARKER = r'Lora_\d+ Model hash:'
+    
+    def is_metadata_matching(self, user_comment: str) -> bool:
+        """Check if the user comment matches the metadata format"""
+        return re.search(self.METADATA_MARKER, user_comment, re.IGNORECASE | re.DOTALL) is not None
+    
+    async def parse_metadata(self, user_comment: str, recipe_scanner=None, civitai_client=None) -> Dict[str, Any]:
+        """Parse metadata from images with meta format metadata (Lora_N Model hash format)"""
+        try:
+            # Get metadata provider instead of using civitai_client directly
+            metadata_provider = await get_default_metadata_provider()
+            
+            # Extract prompt and negative prompt
+            parts = user_comment.split('Negative prompt:', 1)
+            prompt = parts[0].strip()
+            
+            # Initialize metadata
+            metadata = {"prompt": prompt, "loras": []}
+            
+            # Extract negative prompt and parameters if available
+            if len(parts) > 1:
+                negative_and_params = parts[1]
+                
+                # Extract negative prompt - everything until the first parameter (usually "Steps:")
+                param_start = re.search(r'([A-Za-z]+): ', negative_and_params)
+                if param_start:
+                    neg_prompt = negative_and_params[:param_start.start()].strip()
+                    metadata["negative_prompt"] = neg_prompt
+                    params_section = negative_and_params[param_start.start():]
+                else:
+                    params_section = negative_and_params
+                
+                # Extract key-value parameters (Steps, Sampler, Seed, etc.)
+                param_pattern = r'([A-Za-z_0-9 ]+): ([^,]+)'
+                params = re.findall(param_pattern, params_section)
+                for key, value in params:
+                    clean_key = key.strip().lower().replace(' ', '_')
+                    metadata[clean_key] = value.strip()
+            
+            # Extract LoRA information
+            # Pattern to match lora entries: Lora_0 Model name: ArtVador I.safetensors, Lora_0 Model hash: 08f7133a58, etc.
+            lora_pattern = r'Lora_(\d+) Model name: ([^,]+), Lora_\1 Model hash: ([^,]+), Lora_\1 Strength model: ([^,]+), Lora_\1 Strength clip: ([^,]+)'
+            lora_matches = re.findall(lora_pattern, user_comment)
+            
+            # If the regular pattern doesn't match, try a more flexible approach
+            if not lora_matches:
+                # First find all Lora indices
+                lora_indices = set(re.findall(r'Lora_(\d+)', user_comment))
+                
+                # For each index, extract the information
+                for idx in lora_indices:
+                    lora_info = {}
+                    
+                    # Extract model name
+                    name_match = re.search(f'Lora_{idx} Model name: ([^,]+)', user_comment)
+                    if name_match:
+                        lora_info['name'] = name_match.group(1).strip()
+                    
+                    # Extract model hash
+                    hash_match = re.search(f'Lora_{idx} Model hash: ([^,]+)', user_comment)
+                    if hash_match:
+                        lora_info['hash'] = hash_match.group(1).strip()
+                    
+                    # Extract strength model
+                    strength_model_match = re.search(f'Lora_{idx} Strength model: ([^,]+)', user_comment)
+                    if strength_model_match:
+                        lora_info['strength_model'] = float(strength_model_match.group(1).strip())
+                    
+                    # Extract strength clip
+                    strength_clip_match = re.search(f'Lora_{idx} Strength clip: ([^,]+)', user_comment)
+                    if strength_clip_match:
+                        lora_info['strength_clip'] = float(strength_clip_match.group(1).strip())
+                    
+                    # Only add if we have at least name and hash
+                    if 'name' in lora_info and 'hash' in lora_info:
+                        lora_matches.append((idx, lora_info['name'], lora_info['hash'], 
+                                            str(lora_info.get('strength_model', 1.0)), 
+                                            str(lora_info.get('strength_clip', 1.0))))
+            
+            # Process LoRAs
+            base_model_counts = {}
+            loras = []
+            
+            for match in lora_matches:
+                if len(match) == 5:  # Regular pattern match
+                    idx, name, hash_value, strength_model, strength_clip = match
+                else:  # Flexible approach match
+                    continue  # Should not happen now
+                
+                # Clean up the values
+                name = name.strip()
+                if name.endswith('.safetensors'):
+                    name = name[:-12]  # Remove .safetensors extension
+                    
+                hash_value = hash_value.strip()
+                weight = float(strength_model)  # Use model strength as weight
+                
+                # Initialize lora entry with default values
+                lora_entry = {
+                    'name': name,
+                    'type': 'lora',
+                    'weight': weight,
+                    'existsLocally': False,
+                    'localPath': None,
+                    'file_name': name,
+                    'hash': hash_value,
+                    'thumbnailUrl': '/loras_static/images/no-preview.png',
+                    'baseModel': '',
+                    'size': 0,
+                    'downloadUrl': '',
+                    'isDeleted': False
+                }
+                
+                # Get info from Civitai by hash if available
+                if metadata_provider and hash_value:
+                    try:
+                        civitai_info = await metadata_provider.get_model_by_hash(hash_value)
+                        # Populate lora entry with Civitai info
+                        populated_entry = await self.populate_lora_from_civitai(
+                            lora_entry, 
+                            civitai_info, 
+                            recipe_scanner,
+                            base_model_counts,
+                            hash_value
+                        )
+                        if populated_entry is None:
+                            continue  # Skip invalid LoRA types
+                        lora_entry = populated_entry
+                    except Exception as e:
+                        logger.error(f"Error fetching Civitai info for LoRA hash {hash_value}: {e}")
+                
+                loras.append(lora_entry)
+            
+            # Extract checkpoint information from generic Model/Model hash fields
+            checkpoint = None
+            model_hash = metadata.get("model_hash")
+            model_name = metadata.get("model")
+
+            if model_hash or model_name:
+                cleaned_name = None
+                if model_name:
+                    cleaned_name = re.split(r"[\\\\/]", model_name)[-1]
+                    cleaned_name = os.path.splitext(cleaned_name)[0]
+
+                checkpoint_entry = {
+                    'id': 0,
+                    'modelId': 0,
+                    'name': model_name or "Unknown Checkpoint",
+                    'version': '',
+                    'type': 'checkpoint',
+                    'hash': model_hash or "",
+                    'existsLocally': False,
+                    'localPath': None,
+                    'file_name': cleaned_name or (model_name or ""),
+                    'thumbnailUrl': '/loras_static/images/no-preview.png',
+                    'baseModel': '',
+                    'size': 0,
+                    'downloadUrl': '',
+                    'isDeleted': False
+                }
+
+                if metadata_provider and model_hash:
+                    try:
+                        civitai_info = await metadata_provider.get_model_by_hash(model_hash)
+                        checkpoint_entry = await self.populate_checkpoint_from_civitai(
+                            checkpoint_entry,
+                            civitai_info
+                        )
+                    except Exception as e:
+                        logger.error(f"Error fetching Civitai info for checkpoint hash {model_hash}: {e}")
+
+                if checkpoint_entry.get("baseModel"):
+                    base_model_value = checkpoint_entry["baseModel"]
+                    base_model_counts[base_model_value] = base_model_counts.get(base_model_value, 0) + 1
+
+                checkpoint = checkpoint_entry
+            
+            # Set base_model to the most common one from civitai_info or checkpoint
+            base_model = checkpoint["baseModel"] if checkpoint and checkpoint.get("baseModel") else None
+            if not base_model and base_model_counts:
+                base_model = max(base_model_counts.items(), key=lambda x: x[1])[0]
+            
+            # Extract generation parameters for recipe metadata
+            gen_params = {}
+            for key in GEN_PARAM_KEYS:
+                if key in metadata:
+                    gen_params[key] = metadata.get(key, '')
+            
+            # Try to extract size information if available
+            if 'width' in metadata and 'height' in metadata:
+                gen_params['size'] = f"{metadata['width']}x{metadata['height']}"
+            
+            return {
+                'base_model': base_model,
+                'loras': loras,
+                'gen_params': gen_params,
+                'raw_metadata': metadata,
+                'from_meta_format': True,
+                **({'checkpoint': checkpoint, 'model': checkpoint} if checkpoint else {})
+            }
+            
+        except Exception as e:
+            logger.error(f"Error parsing meta format metadata: {e}", exc_info=True)
+            return {"error": str(e), "loras": []}

