Previously, `map_path_to_link` and `map_link_to_path` returned the original input path when no mapping was found, instead of the normalized version. This could cause inconsistencies when paths with different representations (e.g., trailing slashes) were used. Now both methods consistently return the normalized path, ensuring uniform path handling throughout the application.
- Use pointer events (pointerdown/pointermove/pointerup/pointercancel) with proper capture
- Fix drag not updating strength values by avoiding re-renders during drag
- Fix cursor stuck in resize state by ensuring proper cleanup
- Fix cursor showing wrong icon on hover (should be pointer)
- Ensure strength values display fixed width with 2 decimal places
- Remove unnecessary data-capture-wheel attribute (no wheel adjustment in loras widget)
- Add font-variant-numeric: tabular-nums for consistent number display
This ensures loras widget works consistently in both Canvas and Vue DOM render modes.
In Vue DOM render mode, widget.inputEl is not in the DOM, causing autocomplete to fail. This commit:
- Adds findWidgetInputElement() helper to search DOM for actual input elements
- Checks if widget.inputEl is in document before using it
- Falls back to DOM search for Vue-rendered widgets using .lg-node-widget containers
- Implements async initialization with retry logic (20 attempts, 50ms interval)
- Adds debug logging for troubleshooting
- Prevents duplicate initialization with isInitializing flag
Fixes autocomplete functionality for Lora Loader nodes when ComfyUI uses Vue DOM rendering instead of canvas rendering.
- Pass `version_info` parameter through download manager to model update service
- Enhance `_create_record` to use version info when creating records for missing versions
- Add `_extract_single_version` helper method for consistent version extraction
- Improve handling of version metadata during library synchronization
- Initialize logging configuration via `setup_logging()` when not in standalone mode
- Detect standalone mode using environment variables `LORA_MANAGER_STANDALONE` and `HF_HUB_DISABLE_TELEMETRY`
- Remove redundant `STANDALONE_MODE` variable that previously checked `sys.modules`
Changed logging level from INFO to DEBUG for performance-related messages in model management service. This reduces noise in production logs while maintaining debugging capability for performance analysis.
- Add `open_settings_location` method to `FileSystemHandler` to open OS file explorer at settings file location
- Register new POST route `/api/lm/settings/open-location` for settings file access
- Inject `SettingsManager` dependency into `FileSystemHandler` constructor
- Add cross-platform support for Windows, macOS, and Linux file explorers
- Include error handling for missing settings files and system exceptions
- Return cache entry data from model move operations for immediate UI updates
- Add recalculate_type parameter to update_single_model_cache for proper type adjustment
- Propagate cache entry through API layer to frontend MoveManager
- Enable virtual scroller to update moved items with new cache data
- Add pre-processing step to populate missing parameters for candidate samplers, especially for SamplerCustomAdvanced requiring tracing
- Change sampler selection from most recent (closest to downstream) to first in execution order to prioritize base samplers over refine samplers
- Improve parameter handling by updating sampler parameters with traced values before ranking
- Maintain backward compatibility with fallback to first sampler if no criteria match
- Add support for `basic_pipe` nodes in metadata processor to handle pipeline nodes like FromBasicPipe
- Optimize `find_primary_checkpoint` by accepting optional `primary_sampler_id` to avoid redundant calculations
- Update `get_workflow_trace` to pass known primary sampler ID for improved efficiency
Removed the forced normalization of path separators to forward slashes in BaseModelService to maintain platform-specific separators. Updated test cases to use os.sep for constructing expected paths, ensuring tests work correctly across different operating systems while preserving native path representations.
Add `_seed_root_symlink_mappings` method to ensure symlinked root folders are recorded before deep scanning, preventing them from being missed during directory traversal. This ensures that root symlinks are properly captured in the path mappings.
Additionally, normalize separators in relative paths for cross-platform consistency in `BaseModelService`, and update tests to verify root symlinks are preserved in the cache.
