- Add execution_seed and next_seed parameters to support deterministic randomization across batch executions
- Separate UI display generation from execution stack generation to maintain consistency in batch queues
- Update LoraService to accept optional seed parameter for reproducible randomization
- Ensure each execution with a different seed produces unique results without affecting global random state
Introduce a new RecipeFTSIndex class that provides fast prefix-based search across recipe fields (title, tags, LoRA names/models, prompts) using SQLite's FTS5 extension. The implementation supports sub-100ms search times for large datasets (20k+ recipes) and includes asynchronous indexing, incremental updates, and comprehensive unit tests.
CivitAI does not distinguish between checkpoint and diffusion model types -
both are labeled as "checkpoint". For certain base model types like
"ZImageTurbo", all models are actually diffusion models and should be
saved to the unet/diffusion model folder instead of the checkpoint folder.
- Add DIFFUSION_MODEL_BASE_MODELS constant for known diffusion model types
- Add default_unet_root setting with auto-set logic
- Route downloads to unet folder when baseModel matches known diffusion types
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When users paste CivArchive URLs, the system now fetches metadata from
CivArchive API first instead of Civitai. This prevents download failures
when a model has been deleted from Civitai but remains available on
CivArchive with alternative mirrors.
Changes:
- Source-aware metadata fetching: Uses CivArchive API when source='civarchive'
- URL prioritization: Prefers non-Civitai mirrors for CivArchive downloads
- Fallback mechanism: Falls back to default provider if CivArchive fails
Fixes#769
- Add folder_include parameter support in backend API handlers
- Add folder_include to FilterCriteria and implement multi-folder filtering logic
- Update frontend to send all include folders instead of only the first
- Add tests for single/multiple include folders, include with exclude, and non-recursive filtering
Add support for respecting recommended strength values from LoRA usage_tips
when randomizing LoRA selection.
Features:
- New toggle setting to enable/disable recommended strength respect (default off)
- Scale range slider (0-2, default 0.5-1.0) to adjust recommended values
- Uses recommended strength × random(scale) when feature enabled
- Fallbacks to original Model/Clip Strength range when no recommendation exists
- Clip strength recommendations only apply when using Custom Range mode
Backend changes:
- Parse usage_tips JSON string to extract strength/clipStrength
- Apply scale factor to recommended values during randomization
- Pass new parameters through API route and node
Frontend changes:
- Update RandomizerConfig type with new properties
- Add new UI section with toggle and dual-range slider
- Wire up state management and event handlers
- No layout shift (removed description text)
Tests:
- Add tests for enabled/disabled recommended strength in API routes
- Add test verifying config passed to service
- All existing tests pass
Build: Include compiled Vue widgets
- Add blank line after module docstring for better PEP 8 compliance
- Reformat long lines to adhere to 88-character limit using Black-style formatting
- Improve string consistency by using double quotes consistently
- Enhance readability of complex list comprehensions and method calls
- Maintain all existing functionality while improving code structure
- Add `_preprocess_loras_input` method to handle different widget input formats
- Move core randomization logic to `LoraService` for better separation of concerns
- Update `_select_loras` method to use new service-based approach
- Add comprehensive test fixtures for license filtering scenarios
- Include debug print statement for pool config inspection during development
This refactor improves code organization by centralizing business logic in the service layer while maintaining backward compatibility with existing widget inputs.
- Document dual UI systems: standalone web UI and ComfyUI custom node widgets
- Add ComfyUI widget development guidelines including styling and constraints
- Update terminology in LoraRandomizerNode from 'frontend/backend' to 'fixed/always' for clarity
- Include UI constraints for ComfyUI widgets: minimize vertical space, avoid dynamic height changes, keep UI simple
- Implement LoRA locking to prevent specific LoRAs from being changed during randomization
- Add visual styling for locked state with amber accents and distinct backgrounds
- Introduce `roll_mode` configuration with 'backend' (execute current selection while generating new) and 'frontend' (execute newly generated selection) behaviors
- Move LoraPoolNode to 'Lora Manager/randomizer' category and remove standalone class mappings
- Standardize RETURN_NAMES in LoraRandomizerNode for consistency
- Import and register two new nodes: LoraDemoNode and LoraRandomizerNode
- Update import exception handling for better readability with multi-line formatting
- Add comprehensive documentation file `docs/custom-node-ui-output.md` for UI output usage in custom nodes
- Ensure proper node registration in NODE_CLASS_MAPPINGS for ComfyUI integration
- Maintain backward compatibility with existing node structure and import fallbacks
- Update `is_civitai_api_metadata` to exclude both "archive_db" and "civarchive" sources
- Skip Civitai metadata updates when existing metadata is higher quality than incoming archive data
- Add test to verify API metadata is preserved when CivArchive provides lower-quality data
Updated image path resolution logic to prioritize local sibling images in the same directory as recipes. When a stored image path differs from a local sibling, the system now automatically updates the recipe file to use the local path and persists this repair. This improves reliability when recipe assets are moved or reorganized, ensuring images remain accessible even if original paths become invalid.
- 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
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 new move_recipes_bulk endpoint to handle moving multiple recipes simultaneously. This improves efficiency when reorganizing recipe collections by allowing batch operations instead of individual moves.
- Add move_recipes_bulk handler method with proper error handling
- Register new POST /api/lm/recipes/move-bulk route
- Implement bulk move logic in persistence service
- Validate required parameters (recipe_ids and target_path)
- Handle common error cases including validation, not found, and server errors
- Add GET /api/lm/recipes/roots endpoint to retrieve recipe root directories
- Add POST /api/lm/recipe/move endpoint to move recipes between directories
- Register new endpoints in route definitions
- Implement error handling for both new endpoints with proper status codes
- Enable recipe management operations for better file organization
- Add new API endpoints for folder operations: get_folders, get_folder_tree, and get_unified_folder_tree
- Extend recipe listing handler to support folder and recursive filtering parameters
- Register new folder-related routes in route definitions
- Enable users to organize and browse recipes using folder structures
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 `_get_supported_extensions_for_type` method to return allowed extensions per model type
- Rename `_extract_safetensors_from_archive` to `_extract_model_files_from_archive` and extend to filter by allowed extensions
- Update error message to list supported extensions when archive contains no valid files
- Add test for extracting .pt embedding files from zip archives
- Refactor metadata detection to handle nested "meta" objects
- Add support for lowercase "lora:" hash keys
- Extract metadata from nested "meta" field when present
- Update tests to verify nested metadata parsing
- Handle case-insensitive LORA hash detection
The changes ensure proper parsing of Civitai image metadata that may be wrapped in nested structures, improving compatibility with different API response formats.
Add _normalize_checkpoint_entry method to handle legacy checkpoint data formats (strings, tuples) by converting them to dictionaries. This prevents errors during enrichment when checkpoint data is not in the expected dictionary format. Invalid checkpoint entries are now removed instead of causing processing failures.
- Update get_paginated_data and get_recipe_by_id methods to use normalization
- Add test cases for legacy string and tuple checkpoint formats
- Ensure backward compatibility with existing checkpoint handling
Add _normalize_checkpoint_entry method to handle legacy and malformed checkpoint data by:
- Converting string entries to structured dict format
- Handling single-element lists/tuples recursively
- Dropping invalid entries with appropriate warnings
- Maintaining backward compatibility while improving data consistency
Add test case to verify string checkpoint conversion works correctly.
- Add token parsing to support include/exclude search terms using "-" prefix
- Implement token-based matching logic for relative path searches
- Improve search result sorting by prioritizing prefix matches and match position
- Add frontend test for multi-token highlighting with exclusion support