py/recipes/parsers/recipe_format.py

@@ -0,0 +1,202 @@
+"""Parser for dedicated recipe metadata format."""
+
+import re
+import json
+import logging
+from typing import Dict, Any, Optional
+from ...config import config
+from ..base import RecipeMetadataParser
+from ..constants import GEN_PARAM_KEYS
+from ...services.metadata_service import get_default_metadata_provider
+
+logger = logging.getLogger(__name__)
+
+class RecipeFormatParser(RecipeMetadataParser):
+    """Parser for images with dedicated recipe metadata format"""
+    
+    # Regular expression pattern for extracting recipe metadata
+    METADATA_MARKER = r'Recipe metadata: (\{.*\})'
+
+    async def _get_lora_from_version_index(self, recipe_scanner, model_version_id: Any) -> Optional[Dict[str, Any]]:
+        """Return a cached LoRA entry by modelVersionId if available."""
+
+        if not recipe_scanner or not getattr(recipe_scanner, "_lora_scanner", None):
+            return None
+
+        try:
+            normalized_id = int(model_version_id)
+        except (TypeError, ValueError):
+            return None
+
+        try:
+            cache = await recipe_scanner._lora_scanner.get_cached_data()
+        except Exception as exc:  # pragma: no cover - defensive logging
+            logger.debug("Unable to load lora cache for version lookup: %s", exc)
+            return None
+
+        if not cache or not getattr(cache, "version_index", None):
+            return None
+
+        return cache.version_index.get(normalized_id)
+    
+    def is_metadata_matching(self, user_comment: str) -> bool:
+        """Check if the user comment matches the metadata format"""
+        return re.search(self.METADATA_MARKER, user_comment, re.IGNORECASE | re.DOTALL) is not None
+    
+    async def parse_metadata(self, user_comment: str, recipe_scanner=None, civitai_client=None) -> Dict[str, Any]:
+        """Parse metadata from images with dedicated recipe metadata format"""
+        try:
+            # Get metadata provider instead of using civitai_client directly
+            metadata_provider = await get_default_metadata_provider()
+            
+            # Extract recipe metadata from user comment
+            try:
+                # Look for recipe metadata section
+                recipe_match = re.search(self.METADATA_MARKER, user_comment, re.IGNORECASE | re.DOTALL)
+                if not recipe_match:
+                    recipe_metadata = None
+                else:
+                    recipe_json = recipe_match.group(1)
+                    recipe_metadata = json.loads(recipe_json)
+            except Exception as e:
+                logger.error(f"Error extracting recipe metadata: {e}")
+                recipe_metadata = None
+            if not recipe_metadata:
+                return {"error": "No recipe metadata found", "loras": []}
+                
+            # Process the recipe metadata
+            loras = []
+            for lora in recipe_metadata.get('loras', []):
+                # Convert recipe lora format to frontend format
+                lora_entry = {
+                    'id': int(lora.get('modelVersionId', 0)),
+                    'name': lora.get('modelName', ''),
+                    'version': lora.get('modelVersionName', ''),
+                    'type': 'lora',
+                    'weight': lora.get('strength', 1.0),
+                    'file_name': lora.get('file_name', ''),
+                    'hash': lora.get('hash', ''),
+                    'existsLocally': False,
+                    'inLibrary': False,
+                    'localPath': None,
+                    'thumbnailUrl': '/loras_static/images/no-preview.png',
+                    'size': 0
+                }
+                
+                # Check if this LoRA exists locally by SHA256 hash
+                if recipe_scanner:
+                    lora_scanner = recipe_scanner._lora_scanner
+
+                    if lora.get('hash'):
+                        exists_locally = lora_scanner.has_hash(lora['hash'])
+                        if exists_locally:
+                            lora_cache = await lora_scanner.get_cached_data()
+                            lora_item = next((item for item in lora_cache.raw_data if item['sha256'].lower() == lora['hash'].lower()), None)
+                            if lora_item:
+                                lora_entry['existsLocally'] = True
+                                lora_entry['inLibrary'] = True
+                                lora_entry['localPath'] = lora_item['file_path']
+                                lora_entry['file_name'] = lora_item['file_name']
+                                lora_entry['size'] = lora_item['size']
+                                lora_entry['thumbnailUrl'] = config.get_preview_static_url(lora_item['preview_url'])
+                                
+                        else:
+                            lora_entry['existsLocally'] = False
+                            lora_entry['inLibrary'] = False
+                            lora_entry['localPath'] = None
+
+                    # If we still don't have a local match, try matching by modelVersionId
+                    if not lora_entry['existsLocally'] and lora.get('modelVersionId') is not None:
+                        cached_lora = await self._get_lora_from_version_index(recipe_scanner, lora.get('modelVersionId'))
+                        if cached_lora:
+                            lora_entry['existsLocally'] = True
+                            lora_entry['inLibrary'] = True
+                            lora_entry['localPath'] = cached_lora.get('file_path')
+                            lora_entry['file_name'] = cached_lora.get('file_name') or lora_entry['file_name']
+                            lora_entry['size'] = cached_lora.get('size', lora_entry['size'])
+                            if cached_lora.get('sha256'):
+                                lora_entry['hash'] = cached_lora['sha256']
+                            preview_url = cached_lora.get('preview_url')
+                            if preview_url:
+                                lora_entry['thumbnailUrl'] = config.get_preview_static_url(preview_url)
+
+                    # Try to get additional info from Civitai if we have a model version ID and still missing locally
+                    if not lora_entry['existsLocally'] and lora.get('modelVersionId') and metadata_provider:
+                        try:
+                            civitai_info_tuple = await metadata_provider.get_model_version_info(lora['modelVersionId'])
+                            # Populate lora entry with Civitai info
+                            populated_entry = await self.populate_lora_from_civitai(
+                                lora_entry, 
+                                civitai_info_tuple, 
+                                recipe_scanner,
+                                None,  # No need to track base model counts
+                                lora_entry.get('hash', '')
+                            )
+                            if populated_entry is None:
+                                continue  # Skip invalid LoRA types
+                            lora_entry = populated_entry
+                        except Exception as e:
+                            logger.error(f"Error fetching Civitai info for LoRA: {e}")
+                            lora_entry['thumbnailUrl'] = '/loras_static/images/no-preview.png'
+                
+                loras.append(lora_entry)
+            
+            logger.info(f"Found {len(loras)} loras in recipe metadata")
+
+            # Process checkpoint information if present
+            checkpoint = None
+            checkpoint_data = recipe_metadata.get('checkpoint') or {}
+            if isinstance(checkpoint_data, dict) and checkpoint_data:
+                version_id = checkpoint_data.get('modelVersionId') or checkpoint_data.get('id')
+                checkpoint_entry = {
+                    'id': version_id or 0,
+                    'modelId': checkpoint_data.get('modelId', 0),
+                    'name': checkpoint_data.get('name', 'Unknown Checkpoint'),
+                    'version': checkpoint_data.get('version', ''),
+                    'type': checkpoint_data.get('type', 'checkpoint'),
+                    'hash': checkpoint_data.get('hash', ''),
+                    'existsLocally': False,
+                    'localPath': None,
+                    'file_name': checkpoint_data.get('file_name', ''),
+                    'thumbnailUrl': '/loras_static/images/no-preview.png',
+                    'baseModel': '',
+                    'size': 0,
+                    'downloadUrl': '',
+                    'isDeleted': False
+                }
+
+                if metadata_provider:
+                    try:
+                        civitai_info = None
+                        if version_id:
+                            civitai_info = await metadata_provider.get_model_version_info(str(version_id))
+                        elif checkpoint_entry.get('hash'):
+                            civitai_info = await metadata_provider.get_model_by_hash(checkpoint_entry['hash'])
+
+                        if civitai_info:
+                            checkpoint_entry = await self.populate_checkpoint_from_civitai(checkpoint_entry, civitai_info)
+                    except Exception as e:
+                        logger.error(f"Error fetching Civitai info for checkpoint in recipe metadata: {e}")
+
+                checkpoint = checkpoint_entry
+            
+            # Filter gen_params to only include recognized keys
+            filtered_gen_params = {}
+            if 'gen_params' in recipe_metadata:
+                for key, value in recipe_metadata['gen_params'].items():
+                    if key in GEN_PARAM_KEYS:
+                        filtered_gen_params[key] = value
+            
+            return {
+                'base_model': checkpoint['baseModel'] if checkpoint and checkpoint.get('baseModel') else recipe_metadata.get('base_model', ''),
+                'loras': loras,
+                'gen_params': filtered_gen_params,
+                'tags': recipe_metadata.get('tags', []),
+                'title': recipe_metadata.get('title', ''),
+                'from_recipe_metadata': True,
+                **({'checkpoint': checkpoint, 'model': checkpoint} if checkpoint else {})
+            }
+            
+        except Exception as e:
+            logger.error(f"Error parsing recipe format metadata: {e}", exc_info=True)
+            return {"error": str(e), "loras": []}