Add new POST endpoint `/api/lm/example-images/set-nsfw-level` to allow updating NSFW classification for individual example images. The endpoint supports both regular and custom images, validates required parameters, and updates the corresponding model metadata. This enables users to manually adjust NSFW ratings for better content filtering.
Add new model constants for Flux.2 D and ZImageTurbo to the BASE_MODELS object,
along with their corresponding abbreviations in BASE_MODEL_ABBREVIATIONS. Also
include these new models in the appropriate categories within BASE_MODEL_CATEGORIES.
This update ensures the application can properly recognize and handle these
newly supported AI models in the system.
Add KJNodesModelLoaderExtractor to handle metadata extraction from KJNodes loaders that expose model_name. This supports GGUFLoaderKJ and DiffusionModelLoaderKJ nodes, ensuring consistent checkpoint metadata collection across different node types.
Remove legacy HTML scraping implementation of get_model_by_url method
and associated BeautifulSoup dependency. The functionality has been
replaced by API-based implementation in get_model_version method.
This simplifies the codebase and removes the optional bs4 dependency,
making the client more maintainable and reliable.
Add GGUFLoaderExtractor class to handle metadata extraction for GGUF model loaders. Register extractor for both LoaderGGUF and LoaderGGUFAdvanced node types to capture checkpoint metadata from gguf_name input parameter. This enables proper metadata tracking for GGUF model files used in the system.
Update the plugin cleanup process to preserve the model_cache folder
along with settings.json and civitai. This prevents accidental deletion
of cached model files during plugin updates, improving performance by
avoiding unnecessary model re-downloads.
- Implement core LoRA syntax manipulation functions including:
- applyLoraValuesToText for updating LoRA strengths and clip values
- normalizeStrengthValue for consistent numeric formatting
- shouldIncludeClipStrength for clip strength inclusion logic
- cleanupLoraSyntax for text normalization
- debounce utility for input handling
- Add comprehensive test suite covering all utility functions
- Include edge cases for clip strength handling, numeric formatting, and syntax cleanup
- Support both basic and expanded LoRA syntax formats with proper value preservation
- Enable debounced input synchronization for better performance
The utilities provide robust handling of LoRA syntax patterns while maintaining compatibility with existing ComfyUI workflows.
Changed the sync icon to an exclamation circle icon on the updates filter button to provide clearer visual indication of the filter's purpose and improve user interface consistency.
Refactor `_extract_size_bytes` method to prioritize primary model files when calculating size. The new implementation:
- Extracts size parsing into separate `parse_size` function
- Adds logic to prefer files marked as both "model" type and "primary"
- Falls back to first valid size if no primary model file found
- Adds comprehensive tests for primary preference and fallback behavior
This ensures more accurate size reporting for model files, particularly when multiple file types are present in the response.
Add a new filter option to show only models with available updates across all supported languages. This includes:
- Adding "updates" filter translations in all locale files (de, en, es, fr, he, ja, ko)
- Extending BaseModelApiClient to handle update_available_only query parameter
- Implementing update filter button in PageControls component with event listeners
- Adding corresponding CSS styles for active filter state
The feature allows users to quickly identify and focus on models that have updates available, improving the update management workflow.
Replace the native browser confirmation dialog with a custom modal when deleting model versions. This provides better UX with consistent styling, allows displaying version information (name, preview, metadata), and gives users more context before confirming deletion.