py/routes/api_routes.py

@@ -1,818 +0,0 @@
-import os
-import json
-import logging
-from aiohttp import web
-from typing import Dict, List
-
-from ..services.file_monitor import LoraFileMonitor
-from ..services.download_manager import DownloadManager
-from ..services.civitai_client import CivitaiClient
-from ..config import config
-from ..services.lora_scanner import LoraScanner
-from operator import itemgetter
-from ..services.websocket_manager import ws_manager
-from ..services.settings_manager import settings
-import asyncio
-from .update_routes import UpdateRoutes
-
-logger = logging.getLogger(__name__)
-
-class ApiRoutes:
-    """API route handlers for LoRA management"""
-
-    def __init__(self, file_monitor: LoraFileMonitor):
-        self.scanner = LoraScanner()
-        self.civitai_client = CivitaiClient()
-        self.download_manager = DownloadManager(file_monitor)
-        self._download_lock = asyncio.Lock()
-
-    @classmethod
-    def setup_routes(cls, app: web.Application, monitor: LoraFileMonitor):
-        """Register API routes"""
-        routes = cls(monitor)
-        app.router.add_post('/api/delete_model', routes.delete_model)
-        app.router.add_post('/api/fetch-civitai', routes.fetch_civitai)
-        app.router.add_post('/api/replace_preview', routes.replace_preview)
-        app.router.add_get('/api/loras', routes.get_loras)
-        app.router.add_post('/api/fetch-all-civitai', routes.fetch_all_civitai)
-        app.router.add_get('/ws/fetch-progress', ws_manager.handle_connection)
-        app.router.add_get('/api/lora-roots', routes.get_lora_roots)
-        app.router.add_get('/api/civitai/versions/{model_id}', routes.get_civitai_versions)
-        app.router.add_post('/api/download-lora', routes.download_lora)
-        app.router.add_post('/api/settings', routes.update_settings)
-        app.router.add_post('/api/move_model', routes.move_model)
-        app.router.add_get('/api/lora-model-description', routes.get_lora_model_description)  # Add new route
-        app.router.add_post('/loras/api/save-metadata', routes.save_metadata)
-        app.router.add_get('/api/lora-preview-url', routes.get_lora_preview_url)  # Add new route
-        app.router.add_post('/api/move_models_bulk', routes.move_models_bulk)
-        app.router.add_get('/api/top-tags', routes.get_top_tags)  # Add new route for top tags
-
-        # Add update check routes
-        UpdateRoutes.setup_routes(app)
-
-    async def delete_model(self, request: web.Request) -> web.Response:
-        """Handle model deletion request"""
-        try:
-            data = await request.json()
-            file_path = data.get('file_path')
-            if not file_path:
-                return web.Response(text='Model path is required', status=400)
-
-            target_dir = os.path.dirname(file_path)
-            file_name = os.path.splitext(os.path.basename(file_path))[0]
-            
-            deleted_files = await self._delete_model_files(target_dir, file_name)
-            
-            return web.json_response({
-                'success': True,
-                'deleted_files': deleted_files
-            })
-            
-        except Exception as e:
-            logger.error(f"Error deleting model: {e}", exc_info=True)
-            return web.Response(text=str(e), status=500)
-
-    async def fetch_civitai(self, request: web.Request) -> web.Response:
-        """Handle CivitAI metadata fetch request"""
-        try:
-            data = await request.json()
-            metadata_path = os.path.splitext(data['file_path'])[0] + '.metadata.json'
-            
-            # Check if model is from CivitAI
-            local_metadata = await self._load_local_metadata(metadata_path)
-
-            # Fetch and update metadata
-            civitai_metadata = await self.civitai_client.get_model_by_hash(local_metadata["sha256"])
-            if not civitai_metadata:
-                return await self._handle_not_found_on_civitai(metadata_path, local_metadata)
-
-            await self._update_model_metadata(metadata_path, local_metadata, civitai_metadata, self.civitai_client)
-            
-            return web.json_response({"success": True})
-
-        except Exception as e:
-            logger.error(f"Error fetching from CivitAI: {e}", exc_info=True)
-            return web.json_response({"success": False, "error": str(e)}, status=500)
-
-    async def replace_preview(self, request: web.Request) -> web.Response:
-        """Handle preview image replacement request"""
-        try:
-            reader = await request.multipart()
-            preview_data, content_type = await self._read_preview_file(reader)
-            model_path = await self._read_model_path(reader)
-            
-            preview_path = await self._save_preview_file(model_path, preview_data, content_type)
-            await self._update_preview_metadata(model_path, preview_path)
-            
-            # Update preview URL in scanner cache
-            await self.scanner.update_preview_in_cache(model_path, preview_path)
-            
-            return web.json_response({
-                "success": True,
-                "preview_url": config.get_preview_static_url(preview_path)
-            })
-            
-        except Exception as e:
-            logger.error(f"Error replacing preview: {e}", exc_info=True)
-            return web.Response(text=str(e), status=500)
-
-    async def get_loras(self, request: web.Request) -> web.Response:
-        """Handle paginated LoRA data request"""
-        try:
-            # Parse query parameters
-            page = int(request.query.get('page', '1'))
-            page_size = int(request.query.get('page_size', '20'))
-            sort_by = request.query.get('sort_by', 'name')
-            folder = request.query.get('folder')
-            search = request.query.get('search', '').lower()
-            fuzzy = request.query.get('fuzzy', 'false').lower() == 'true'
-            recursive = request.query.get('recursive', 'false').lower() == 'true'
-            
-            # Parse base models filter parameter
-            base_models = request.query.get('base_models', '').split(',')
-            base_models = [model.strip() for model in base_models if model.strip()]
-            
-            # Parse search options
-            search_filename = request.query.get('search_filename', 'true').lower() == 'true'
-            search_modelname = request.query.get('search_modelname', 'true').lower() == 'true'
-            search_tags = request.query.get('search_tags', 'false').lower() == 'true'
-            
-            # Validate parameters
-            if page < 1 or page_size < 1 or page_size > 100:
-                return web.json_response({
-                    'error': 'Invalid pagination parameters'
-                }, status=400)
-            
-            if sort_by not in ['date', 'name']:
-                return web.json_response({
-                    'error': 'Invalid sort parameter'
-                }, status=400)
-            
-            # Parse tags filter parameter
-            tags = request.query.get('tags', '').split(',')
-            tags = [tag.strip() for tag in tags if tag.strip()]
-            
-            # Get paginated data with search and filters
-            result = await self.scanner.get_paginated_data(
-                page=page,
-                page_size=page_size,
-                sort_by=sort_by,
-                folder=folder,
-                search=search,
-                fuzzy=fuzzy,
-                recursive=recursive,
-                base_models=base_models,  # Pass base models filter
-                tags=tags,  # Add tags parameter
-                search_options={
-                    'filename': search_filename,
-                    'modelname': search_modelname,
-                    'tags': search_tags
-                }
-            )
-            
-            # Format the response data
-            formatted_items = [
-                self._format_lora_response(item) 
-                for item in result['items']
-            ]
-
-            # Get all available folders from cache
-            cache = await self.scanner.get_cached_data()
-            
-            return web.json_response({
-                'items': formatted_items,
-                'total': result['total'],
-                'page': result['page'],
-                'page_size': result['page_size'],
-                'total_pages': result['total_pages'],
-                'folders': cache.folders
-            })
-            
-        except Exception as e:
-            logger.error(f"Error in get_loras: {str(e)}", exc_info=True)
-            return web.json_response({
-                'error': 'Internal server error'
-            }, status=500)
-
-    def _format_lora_response(self, lora: Dict) -> Dict:
-        """Format LoRA data for API response"""
-        return {
-            "model_name": lora["model_name"],
-            "file_name": lora["file_name"],
-            "preview_url": config.get_preview_static_url(lora["preview_url"]),
-            "base_model": lora["base_model"],
-            "folder": lora["folder"],
-            "sha256": lora["sha256"],
-            "file_path": lora["file_path"].replace(os.sep, "/"),
-            "file_size": lora["size"],
-            "modified": lora["modified"],
-            "tags": lora["tags"],
-            "modelDescription": lora["modelDescription"],
-            "from_civitai": lora.get("from_civitai", True),
-            "usage_tips": lora.get("usage_tips", ""),
-            "notes": lora.get("notes", ""),
-            "civitai": self._filter_civitai_data(lora.get("civitai", {}))
-        }
-
-    def _filter_civitai_data(self, data: Dict) -> Dict:
-        """Filter relevant fields from CivitAI data"""
-        if not data:
-            return {}
-            
-        fields = [
-            "id", "modelId", "name", "createdAt", "updatedAt", 
-            "publishedAt", "trainedWords", "baseModel", "description",
-            "model", "images"
-        ]
-        return {k: data[k] for k in fields if k in data}
-
-    # Private helper methods
-    async def _delete_model_files(self, target_dir: str, file_name: str) -> List[str]:
-        """Delete model and associated files"""
-        patterns = [
-            f"{file_name}.safetensors",  # Required
-            f"{file_name}.metadata.json",
-            f"{file_name}.preview.png",
-            f"{file_name}.preview.jpg",
-            f"{file_name}.preview.jpeg",
-            f"{file_name}.preview.webp",
-            f"{file_name}.preview.mp4",
-            f"{file_name}.png",
-            f"{file_name}.jpg",
-            f"{file_name}.jpeg",
-            f"{file_name}.webp",
-            f"{file_name}.mp4"
-        ]
-        
-        deleted = []
-        main_file = patterns[0]
-        main_path = os.path.join(target_dir, main_file).replace(os.sep, '/')
-        
-        if os.path.exists(main_path):
-            # Notify file monitor to ignore delete event
-            self.download_manager.file_monitor.handler.add_ignore_path(main_path, 0)
-            
-            # Delete file
-            os.remove(main_path)
-            deleted.append(main_path)
-        else:
-            logger.warning(f"Model file not found: {main_file}")
-
-        # Remove from cache
-        cache = await self.scanner.get_cached_data()
-        cache.raw_data = [item for item in cache.raw_data if item['file_path'] != main_path]
-        await cache.resort()
-        
-        # Delete optional files
-        for pattern in patterns[1:]:
-            path = os.path.join(target_dir, pattern)
-            if os.path.exists(path):
-                try:
-                    os.remove(path)
-                    deleted.append(pattern)
-                except Exception as e:
-                    logger.warning(f"Failed to delete {pattern}: {e}")
-                    
-        return deleted
-
-    async def _read_preview_file(self, reader) -> tuple[bytes, str]:
-        """Read preview file and content type from multipart request"""
-        field = await reader.next()
-        if field.name != 'preview_file':
-            raise ValueError("Expected 'preview_file' field")
-        content_type = field.headers.get('Content-Type', 'image/png')
-        return await field.read(), content_type
-
-    async def _read_model_path(self, reader) -> str:
-        """Read model path from multipart request"""
-        field = await reader.next()
-        if field.name != 'model_path':
-            raise ValueError("Expected 'model_path' field")
-        return (await field.read()).decode()
-
-    async def _save_preview_file(self, model_path: str, preview_data: bytes, content_type: str) -> str:
-        """Save preview file and return its path"""
-        # Determine file extension based on content type
-        if content_type.startswith('video/'):
-            extension = '.preview.mp4'
-        else:
-            extension = '.preview.png'
-        
-        base_name = os.path.splitext(os.path.basename(model_path))[0]
-        folder = os.path.dirname(model_path)
-        preview_path = os.path.join(folder, base_name + extension).replace(os.sep, '/')
-        
-        with open(preview_path, 'wb') as f:
-            f.write(preview_data)
-            
-        return preview_path
-
-    async def _update_preview_metadata(self, model_path: str, preview_path: str):
-        """Update preview path in metadata"""
-        metadata_path = os.path.splitext(model_path)[0] + '.metadata.json'
-        if os.path.exists(metadata_path):
-            try:
-                with open(metadata_path, 'r', encoding='utf-8') as f:
-                    metadata = json.load(f)
-                
-                # Update preview_url directly in the metadata dict
-                metadata['preview_url'] = preview_path
-                
-                with open(metadata_path, 'w', encoding='utf-8') as f:
-                    json.dump(metadata, f, indent=2, ensure_ascii=False)
-            except Exception as e:
-                logger.error(f"Error updating metadata: {e}")
-
-    async def _load_local_metadata(self, metadata_path: str) -> Dict:
-        """Load local metadata file"""
-        if os.path.exists(metadata_path):
-            try:
-                with open(metadata_path, 'r', encoding='utf-8') as f:
-                    return json.load(f)
-            except Exception as e:
-                logger.error(f"Error loading metadata from {metadata_path}: {e}")
-        return {}
-
-    async def _handle_not_found_on_civitai(self, metadata_path: str, local_metadata: Dict) -> web.Response:
-        """Handle case when model is not found on CivitAI"""
-        local_metadata['from_civitai'] = False
-        with open(metadata_path, 'w', encoding='utf-8') as f:
-            json.dump(local_metadata, f, indent=2, ensure_ascii=False)
-        return web.json_response(
-            {"success": False, "error": "Not found on CivitAI"}, 
-            status=404
-        )
-
-    async def _update_model_metadata(self, metadata_path: str, local_metadata: Dict, 
-                                   civitai_metadata: Dict, client: CivitaiClient) -> None:
-        """Update local metadata with CivitAI data"""
-        local_metadata['civitai'] = civitai_metadata
-        
-        # Update model name if available
-        if 'model' in civitai_metadata:
-            local_metadata['model_name'] = civitai_metadata['model'].get('name', 
-                                                                       local_metadata.get('model_name'))
-        
-            # Fetch additional model metadata (description and tags) if we have model ID
-            model_id = civitai_metadata['modelId']
-            if model_id:
-                model_metadata = await client.get_model_metadata(str(model_id))
-                if model_metadata:
-                    local_metadata['modelDescription'] = model_metadata.get('description', '')
-                    local_metadata['tags'] = model_metadata.get('tags', [])
-        
-        # Update base model
-        local_metadata['base_model'] = civitai_metadata.get('baseModel')
-        
-        # Update preview if needed
-        if not local_metadata.get('preview_url') or not os.path.exists(local_metadata['preview_url']):
-            first_preview = next((img for img in civitai_metadata.get('images', [])), None)
-            if first_preview:
-                preview_ext = '.mp4' if first_preview['type'] == 'video' else os.path.splitext(first_preview['url'])[-1]
-                base_name = os.path.splitext(os.path.splitext(os.path.basename(metadata_path))[0])[0]
-                preview_filename = base_name + preview_ext
-                preview_path = os.path.join(os.path.dirname(metadata_path), preview_filename)
-                
-                if await client.download_preview_image(first_preview['url'], preview_path):
-                    local_metadata['preview_url'] = preview_path.replace(os.sep, '/')
-
-        # Save updated metadata
-        with open(metadata_path, 'w', encoding='utf-8') as f:
-            json.dump(local_metadata, f, indent=2, ensure_ascii=False)
-
-        await self.scanner.update_single_lora_cache(local_metadata['file_path'], local_metadata['file_path'], local_metadata)
-
-    async def fetch_all_civitai(self, request: web.Request) -> web.Response:
-        """Fetch CivitAI metadata for all loras in the background"""
-        try:
-            cache = await self.scanner.get_cached_data()
-            total = len(cache.raw_data)
-            processed = 0
-            success = 0
-            needs_resort = False
-            
-            # 准备要处理的 loras
-            to_process = [
-                lora for lora in cache.raw_data 
-                if lora.get('sha256') and (not lora.get('civitai') or 'id' not in lora.get('civitai')) and lora.get('from_civitai')  # TODO: for lora not from CivitAI but added traineWords
-            ]
-            total_to_process = len(to_process)
-            
-            # 发送初始进度
-            await ws_manager.broadcast({
-                'status': 'started',
-                'total': total_to_process,
-                'processed': 0,
-                'success': 0
-            })
-            
-            for lora in to_process:
-                try:
-                    original_name = lora.get('model_name')
-                    if await self._fetch_and_update_single_lora(
-                        sha256=lora['sha256'],
-                        file_path=lora['file_path'],
-                        lora=lora
-                    ):
-                        success += 1
-                        if original_name != lora.get('model_name'):
-                            needs_resort = True
-                    
-                    processed += 1
-                    
-                    # 每处理一个就发送进度更新
-                    await ws_manager.broadcast({
-                        'status': 'processing',
-                        'total': total_to_process,
-                        'processed': processed,
-                        'success': success,
-                        'current_name': lora.get('model_name', 'Unknown')
-                    })
-                    
-                except Exception as e:
-                    logger.error(f"Error fetching CivitAI data for {lora['file_path']}: {e}")
-            
-            if needs_resort:
-                await cache.resort(name_only=True)
-            
-            # 发送完成消息
-            await ws_manager.broadcast({
-                'status': 'completed',
-                'total': total_to_process,
-                'processed': processed,
-                'success': success
-            })
-                    
-            return web.json_response({
-                "success": True,
-                "message": f"Successfully updated {success} of {processed} processed loras (total: {total})"
-            })
-            
-        except Exception as e:
-            # 发送错误消息
-            await ws_manager.broadcast({
-                'status': 'error',
-                'error': str(e)
-            })
-            logger.error(f"Error in fetch_all_civitai: {e}")
-            return web.Response(text=str(e), status=500)
-
-    async def _fetch_and_update_single_lora(self, sha256: str, file_path: str, lora: dict) -> bool:
-        """Fetch and update metadata for a single lora without sorting
-        
-        Args:
-            sha256: SHA256 hash of the lora file
-            file_path: Path to the lora file
-            lora: The lora object in cache to update
-            
-        Returns:
-            bool: True if successful, False otherwise
-        """
-        client = CivitaiClient()
-        try:
-            metadata_path = os.path.splitext(file_path)[0] + '.metadata.json'
-            
-            # Check if model is from CivitAI
-            local_metadata = await self._load_local_metadata(metadata_path)
-
-            # Fetch metadata
-            civitai_metadata = await client.get_model_by_hash(sha256)
-            if not civitai_metadata:
-                # Mark as not from CivitAI if not found
-                local_metadata['from_civitai'] = False
-                lora['from_civitai'] = False
-                with open(metadata_path, 'w', encoding='utf-8') as f:
-                    json.dump(local_metadata, f, indent=2, ensure_ascii=False)
-                return False
-
-            # Update metadata
-            await self._update_model_metadata(
-                metadata_path, 
-                local_metadata, 
-                civitai_metadata, 
-                client
-            )
-            
-            # Update cache object directly
-            lora.update({
-                'model_name': local_metadata.get('model_name'),
-                'preview_url': local_metadata.get('preview_url'),
-                'from_civitai': True,
-                'civitai': civitai_metadata
-            })
-                
-            return True
-
-        except Exception as e:
-            logger.error(f"Error fetching CivitAI data: {e}")
-            return False
-        finally:
-            await client.close()
-
-    async def get_lora_roots(self, request: web.Request) -> web.Response:
-        """Get all configured LoRA root directories"""
-        return web.json_response({
-            'roots': config.loras_roots
-        })
-
-    async def get_civitai_versions(self, request: web.Request) -> web.Response:
-        """Get available versions for a Civitai model with local availability info"""
-        try:
-            model_id = request.match_info['model_id']
-            versions = await self.civitai_client.get_model_versions(model_id)
-            if not versions:
-                return web.Response(status=404, text="Model not found")
-            
-            # Check local availability for each version
-            for version in versions:
-                for file in version.get('files', []):
-                    sha256 = file.get('hashes', {}).get('SHA256')
-                    if sha256:
-                        file['existsLocally'] = self.scanner.has_lora_hash(sha256)
-                        if file['existsLocally']:
-                            file['localPath'] = self.scanner.get_lora_path_by_hash(sha256)
-                        
-            return web.json_response(versions)
-        except Exception as e:
-            logger.error(f"Error fetching model versions: {e}")
-            return web.Response(status=500, text=str(e))
-
-    async def download_lora(self, request: web.Request) -> web.Response:
-        async with self._download_lock:
-            try:
-                data = await request.json()
-                
-                # Create progress callback
-                async def progress_callback(progress):
-                    await ws_manager.broadcast({
-                        'status': 'progress',
-                        'progress': progress
-                    })
-                
-                result = await self.download_manager.download_from_civitai(
-                    download_url=data.get('download_url'),
-                    save_dir=data.get('lora_root'),
-                    relative_path=data.get('relative_path'),
-                    progress_callback=progress_callback  # Add progress callback
-                )
-                
-                if not result.get('success', False):
-                    return web.Response(status=500, text=result.get('error', 'Unknown error'))
-                
-                return web.json_response(result)
-            except Exception as e:
-                logger.error(f"Error downloading LoRA: {e}")
-                return web.Response(status=500, text=str(e))
-
-    async def update_settings(self, request: web.Request) -> web.Response:
-        """Update application settings"""
-        try:
-            data = await request.json()
-            
-            # Validate and update settings
-            if 'civitai_api_key' in data:
-                settings.set('civitai_api_key', data['civitai_api_key'])
-            
-            return web.json_response({'success': True})
-        except Exception as e:
-            logger.error(f"Error updating settings: {e}", exc_info=True)  # 添加 exc_info=True 以获取完整堆栈
-            return web.Response(status=500, text=str(e))
-
-    async def move_model(self, request: web.Request) -> web.Response:
-        """Handle model move request"""
-        try:
-            data = await request.json()
-            file_path = data.get('file_path')
-            target_path = data.get('target_path')
-            
-            if not file_path or not target_path:
-                return web.Response(text='File path and target path are required', status=400)
-
-            # Call scanner to handle the move operation
-            success = await self.scanner.move_model(file_path, target_path)
-            
-            if success:
-                return web.json_response({'success': True})
-            else:
-                return web.Response(text='Failed to move model', status=500)
-                
-        except Exception as e:
-            logger.error(f"Error moving model: {e}", exc_info=True)
-            return web.Response(text=str(e), status=500)
-
-    @classmethod
-    async def cleanup(cls):
-        """Add cleanup method for application shutdown"""
-        if hasattr(cls, '_instance'):
-            await cls._instance.civitai_client.close()
-
-    async def save_metadata(self, request: web.Request) -> web.Response:
-        """Handle saving metadata updates"""
-        try:
-            data = await request.json()
-            file_path = data.get('file_path')
-            if not file_path:
-                return web.Response(text='File path is required', status=400)
-
-            # Remove file path from data to avoid saving it
-            metadata_updates = {k: v for k, v in data.items() if k != 'file_path'}
-            
-            # Get metadata file path
-            metadata_path = os.path.splitext(file_path)[0] + '.metadata.json'
-            
-            # Load existing metadata
-            if os.path.exists(metadata_path):
-                with open(metadata_path, 'r', encoding='utf-8') as f:
-                    metadata = json.load(f)
-            else:
-                metadata = {}
-
-            # Handle nested updates (for civitai.trainedWords)
-            for key, value in metadata_updates.items():
-                if isinstance(value, dict) and key in metadata and isinstance(metadata[key], dict):
-                    # Deep update for nested dictionaries
-                    for nested_key, nested_value in value.items():
-                        metadata[key][nested_key] = nested_value
-                else:
-                    # Regular update for top-level keys
-                    metadata[key] = value
-
-            # Save updated metadata
-            with open(metadata_path, 'w', encoding='utf-8') as f:
-                json.dump(metadata, f, indent=2, ensure_ascii=False)
-
-            # Update cache
-            await self.scanner.update_single_lora_cache(file_path, file_path, metadata)
-
-            # If model_name was updated, resort the cache
-            if 'model_name' in metadata_updates:
-                cache = await self.scanner.get_cached_data()
-                await cache.resort(name_only=True)
-
-            return web.json_response({'success': True})
-
-        except Exception as e:
-            logger.error(f"Error saving metadata: {e}", exc_info=True)
-            return web.Response(text=str(e), status=500)
-
-    async def get_lora_preview_url(self, request: web.Request) -> web.Response:
-        """Get the static preview URL for a LoRA file"""
-        try:
-            # Get lora file name from query parameters
-            lora_name = request.query.get('name')
-            if not lora_name:
-                return web.Response(text='Lora file name is required', status=400)
-
-            # Get cache data
-            cache = await self.scanner.get_cached_data()
-            
-            # Search for the lora in cache data
-            for lora in cache.raw_data:
-                file_name = lora['file_name']
-                if file_name == lora_name:
-                    if preview_url := lora.get('preview_url'):
-                        # Convert preview path to static URL
-                        static_url = config.get_preview_static_url(preview_url)
-                        if static_url:
-                            return web.json_response({
-                                'success': True,
-                                'preview_url': static_url
-                            })
-                    break
-
-            # If no preview URL found
-            return web.json_response({
-                'success': False,
-                'error': 'No preview URL found for the specified lora'
-            }, status=404)
-
-        except Exception as e:
-            logger.error(f"Error getting lora preview URL: {e}", exc_info=True)
-            return web.Response(text=str(e), status=500)
-
-    async def move_models_bulk(self, request: web.Request) -> web.Response:
-        """Handle bulk model move request"""
-        try:
-            data = await request.json()
-            file_paths = data.get('file_paths', [])
-            target_path = data.get('target_path')
-            
-            if not file_paths or not target_path:
-                return web.Response(text='File paths and target path are required', status=400)
-
-            results = []
-            for file_path in file_paths:
-                success = await self.scanner.move_model(file_path, target_path)
-                results.append({"path": file_path, "success": success})
-            
-            # Count successes
-            success_count = sum(1 for r in results if r["success"])
-            
-            if success_count == len(file_paths):
-                return web.json_response({
-                    'success': True,
-                    'message': f'Successfully moved {success_count} models'
-                })
-            elif success_count > 0:
-                return web.json_response({
-                    'success': True,
-                    'message': f'Moved {success_count} of {len(file_paths)} models',
-                    'results': results
-                })
-            else:
-                return web.Response(text='Failed to move any models', status=500)
-                
-        except Exception as e:
-            logger.error(f"Error moving models in bulk: {e}", exc_info=True)
-            return web.Response(text=str(e), status=500)
-
-    async def get_lora_model_description(self, request: web.Request) -> web.Response:
-        """Get model description for a Lora model"""
-        try:
-            # Get parameters
-            model_id = request.query.get('model_id')
-            file_path = request.query.get('file_path')
-            
-            if not model_id:
-                return web.json_response({
-                    'success': False, 
-                    'error': 'Model ID is required'
-                }, status=400)
-            
-            # Check if we already have the description stored in metadata
-            description = None
-            tags = []
-            if file_path:
-                metadata_path = os.path.splitext(file_path)[0] + '.metadata.json'
-                if os.path.exists(metadata_path):
-                    try:
-                        with open(metadata_path, 'r', encoding='utf-8') as f:
-                            metadata = json.load(f)
-                            description = metadata.get('modelDescription')
-                            tags = metadata.get('tags', [])
-                    except Exception as e:
-                        logger.error(f"Error loading metadata from {metadata_path}: {e}")
-            
-            # If description is not in metadata, fetch from CivitAI
-            if not description:
-                logger.info(f"Fetching model metadata for model ID: {model_id}")
-                model_metadata = await self.civitai_client.get_model_metadata(model_id)
-                
-                if model_metadata:
-                    description = model_metadata.get('description')
-                    tags = model_metadata.get('tags', [])
-                
-                    # Save the metadata to file if we have a file path and got metadata
-                    if file_path:
-                        try:
-                            metadata_path = os.path.splitext(file_path)[0] + '.metadata.json'
-                            if os.path.exists(metadata_path):
-                                with open(metadata_path, 'r', encoding='utf-8') as f:
-                                    metadata = json.load(f)
-                                
-                                metadata['modelDescription'] = description
-                                metadata['tags'] = tags
-                                
-                                with open(metadata_path, 'w', encoding='utf-8') as f:
-                                    json.dump(metadata, f, indent=2, ensure_ascii=False)
-                                    logger.info(f"Saved model metadata to file for {file_path}")
-                        except Exception as e:
-                            logger.error(f"Error saving model metadata: {e}")
-            
-            return web.json_response({
-                'success': True,
-                'description': description or "<p>No model description available.</p>",
-                'tags': tags
-            })
-            
-        except Exception as e:
-            logger.error(f"Error getting model metadata: {e}", exc_info=True)
-            return web.json_response({
-                'success': False,
-                'error': str(e)
-            }, status=500)
-
-    async def get_top_tags(self, request: web.Request) -> web.Response:
-        """Handle request for top tags sorted by frequency"""
-        try:
-            # Parse query parameters
-            limit = int(request.query.get('limit', '20'))
-            
-            # Validate limit
-            if limit < 1 or limit > 100:
-                limit = 20  # Default to a reasonable limit
-                
-            # Get top tags
-            top_tags = await self.scanner.get_top_tags(limit)
-            
-            return web.json_response({
-                'success': True,
-                'tags': top_tags
-            })
-            
-        except Exception as e:
-            logger.error(f"Error getting top tags: {str(e)}", exc_info=True)
-            return web.json_response({
-                'success': False,
-                'error': 'Internal server error'
-            }, status=500)