Key changes:
- Added modalManager import
- Created showDeleteVersionModal function to display deletion confirmation modal
- Enhanced performDeleteVersion function with better error handling and button state management
- Modal shows version preview, name, base model, and metadata
- Improved accessibility with proper modal interactions
- Maintains existing deletion functionality with enhanced user experience
- Replace inline styles with CSS classes for better maintainability
- Update class names to use consistent 'lm-' prefix across components
- Add comprehensive CSS stylesheet with tooltip system and responsive layouts
- Improve accessibility with proper focus states and keyboard navigation
- Implement hover and active state transitions for enhanced UX
- Refactor expand button to use CSS classes instead of inline styles
- Update test files to reflect new class naming convention
- Rename ui_utils.js to top_menu_extension.js to better reflect functionality
- Replace custom button creation with ComfyUI Button and ButtonGroup components
- Implement proper top menu integration using ComfyUI's menu system
- Simplify window opening logic with shift-click support for new windows
- Add retry mechanism for attaching button to menu
- Improve code organization and maintainability by leveraging existing UI components
Refactor focus navigation logic in LoRA widget to separate focus queueing from execution. Added helper functions for finding focus entries and managing focus queue. This improves code organization and prevents focus issues when tabbing between strength and clip inputs.
Key changes:
- Extract focus navigation logic into reusable functions
- Separate focus queueing from focus execution
- Maintain same keyboard navigation behavior while improving code structure
- Fix potential focus loss when tabbing between inputs
The refactoring makes the focus navigation code more maintainable and reduces duplication while preserving the existing tab navigation functionality.
Add pendingFocusTarget state to track which LoRA strength input is being interacted with. This ensures proper focus behavior when clicking on strength inputs, particularly when the widget is being re-rendered. The focus is now properly restored to the correct input after UI updates.
Add support for {model_name} and {version_name} placeholders in download path templates. These new placeholders allow for more flexible and descriptive file organization by including the actual model name and version name in the download directory structure.
Changes include:
- Updated download_manager.py and utils.py to handle new placeholders
- Added placeholders to constants.js for UI reference
- Updated settings modal template to show available placeholders
- Added comprehensive tests to verify placeholder functionality
This enhancement provides users with more control over how downloaded models are organized on their file system.
Add initialization tracking to prevent multiple event listener attachments
in context menu components. Use dataset.initialized flag to ensure NSFW
selector events are only set up once per component instance.
In ModelContextMenuMixin, replace DOM elements and reattach event listeners
to avoid duplicates when components are reinitialized. This fixes issues
where multiple click handlers could be attached to the same elements.
- Pass ModelUpdateService to CheckpointService, EmbeddingService, and LoraService constructors
- Add has_update query parameter filter to model listing handler
- Update BaseModelService to accept optional update_service parameter
These changes enable model update functionality across different model types and provide filtering capability for models with available updates.
- Add new model name display setting with options to show model name or file name
- Implement helper function to determine display name based on user preference
- Update model card footer to use dynamic display name
- Include model name display setting in settings modal and state management
- Remove redundant labels from display density descriptions in multiple locales
- Simplify card info display descriptions by removing duplicate text
The changes provide cleaner UI text and add flexibility for users to choose between displaying model names or file names in card footers.
- Add RateLimitError import and exception handling in API methods
- Create _make_request wrapper to surface rate limit errors from downloader
- Add test case to verify rate limit error propagation
- Set default provider as "civarchive_api" for rate limit errors
This ensures rate limit errors are properly propagated and handled throughout the CivArchive client, improving error reporting and allowing callers to implement appropriate retry logic.
- Add RateLimitError import and _make_request wrapper method to handle rate limiting
- Update API methods to use _make_request wrapper instead of direct downloader calls
- Add explicit RateLimitError handling in API methods to properly propagate rate limit errors
- Add _extract_retry_after method to parse Retry-After headers
- Improve error handling by surfacing rate limit information to callers
These changes ensure that rate limiting from the Civitai API is properly detected and handled, allowing callers to implement appropriate backoff strategies when rate limits are encountered.
Change bulk action labels from "All" to "Selected" in both English and Chinese locales to accurately reflect that these actions apply only to selected items rather than all items. This improves user interface clarity and prevents potential confusion about the scope of bulk operations.