py/routes/base_model_routes.py

@@ -0,0 +1,300 @@
+from __future__ import annotations
+
+import logging
+from abc import ABC, abstractmethod
+from typing import TYPE_CHECKING, Callable, Dict, Mapping
+
+import jinja2
+from aiohttp import web
+
+from ..config import config
+from ..services.download_coordinator import DownloadCoordinator
+from ..services.downloader import get_downloader
+from ..services.metadata_service import get_default_metadata_provider, get_metadata_provider
+from ..services.metadata_sync_service import MetadataSyncService
+from ..services.model_file_service import ModelFileService, ModelMoveService
+from ..services.model_lifecycle_service import ModelLifecycleService
+from ..services.preview_asset_service import PreviewAssetService
+from ..services.server_i18n import server_i18n as default_server_i18n
+from ..services.service_registry import ServiceRegistry
+from ..services.settings_manager import get_settings_manager
+from ..services.tag_update_service import TagUpdateService
+from ..services.websocket_manager import ws_manager as default_ws_manager
+from ..services.use_cases import (
+    AutoOrganizeUseCase,
+    BulkMetadataRefreshUseCase,
+    DownloadModelUseCase,
+)
+from ..services.websocket_progress_callback import (
+    WebSocketBroadcastCallback,
+    WebSocketProgressCallback,
+)
+from ..utils.exif_utils import ExifUtils
+from ..utils.metadata_manager import MetadataManager
+from .model_route_registrar import COMMON_ROUTE_DEFINITIONS, ModelRouteRegistrar
+from .handlers.model_handlers import (
+    ModelAutoOrganizeHandler,
+    ModelCivitaiHandler,
+    ModelDownloadHandler,
+    ModelHandlerSet,
+    ModelListingHandler,
+    ModelManagementHandler,
+    ModelMoveHandler,
+    ModelPageView,
+    ModelQueryHandler,
+    ModelUpdateHandler,
+)
+
+if TYPE_CHECKING:
+    from ..services.model_update_service import ModelUpdateService
+
+logger = logging.getLogger(__name__)
+
+
+class BaseModelRoutes(ABC):
+    """Base route controller for all model types."""
+
+    template_name: str | None = None
+
+    def __init__(
+        self,
+        service=None,
+        *,
+        settings_service=None,
+        ws_manager=default_ws_manager,
+        server_i18n=default_server_i18n,
+        metadata_provider_factory=get_default_metadata_provider,
+    ) -> None:
+        self.service = None
+        self.model_type = ""
+        self._settings = settings_service or get_settings_manager()
+        self._ws_manager = ws_manager
+        self._server_i18n = server_i18n
+        self._metadata_provider_factory = metadata_provider_factory
+
+        self.template_env = jinja2.Environment(
+            loader=jinja2.FileSystemLoader(config.templates_path),
+            autoescape=True,
+        )
+
+        self.model_file_service: ModelFileService | None = None
+        self.model_move_service: ModelMoveService | None = None
+        self.model_lifecycle_service: ModelLifecycleService | None = None
+        self.websocket_progress_callback = WebSocketProgressCallback()
+        self.metadata_progress_callback = WebSocketBroadcastCallback()
+
+        self._handler_set: ModelHandlerSet | None = None
+        self._handler_mapping: Dict[str, Callable[[web.Request], web.StreamResponse]] | None = None
+
+        self._preview_service = PreviewAssetService(
+            metadata_manager=MetadataManager,
+            downloader_factory=get_downloader,
+            exif_utils=ExifUtils,
+        )
+        self._metadata_sync_service = MetadataSyncService(
+            metadata_manager=MetadataManager,
+            preview_service=self._preview_service,
+            settings=self._settings,
+            default_metadata_provider_factory=metadata_provider_factory,
+            metadata_provider_selector=get_metadata_provider,
+        )
+        self._tag_update_service = TagUpdateService(metadata_manager=MetadataManager)
+        self._download_coordinator = DownloadCoordinator(
+            ws_manager=self._ws_manager,
+            download_manager_factory=ServiceRegistry.get_download_manager,
+        )
+        self._model_update_service: ModelUpdateService | None = None
+
+        if service is not None:
+            self.attach_service(service)
+
+    def set_model_update_service(self, service: "ModelUpdateService") -> None:
+        """Attach the model update tracking service."""
+
+        self._model_update_service = service
+        self._handler_set = None
+        self._handler_mapping = None
+
+    def attach_service(self, service) -> None:
+        """Attach a model service and rebuild handler dependencies."""
+        self.service = service
+        self.model_type = service.model_type
+        self.model_file_service = ModelFileService(service.scanner, service.model_type)
+        self.model_move_service = ModelMoveService(service.scanner, service.model_type)
+        self.model_lifecycle_service = ModelLifecycleService(
+            scanner=service.scanner,
+            metadata_manager=MetadataManager,
+            metadata_loader=self._metadata_sync_service.load_local_metadata,
+            recipe_scanner_factory=ServiceRegistry.get_recipe_scanner,
+            update_service=self._model_update_service,
+        )
+        self._handler_set = None
+        self._handler_mapping = None
+
+    def _ensure_handler_mapping(self) -> Mapping[str, Callable[[web.Request], web.StreamResponse]]:
+        if self._handler_mapping is None:
+            handler_set = self._create_handler_set()
+            self._handler_set = handler_set
+            self._handler_mapping = handler_set.to_route_mapping()
+        return self._handler_mapping
+
+    def _create_handler_set(self) -> ModelHandlerSet:
+        service = self._ensure_service()
+        update_service = self._ensure_model_update_service()
+        page_view = ModelPageView(
+            template_env=self.template_env,
+            template_name=self.template_name or "",
+            service=service,
+            settings_service=self._settings,
+            server_i18n=self._server_i18n,
+            logger=logger,
+        )
+        listing = ModelListingHandler(
+            service=service,
+            parse_specific_params=self._parse_specific_params,
+            logger=logger,
+        )
+        management = ModelManagementHandler(
+            service=service,
+            logger=logger,
+            metadata_sync=self._metadata_sync_service,
+            preview_service=self._preview_service,
+            tag_update_service=self._tag_update_service,
+            lifecycle_service=self._ensure_lifecycle_service(),
+        )
+        query = ModelQueryHandler(service=service, logger=logger)
+        download_use_case = DownloadModelUseCase(download_coordinator=self._download_coordinator)
+        download = ModelDownloadHandler(
+            ws_manager=self._ws_manager,
+            logger=logger,
+            download_use_case=download_use_case,
+            download_coordinator=self._download_coordinator,
+        )
+        metadata_refresh_use_case = BulkMetadataRefreshUseCase(
+            service=service,
+            metadata_sync=self._metadata_sync_service,
+            settings_service=self._settings,
+            logger=logger,
+        )
+        civitai = ModelCivitaiHandler(
+            service=service,
+            settings_service=self._settings,
+            ws_manager=self._ws_manager,
+            logger=logger,
+            metadata_provider_factory=self._metadata_provider_factory,
+            validate_model_type=self._validate_civitai_model_type,
+            expected_model_types=self._get_expected_model_types,
+            find_model_file=self._find_model_file,
+            metadata_sync=self._metadata_sync_service,
+            metadata_refresh_use_case=metadata_refresh_use_case,
+            metadata_progress_callback=self.metadata_progress_callback,
+        )
+        move = ModelMoveHandler(move_service=self._ensure_move_service(), logger=logger)
+        auto_organize_use_case = AutoOrganizeUseCase(
+            file_service=self._ensure_file_service(),
+            lock_provider=self._ws_manager,
+        )
+        auto_organize = ModelAutoOrganizeHandler(
+            use_case=auto_organize_use_case,
+            progress_callback=self.websocket_progress_callback,
+            ws_manager=self._ws_manager,
+            logger=logger,
+        )
+        updates = ModelUpdateHandler(
+            service=service,
+            update_service=update_service,
+            metadata_provider_selector=get_metadata_provider,
+            logger=logger,
+        )
+        return ModelHandlerSet(
+            page_view=page_view,
+            listing=listing,
+            management=management,
+            query=query,
+            download=download,
+            civitai=civitai,
+            move=move,
+            auto_organize=auto_organize,
+            updates=updates,
+        )
+
+    @property
+    def route_handlers(self) -> Mapping[str, Callable[[web.Request], web.StreamResponse]]:
+        return self._ensure_handler_mapping()
+
+    def setup_routes(self, app: web.Application, prefix: str) -> None:
+        registrar = ModelRouteRegistrar(app)
+        handler_lookup = {
+            definition.handler_name: self._make_handler_proxy(definition.handler_name)
+            for definition in COMMON_ROUTE_DEFINITIONS
+        }
+        registrar.register_common_routes(prefix, handler_lookup)
+        self.setup_specific_routes(registrar, prefix)
+
+    @abstractmethod
+    def setup_specific_routes(self, registrar: ModelRouteRegistrar, prefix: str) -> None:
+        """Setup model-specific routes."""
+        raise NotImplementedError
+
+    def _parse_specific_params(self, request: web.Request) -> Dict:
+        """Parse model-specific parameters - to be overridden by subclasses."""
+        return {}
+
+    def _validate_civitai_model_type(self, model_type: str) -> bool:
+        """Validate CivitAI model type - to be overridden by subclasses."""
+        return True
+
+    def _get_expected_model_types(self) -> str:
+        """Get expected model types string for error messages - to be overridden by subclasses."""
+        return "any model type"
+
+    def _find_model_file(self, files):
+        """Find the appropriate model file from the files list - can be overridden by subclasses."""
+        return next((file for file in files if file.get("type") == "Model" and file.get("primary") is True), None)
+
+    def get_handler(self, name: str) -> Callable[[web.Request], web.StreamResponse]:
+        """Expose handlers for subclasses or tests."""
+        return self._ensure_handler_mapping()[name]
+
+    def _ensure_service(self):
+        if self.service is None:
+            raise RuntimeError("Model service has not been attached")
+        return self.service
+
+    def _ensure_file_service(self) -> ModelFileService:
+        if self.model_file_service is None:
+            service = self._ensure_service()
+            self.model_file_service = ModelFileService(service.scanner, service.model_type)
+        return self.model_file_service
+
+    def _ensure_move_service(self) -> ModelMoveService:
+        if self.model_move_service is None:
+            service = self._ensure_service()
+            self.model_move_service = ModelMoveService(service.scanner, service.model_type)
+        return self.model_move_service
+
+    def _ensure_lifecycle_service(self) -> ModelLifecycleService:
+        if self.model_lifecycle_service is None:
+            service = self._ensure_service()
+            self.model_lifecycle_service = ModelLifecycleService(
+                scanner=service.scanner,
+                metadata_manager=MetadataManager,
+                metadata_loader=self._metadata_sync_service.load_local_metadata,
+                recipe_scanner_factory=ServiceRegistry.get_recipe_scanner,
+            )
+        return self.model_lifecycle_service
+
+    def _make_handler_proxy(self, name: str) -> Callable[[web.Request], web.StreamResponse]:
+        async def proxy(request: web.Request) -> web.StreamResponse:
+            try:
+                handler = self.get_handler(name)
+            except RuntimeError:
+                return web.json_response({"success": False, "error": "Service not ready"}, status=503)
+            return await handler(request)
+
+        return proxy
+
+    def _ensure_model_update_service(self) -> "ModelUpdateService":
+        if self._model_update_service is None:
+            raise RuntimeError("Model update service has not been attached")
+        return self._model_update_service

py/routes/base_recipe_routes.py

@@ -0,0 +1,200 @@
+"""Base infrastructure shared across recipe routes."""
+from __future__ import annotations
+
+import logging
+import os
+from typing import Callable, Mapping
+
+import jinja2
+from aiohttp import web
+
+from ..config import config
+from ..recipes import RecipeParserFactory
+from ..services.downloader import get_downloader
+from ..services.recipes import (
+    RecipeAnalysisService,
+    RecipePersistenceService,
+    RecipeSharingService,
+)
+from ..services.server_i18n import server_i18n
+from ..services.service_registry import ServiceRegistry
+from ..services.settings_manager import get_settings_manager
+from ..utils.constants import CARD_PREVIEW_WIDTH
+from ..utils.exif_utils import ExifUtils
+from .handlers.recipe_handlers import (
+    RecipeAnalysisHandler,
+    RecipeHandlerSet,
+    RecipeListingHandler,
+    RecipeManagementHandler,
+    RecipePageView,
+    RecipeQueryHandler,
+    RecipeSharingHandler,
+)
+from .recipe_route_registrar import ROUTE_DEFINITIONS
+
+logger = logging.getLogger(__name__)
+
+
+class BaseRecipeRoutes:
+    """Common dependency and startup wiring for recipe routes."""
+
+    _HANDLER_NAMES: tuple[str, ...] = tuple(
+        definition.handler_name for definition in ROUTE_DEFINITIONS
+    )
+
+    template_name: str = "recipes.html"
+
+    def __init__(self) -> None:
+        self.recipe_scanner = None
+        self.lora_scanner = None
+        self.civitai_client = None
+        self.settings = get_settings_manager()
+        self.server_i18n = server_i18n
+        self.template_env = jinja2.Environment(
+            loader=jinja2.FileSystemLoader(config.templates_path),
+            autoescape=True,
+        )
+
+        self._i18n_registered = False
+        self._startup_hooks_registered = False
+        self._handler_set: RecipeHandlerSet | None = None
+        self._handler_mapping: dict[str, Callable] | None = None
+
+    async def attach_dependencies(self, app: web.Application | None = None) -> None:
+        """Resolve shared services from the registry."""
+
+        await self._ensure_services()
+        self._ensure_i18n_filter()
+
+    async def ensure_dependencies_ready(self) -> None:
+        """Ensure dependencies are available for request handlers."""
+
+        if self.recipe_scanner is None or self.civitai_client is None:
+            await self.attach_dependencies()
+
+    def register_startup_hooks(self, app: web.Application) -> None:
+        """Register startup hooks once for dependency wiring."""
+
+        if self._startup_hooks_registered:
+            return
+
+        app.on_startup.append(self.attach_dependencies)
+        self._startup_hooks_registered = True
+
+    def to_route_mapping(self) -> Mapping[str, Callable]:
+        """Return a mapping of handler name to coroutine for registrar binding."""
+
+        if self._handler_mapping is None:
+            handler_set = self._create_handler_set()
+            self._handler_set = handler_set
+            self._handler_mapping = handler_set.to_route_mapping()
+        return self._handler_mapping
+
+    # Internal helpers -------------------------------------------------
+
+    async def _ensure_services(self) -> None:
+        if self.recipe_scanner is None:
+            self.recipe_scanner = await ServiceRegistry.get_recipe_scanner()
+            self.lora_scanner = getattr(self.recipe_scanner, "_lora_scanner", None)
+
+        if self.civitai_client is None:
+            self.civitai_client = await ServiceRegistry.get_civitai_client()
+
+    def _ensure_i18n_filter(self) -> None:
+        if not self._i18n_registered:
+            self.template_env.filters["t"] = self.server_i18n.create_template_filter()
+            self._i18n_registered = True
+
+    def get_handler_owner(self):
+        """Return the object supplying bound handler coroutines."""
+
+        if self._handler_set is None:
+            self._handler_set = self._create_handler_set()
+        return self._handler_set
+
+    def _create_handler_set(self) -> RecipeHandlerSet:
+        recipe_scanner_getter = lambda: self.recipe_scanner
+        civitai_client_getter = lambda: self.civitai_client
+
+        standalone_mode = os.environ.get("LORA_MANAGER_STANDALONE", "0") == "1" or os.environ.get("HF_HUB_DISABLE_TELEMETRY", "0") == "0"
+        if not standalone_mode:
+            from ..metadata_collector import get_metadata  # type: ignore[import-not-found]
+            from ..metadata_collector.metadata_processor import (  # type: ignore[import-not-found]
+                MetadataProcessor,
+            )
+            from ..metadata_collector.metadata_registry import (  # type: ignore[import-not-found]
+                MetadataRegistry,
+            )
+        else:  # pragma: no cover - optional dependency path
+            get_metadata = None  # type: ignore[assignment]
+            MetadataProcessor = None  # type: ignore[assignment]
+            MetadataRegistry = None  # type: ignore[assignment]
+
+        analysis_service = RecipeAnalysisService(
+            exif_utils=ExifUtils,
+            recipe_parser_factory=RecipeParserFactory,
+            downloader_factory=get_downloader,
+            metadata_collector=get_metadata,
+            metadata_processor_cls=MetadataProcessor,
+            metadata_registry_cls=MetadataRegistry,
+            standalone_mode=standalone_mode,
+            logger=logger,
+        )
+        persistence_service = RecipePersistenceService(
+            exif_utils=ExifUtils,
+            card_preview_width=CARD_PREVIEW_WIDTH,
+            logger=logger,
+        )
+        sharing_service = RecipeSharingService(logger=logger)
+
+        page_view = RecipePageView(
+            ensure_dependencies_ready=self.ensure_dependencies_ready,
+            settings_service=self.settings,
+            server_i18n=self.server_i18n,
+            template_env=self.template_env,
+            template_name=self.template_name,
+            recipe_scanner_getter=recipe_scanner_getter,
+            logger=logger,
+        )
+        listing = RecipeListingHandler(
+            ensure_dependencies_ready=self.ensure_dependencies_ready,
+            recipe_scanner_getter=recipe_scanner_getter,
+            logger=logger,
+        )
+        query = RecipeQueryHandler(
+            ensure_dependencies_ready=self.ensure_dependencies_ready,
+            recipe_scanner_getter=recipe_scanner_getter,
+            format_recipe_file_url=listing.format_recipe_file_url,
+            logger=logger,
+        )
+        management = RecipeManagementHandler(
+            ensure_dependencies_ready=self.ensure_dependencies_ready,
+            recipe_scanner_getter=recipe_scanner_getter,
+            logger=logger,
+            persistence_service=persistence_service,
+            analysis_service=analysis_service,
+            downloader_factory=get_downloader,
+            civitai_client_getter=civitai_client_getter,
+        )
+        analysis = RecipeAnalysisHandler(
+            ensure_dependencies_ready=self.ensure_dependencies_ready,
+            recipe_scanner_getter=recipe_scanner_getter,
+            civitai_client_getter=civitai_client_getter,
+            logger=logger,
+            analysis_service=analysis_service,
+        )
+        sharing = RecipeSharingHandler(
+            ensure_dependencies_ready=self.ensure_dependencies_ready,
+            recipe_scanner_getter=recipe_scanner_getter,
+            logger=logger,
+            sharing_service=sharing_service,
+        )
+
+        return RecipeHandlerSet(
+            page_view=page_view,
+            listing=listing,
+            query=query,
+            management=management,
+            analysis=analysis,
+            sharing=sharing,
+        )