- Remove 'clear' action from context menu
- Reorder menu items to prioritize common operations
- Move destructive operations (delete, move) to bottom section
- Add visual separation between action groups
- Maintain all existing functionality with improved organization
- Replace setdefault with get and explicit dict initialization in MetadataUpdater
- Change civitai field type from Optional[Dict] to Dict[str, Any] with default_factory
- Add None check and dict initialization in BaseModelMetadata.__post_init__
- Ensures civitai data is always a dictionary, preventing type errors and improving code reliability
Remove direct imports of IO type constants from comfy.comfy_types and replace them with string literals "STRING" in input type definitions and return types. This improves code portability and reduces dependency on external type definitions.
Changes made across multiple files:
- Remove `from comfy.comfy_types import IO` imports
- Replace `IO.STRING` with "STRING" in INPUT_TYPES and RETURN_TYPES
- Move CLIPTextEncode import to function scope in prompt.py for better dependency management
This refactor maintains the same functionality while making the code more self-contained and reducing external dependencies.
- Add metadata_source field to track origin of model metadata
- Define MODEL_COLUMNS constants for consistent column management
- Refactor SQL queries to use dynamic column selection
- Improve Civitai data detection to include creator_username and trained_words
- Update database operations to handle new metadata field and tag management
Refactored `createModelApiClient` to pass the specific model type as a parameter to each client constructor. Introduced `isValidModelType` for validation and added logic to set a default model type if provided type is invalid or not specified. Updated `getModelApiClient` function to utilize these improvements, ensuring robust model API instantiation.
Addressed the aiohttp warnings by aligning the test scaffolding with current best practices. Added an AppKey constant and stored the route tracking list through it to satisfy aiohttp’s NotAppKeyWarning expectations. Swapped the websocket lambdas for async no-op handlers so the registered routes now point to coroutine callables, clearing the deprecation warning about bare functions.
Updated the `ModelScanner` class to extract and format the creator username from Civitai data. This enhancement ensures that the creator information is properly included in slim model data.
This commit implements drag-and-drop functionality for sidebar nodes,
adding visual feedback via highlight styling when dragging over
valid drop targets. The CSS introduces new classes to style
drop indicators using the lora-accent color scheme, while the JS
adds event handlers to manage drag operations and update the UI
state accordingly. This improves user interaction by providing
clear visual cues for valid drop areas during file operations.
- Pass modelType parameter to setupTagEditMode function
- Implement model type aware priority tag suggestions
- Add model type normalization and resolution logic
- Handle suggestion state reset when model type changes
- Maintain backward compatibility with existing functionality
The changes enable context-aware tag suggestions based on model type, improving tag relevance and user experience when editing tags for different model types.
- Rename `.settings-open-location-button` to `.settings-action-link` for better semantic meaning
- Add enhanced hover/focus states with accent colors and border transitions
- Implement tabbed interface for priority tags with LoRA, checkpoint, and embedding sections
- Improve input styling with consistent error states and example code formatting
- Remove deprecated grid layout in favor of tab-based organization
- Add responsive tab navigation with proper focus management and visual feedback
Refactor metadata sync service to use a prioritized provider fallback system when handling deleted CivitAI models. The new approach:
1. Attempts civarchive_api provider first for deleted models
2. Falls back to sqlite provider if archive DB is enabled
3. Maintains existing default provider behavior for non-deleted models
4. Tracks provider attempts and errors for better debugging
This improves reliability when fetching metadata for deleted models by trying multiple sources before giving up, and provides clearer error messages based on which providers were attempted.
- Fix CivArchive model ID redirect logic to only follow redirects when context points to original model
- Rename CivitaiModelMetadataProvider to CivArchiveModelMetadataProvider for consistency
- Reorder fallback metadata providers to prioritize Civitai API over CivArchive API for better metadata quality
- Remove unused asyncio import and redundant logging from metadata sync service
Introduce `_request_json` method for async JSON requests and improved error handling. Add static methods `_normalize_payload`, `_split_context`, `_ensure_list`, and `_build_model_info` to parse and normalize API responses. These changes improve the robustness of the CivArchiveClient by ensuring consistent data structures and handling potential API response issues gracefully.
Will Miao