py/routes/checkpoint_routes.py

@@ -0,0 +1,112 @@
+import logging
+from typing import Dict
+from aiohttp import web
+
+from .base_model_routes import BaseModelRoutes
+from .model_route_registrar import ModelRouteRegistrar
+from ..services.checkpoint_service import CheckpointService
+from ..services.service_registry import ServiceRegistry
+from ..config import config
+
+logger = logging.getLogger(__name__)
+
+class CheckpointRoutes(BaseModelRoutes):
+    """Checkpoint-specific route controller"""
+    
+    def __init__(self):
+        """Initialize Checkpoint routes with Checkpoint service"""
+        super().__init__()
+        self.template_name = "checkpoints.html"
+    
+    async def initialize_services(self):
+        """Initialize services from ServiceRegistry"""
+        checkpoint_scanner = await ServiceRegistry.get_checkpoint_scanner()
+        update_service = await ServiceRegistry.get_model_update_service()
+        self.service = CheckpointService(checkpoint_scanner, update_service=update_service)
+        self.set_model_update_service(update_service)
+
+        # Attach service dependencies
+        self.attach_service(self.service)
+    
+    def setup_routes(self, app: web.Application):
+        """Setup Checkpoint routes"""
+        # Schedule service initialization on app startup
+        app.on_startup.append(lambda _: self.initialize_services())
+        
+        # Setup common routes with 'checkpoints' prefix (includes page route)
+        super().setup_routes(app, 'checkpoints')
+    
+    def setup_specific_routes(self, registrar: ModelRouteRegistrar, prefix: str):
+        """Setup Checkpoint-specific routes"""
+        # Checkpoint info by name
+        registrar.add_prefixed_route('GET', '/api/lm/{prefix}/info/{name}', prefix, self.get_checkpoint_info)
+
+        # Checkpoint roots and Unet roots
+        registrar.add_prefixed_route('GET', '/api/lm/{prefix}/checkpoints_roots', prefix, self.get_checkpoints_roots)
+        registrar.add_prefixed_route('GET', '/api/lm/{prefix}/unet_roots', prefix, self.get_unet_roots)
+    
+    def _validate_civitai_model_type(self, model_type: str) -> bool:
+        """Validate CivitAI model type for Checkpoint"""
+        return model_type.lower() == 'checkpoint'
+    
+    def _get_expected_model_types(self) -> str:
+        """Get expected model types string for error messages"""
+        return "Checkpoint"
+
+    def _parse_specific_params(self, request: web.Request) -> Dict:
+        """Parse Checkpoint-specific parameters"""
+        params: Dict = {}
+
+        if 'checkpoint_hash' in request.query:
+            params['hash_filters'] = {'single_hash': request.query['checkpoint_hash'].lower()}
+        elif 'checkpoint_hashes' in request.query:
+            params['hash_filters'] = {
+                'multiple_hashes': [h.lower() for h in request.query['checkpoint_hashes'].split(',')]
+            }
+
+        return params
+    
+    async def get_checkpoint_info(self, request: web.Request) -> web.Response:
+        """Get detailed information for a specific checkpoint by name"""
+        try:
+            name = request.match_info.get('name', '')
+            checkpoint_info = await self.service.get_model_info_by_name(name)
+            
+            if checkpoint_info:
+                return web.json_response(checkpoint_info)
+            else:
+                return web.json_response({"error": "Checkpoint not found"}, status=404)
+                
+        except Exception as e:
+            logger.error(f"Error in get_checkpoint_info: {e}", exc_info=True)
+            return web.json_response({"error": str(e)}, status=500)
+    
+    async def get_checkpoints_roots(self, request: web.Request) -> web.Response:
+        """Return the list of checkpoint roots from config"""
+        try:
+            roots = config.checkpoints_roots
+            return web.json_response({
+                "success": True,
+                "roots": roots
+            })
+        except Exception as e:
+            logger.error(f"Error getting checkpoint roots: {e}", exc_info=True)
+            return web.json_response({
+                "success": False,
+                "error": str(e)
+            }, status=500)
+
+    async def get_unet_roots(self, request: web.Request) -> web.Response:
+        """Return the list of unet roots from config"""
+        try:
+            roots = config.unet_roots
+            return web.json_response({
+                "success": True,
+                "roots": roots
+            })
+        except Exception as e:
+            logger.error(f"Error getting unet roots: {e}", exc_info=True)
+            return web.json_response({
+                "success": False,
+                "error": str(e)
+            }, status=500)

py/routes/embedding_routes.py

@@ -0,0 +1,63 @@
+import logging
+from aiohttp import web
+
+from .base_model_routes import BaseModelRoutes
+from .model_route_registrar import ModelRouteRegistrar
+from ..services.embedding_service import EmbeddingService
+from ..services.service_registry import ServiceRegistry
+
+logger = logging.getLogger(__name__)
+
+class EmbeddingRoutes(BaseModelRoutes):
+    """Embedding-specific route controller"""
+    
+    def __init__(self):
+        """Initialize Embedding routes with Embedding service"""
+        super().__init__()
+        self.template_name = "embeddings.html"
+    
+    async def initialize_services(self):
+        """Initialize services from ServiceRegistry"""
+        embedding_scanner = await ServiceRegistry.get_embedding_scanner()
+        update_service = await ServiceRegistry.get_model_update_service()
+        self.service = EmbeddingService(embedding_scanner, update_service=update_service)
+        self.set_model_update_service(update_service)
+
+        # Attach service dependencies
+        self.attach_service(self.service)
+    
+    def setup_routes(self, app: web.Application):
+        """Setup Embedding routes"""
+        # Schedule service initialization on app startup
+        app.on_startup.append(lambda _: self.initialize_services())
+        
+        # Setup common routes with 'embeddings' prefix (includes page route)
+        super().setup_routes(app, 'embeddings')
+    
+    def setup_specific_routes(self, registrar: ModelRouteRegistrar, prefix: str):
+        """Setup Embedding-specific routes"""
+        # Embedding info by name
+        registrar.add_prefixed_route('GET', '/api/lm/{prefix}/info/{name}', prefix, self.get_embedding_info)
+    
+    def _validate_civitai_model_type(self, model_type: str) -> bool:
+        """Validate CivitAI model type for Embedding"""
+        return model_type.lower() == 'textualinversion'
+    
+    def _get_expected_model_types(self) -> str:
+        """Get expected model types string for error messages"""
+        return "TextualInversion"
+    
+    async def get_embedding_info(self, request: web.Request) -> web.Response:
+        """Get detailed information for a specific embedding by name"""
+        try:
+            name = request.match_info.get('name', '')
+            embedding_info = await self.service.get_model_info_by_name(name)
+            
+            if embedding_info:
+                return web.json_response(embedding_info)
+            else:
+                return web.json_response({"error": "Embedding not found"}, status=404)
+                
+        except Exception as e:
+            logger.error(f"Error in get_embedding_info: {e}", exc_info=True)
+            return web.json_response({"error": str(e)}, status=500)

py/routes/example_images_route_registrar.py

@@ -0,0 +1,65 @@
+"""Route registrar for example image endpoints."""
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Callable, Iterable, Mapping
+
+from aiohttp import web
+
+
+@dataclass(frozen=True)
+class RouteDefinition:
+    """Declarative configuration for a HTTP route."""
+
+    method: str
+    path: str
+    handler_name: str
+
+
+ROUTE_DEFINITIONS: tuple[RouteDefinition, ...] = (
+    RouteDefinition("POST", "/api/lm/download-example-images", "download_example_images"),
+    RouteDefinition("POST", "/api/lm/import-example-images", "import_example_images"),
+    RouteDefinition("GET", "/api/lm/example-images-status", "get_example_images_status"),
+    RouteDefinition("POST", "/api/lm/pause-example-images", "pause_example_images"),
+    RouteDefinition("POST", "/api/lm/resume-example-images", "resume_example_images"),
+    RouteDefinition("POST", "/api/lm/stop-example-images", "stop_example_images"),
+    RouteDefinition("POST", "/api/lm/open-example-images-folder", "open_example_images_folder"),
+    RouteDefinition("GET", "/api/lm/example-image-files", "get_example_image_files"),
+    RouteDefinition("GET", "/api/lm/has-example-images", "has_example_images"),
+    RouteDefinition("POST", "/api/lm/delete-example-image", "delete_example_image"),
+    RouteDefinition("POST", "/api/lm/force-download-example-images", "force_download_example_images"),
+    RouteDefinition("POST", "/api/lm/cleanup-example-image-folders", "cleanup_example_image_folders"),
+    RouteDefinition("POST", "/api/lm/example-images/set-nsfw-level", "set_example_image_nsfw_level"),
+    RouteDefinition("POST", "/api/lm/check-example-images-needed", "check_example_images_needed"),
+)
+
+
+class ExampleImagesRouteRegistrar:
+    """Bind declarative example image routes to an aiohttp router."""
+
+    _METHOD_MAP = {
+        "GET": "add_get",
+        "POST": "add_post",
+        "PUT": "add_put",
+        "DELETE": "add_delete",
+    }
+
+    def __init__(self, app: web.Application) -> None:
+        self._app = app
+
+    def register_routes(
+        self,
+        handler_lookup: Mapping[str, Callable[[web.Request], object]],
+        *,
+        definitions: Iterable[RouteDefinition] = ROUTE_DEFINITIONS,
+    ) -> None:
+        """Register each route definition using the supplied handlers."""
+
+        for definition in definitions:
+            handler = handler_lookup[definition.handler_name]
+            self._bind_route(definition.method, definition.path, handler)
+
+    def _bind_route(self, method: str, path: str, handler: Callable[[web.Request], object]) -> None:
+        add_method_name = self._METHOD_MAP[method.upper()]
+        add_method = getattr(self._app.router, add_method_name)
+        add_method(path, handler)

py/routes/example_images_routes.py

@@ -0,0 +1,88 @@
+from __future__ import annotations
+
+import logging
+from typing import Callable, Mapping
+
+from aiohttp import web
+
+from .example_images_route_registrar import ExampleImagesRouteRegistrar
+from .handlers.example_images_handlers import (
+    ExampleImagesDownloadHandler,
+    ExampleImagesFileHandler,
+    ExampleImagesHandlerSet,
+    ExampleImagesManagementHandler,
+)
+from ..services.use_cases.example_images import (
+    DownloadExampleImagesUseCase,
+    ImportExampleImagesUseCase,
+)
+from ..utils.example_images_download_manager import (
+    DownloadManager,
+    get_default_download_manager,
+)
+from ..utils.example_images_file_manager import ExampleImagesFileManager
+from ..utils.example_images_processor import ExampleImagesProcessor
+from ..services.example_images_cleanup_service import ExampleImagesCleanupService
+
+logger = logging.getLogger(__name__)
+
+
+class ExampleImagesRoutes:
+    """Route controller for example image endpoints."""
+
+    def __init__(
+        self,
+        *,
+        ws_manager,
+        download_manager: DownloadManager | None = None,
+        processor=ExampleImagesProcessor,
+        file_manager=ExampleImagesFileManager,
+        cleanup_service: ExampleImagesCleanupService | None = None,
+    ) -> None:
+        if ws_manager is None:
+            raise ValueError("ws_manager is required")
+        self._download_manager = download_manager or get_default_download_manager(ws_manager)
+        self._processor = processor
+        self._file_manager = file_manager
+        self._cleanup_service = cleanup_service or ExampleImagesCleanupService()
+        self._handler_set: ExampleImagesHandlerSet | None = None
+        self._handler_mapping: Mapping[str, Callable[[web.Request], web.StreamResponse]] | None = None
+
+    @classmethod
+    def setup_routes(cls, app: web.Application, *, ws_manager) -> None:
+        """Register routes on the given aiohttp application using default wiring."""
+
+        controller = cls(ws_manager=ws_manager)
+        controller.register(app)
+
+    def register(self, app: web.Application) -> None:
+        """Bind the controller's handlers to the aiohttp router."""
+
+        registrar = ExampleImagesRouteRegistrar(app)
+        registrar.register_routes(self.to_route_mapping())
+
+    def to_route_mapping(self) -> Mapping[str, Callable[[web.Request], web.StreamResponse]]:
+        """Return the registrar-compatible mapping of handler names to callables."""
+
+        if self._handler_mapping is None:
+            handler_set = self._build_handler_set()
+            self._handler_set = handler_set
+            self._handler_mapping = handler_set.to_route_mapping()
+        return self._handler_mapping
+
+    def _build_handler_set(self) -> ExampleImagesHandlerSet:
+        logger.debug("Building ExampleImagesHandlerSet with %s, %s, %s", self._download_manager, self._processor, self._file_manager)
+        download_use_case = DownloadExampleImagesUseCase(download_manager=self._download_manager)
+        download_handler = ExampleImagesDownloadHandler(download_use_case, self._download_manager)
+        import_use_case = ImportExampleImagesUseCase(processor=self._processor)
+        management_handler = ExampleImagesManagementHandler(
+            import_use_case,
+            self._processor,
+            self._cleanup_service,
+        )
+        file_handler = ExampleImagesFileHandler(self._file_manager)
+        return ExampleImagesHandlerSet(
+            download=download_handler,
+            management=management_handler,
+            files=file_handler,
+        )

py/routes/handlers/example_images_handlers.py

@@ -0,0 +1,185 @@
+"""Handler set for example image routes."""
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Callable, Mapping
+
+from aiohttp import web
+
+from ...services.use_cases.example_images import (
+    DownloadExampleImagesConfigurationError,
+    DownloadExampleImagesInProgressError,
+    DownloadExampleImagesUseCase,
+    ImportExampleImagesUseCase,
+    ImportExampleImagesValidationError,
+)
+from ...utils.example_images_download_manager import (
+    DownloadConfigurationError,
+    DownloadInProgressError,
+    DownloadNotRunningError,
+    ExampleImagesDownloadError,
+)
+from ...utils.example_images_processor import ExampleImagesImportError
+
+
+class ExampleImagesDownloadHandler:
+    """HTTP adapters for download-related example image endpoints."""
+
+    def __init__(
+        self,
+        download_use_case: DownloadExampleImagesUseCase,
+        download_manager,
+    ) -> None:
+        self._download_use_case = download_use_case
+        self._download_manager = download_manager
+
+    async def download_example_images(self, request: web.Request) -> web.StreamResponse:
+        try:
+            payload = await request.json()
+            result = await self._download_use_case.execute(payload)
+            return web.json_response(result)
+        except DownloadExampleImagesInProgressError as exc:
+            response = {
+                'success': False,
+                'error': str(exc),
+                'status': exc.progress,
+            }
+            return web.json_response(response, status=400)
+        except DownloadExampleImagesConfigurationError as exc:
+            return web.json_response({'success': False, 'error': str(exc)}, status=400)
+        except ExampleImagesDownloadError as exc:
+            return web.json_response({'success': False, 'error': str(exc)}, status=500)
+
+    async def get_example_images_status(self, request: web.Request) -> web.StreamResponse:
+        result = await self._download_manager.get_status(request)
+        return web.json_response(result)
+
+    async def pause_example_images(self, request: web.Request) -> web.StreamResponse:
+        try:
+            result = await self._download_manager.pause_download(request)
+            return web.json_response(result)
+        except DownloadNotRunningError as exc:
+            return web.json_response({'success': False, 'error': str(exc)}, status=400)
+
+    async def resume_example_images(self, request: web.Request) -> web.StreamResponse:
+        try:
+            result = await self._download_manager.resume_download(request)
+            return web.json_response(result)
+        except DownloadNotRunningError as exc:
+            return web.json_response({'success': False, 'error': str(exc)}, status=400)
+
+    async def stop_example_images(self, request: web.Request) -> web.StreamResponse:
+        try:
+            result = await self._download_manager.stop_download(request)
+            return web.json_response(result)
+        except DownloadNotRunningError as exc:
+            return web.json_response({'success': False, 'error': str(exc)}, status=400)
+
+    async def force_download_example_images(self, request: web.Request) -> web.StreamResponse:
+        try:
+            payload = await request.json()
+            result = await self._download_manager.start_force_download(payload)
+            return web.json_response(result)
+        except DownloadInProgressError as exc:
+            response = {
+                'success': False,
+                'error': str(exc),
+                'status': exc.progress_snapshot,
+            }
+            return web.json_response(response, status=400)
+        except DownloadConfigurationError as exc:
+            return web.json_response({'success': False, 'error': str(exc)}, status=400)
+        except ExampleImagesDownloadError as exc:
+            return web.json_response({'success': False, 'error': str(exc)}, status=500)
+
+    async def check_example_images_needed(self, request: web.Request) -> web.StreamResponse:
+        """Lightweight check to see if any models need example images downloaded."""
+        try:
+            payload = await request.json()
+            model_types = payload.get('model_types', ['lora', 'checkpoint', 'embedding'])
+            result = await self._download_manager.check_pending_models(model_types)
+            return web.json_response(result)
+        except Exception as exc:
+            return web.json_response(
+                {'success': False, 'error': str(exc)},
+                status=500
+            )
+
+
+class ExampleImagesManagementHandler:
+    """HTTP adapters for import/delete endpoints."""
+
+    def __init__(self, import_use_case: ImportExampleImagesUseCase, processor, cleanup_service) -> None:
+        self._import_use_case = import_use_case
+        self._processor = processor
+        self._cleanup_service = cleanup_service
+
+    async def import_example_images(self, request: web.Request) -> web.StreamResponse:
+        try:
+            result = await self._import_use_case.execute(request)
+            return web.json_response(result)
+        except ImportExampleImagesValidationError as exc:
+            return web.json_response({'success': False, 'error': str(exc)}, status=400)
+        except ExampleImagesImportError as exc:
+            return web.json_response({'success': False, 'error': str(exc)}, status=500)
+
+    async def delete_example_image(self, request: web.Request) -> web.StreamResponse:
+        return await self._processor.delete_custom_image(request)
+
+    async def set_example_image_nsfw_level(self, request: web.Request) -> web.StreamResponse:
+        return await self._processor.set_example_image_nsfw_level(request)
+
+    async def cleanup_example_image_folders(self, request: web.Request) -> web.StreamResponse:
+        result = await self._cleanup_service.cleanup_example_image_folders()
+
+        if result.get('success') or result.get('partial_success'):
+            return web.json_response(result, status=200)
+
+        error_code = result.get('error_code')
+        status = 400 if error_code in {'path_not_configured', 'path_not_found'} else 500
+        return web.json_response(result, status=status)
+
+
+class ExampleImagesFileHandler:
+    """HTTP adapters for filesystem-centric endpoints."""
+
+    def __init__(self, file_manager) -> None:
+        self._file_manager = file_manager
+
+    async def open_example_images_folder(self, request: web.Request) -> web.StreamResponse:
+        return await self._file_manager.open_folder(request)
+
+    async def get_example_image_files(self, request: web.Request) -> web.StreamResponse:
+        return await self._file_manager.get_files(request)
+
+    async def has_example_images(self, request: web.Request) -> web.StreamResponse:
+        return await self._file_manager.has_images(request)
+
+
+@dataclass(frozen=True)
+class ExampleImagesHandlerSet:
+    """Aggregate of handlers exposed to the registrar."""
+
+    download: ExampleImagesDownloadHandler
+    management: ExampleImagesManagementHandler
+    files: ExampleImagesFileHandler
+
+    def to_route_mapping(self) -> Mapping[str, Callable[[web.Request], web.StreamResponse]]:
+        """Flatten handler methods into the registrar mapping."""
+
+        return {
+            "download_example_images": self.download.download_example_images,
+            "get_example_images_status": self.download.get_example_images_status,
+            "pause_example_images": self.download.pause_example_images,
+            "resume_example_images": self.download.resume_example_images,
+            "stop_example_images": self.download.stop_example_images,
+            "force_download_example_images": self.download.force_download_example_images,
+            "check_example_images_needed": self.download.check_example_images_needed,
+            "import_example_images": self.management.import_example_images,
+            "delete_example_image": self.management.delete_example_image,
+            "set_example_image_nsfw_level": self.management.set_example_image_nsfw_level,
+            "cleanup_example_image_folders": self.management.cleanup_example_image_folders,
+            "open_example_images_folder": self.files.open_example_images_folder,
+            "get_example_image_files": self.files.get_example_image_files,
+            "has_example_images": self.files.has_example_images,
+        }