When batch-downloading different versions of the same model, dedup by
modelId alone discards the second URL. Use modelId:modelVersionId as
the dedup key so users can download, e.g., latest + a specific version.
- Add 'Node 2.0: Maximum visible LoRA entries' setting (default 12)
- Apply max-height to loras container in Vue mode to prevent unbounded growth
- Add enableListWheelScroll: window capture-phase wheel hook so scroll
inside the widget scrolls the list instead of zooming the canvas
The refresh_model_updates handler was calling record.has_update() with
default hide_early_access=False, causing the toast to report early-access
updates that the Updates filter (which uses the user's hide_early_access
setting) would then hide. This resulted in misleading "Found N updates"
toasts followed by an empty Updates view.
Now the handler reads hide_early_access_updates from settings and passes
it to has_update(), matching the behavior of _serialize_record and
_annotate_update_flags.
In Vue/Node 2.0 mode, the AutocompleteTextWidget's textarea wheel events were intercepted by TransformPane @wheel.capture before reaching the @wheel handler, causing canvas zoom instead of text scrolling.
- Add lm-wheel-scrollable class in Vue mode to hook into the window capture-phase handler (enableListWheelScroll) which scrolls the textarea manually before TransformPane can react.
- Add maxHeight prop and container max-height for Lora Loader/Stacker/WanVideo nodes (modelType === 'loras'), matching canvas mode's height cap. Prompt/Text nodes remain uncapped.
In Nodes 2.0 / Vue node mode the Lora Loader list could not be capped
and the node grew to show every row, unlike classic mode which fixes the
list area to 12 rows. The Vue layout engine measures the rendered DOM, so
CSS variables and computeLayoutSize alone were ignored.
- Physically cap the container via max-height so the rendered element is
bounded to the 12-row height; extra rows scroll (overflow: auto).
- Report the capped height through computeSize / computeLayoutSize /
getHeight / getMinHeight so the node background matches the list.
- Add enableListWheelScroll: a window capture-phase wheel hook that scrolls
the hovered list instead of letting ComfyUI zoom the canvas, which fires
on the document/canvas in capture and beat a container-level listener.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Both restore_suffixed_filenames.py and migrate_legacy_metadata.py
hardcoded Path.home() / '.config' / APP_NAME for finding settings.json,
which only works on Linux. On Windows this resolves to the wrong path
(~/.config/ instead of %LOCALAPPDATA%).
Replace the hand-rolled fallback with platformdirs.user_config_dir(),
which correctly resolves to the OS-appropriate config directory on all
platforms (Windows: %%LOCALAPPDATA%%, macOS: ~/Library/Application Support,
Linux: ~/.config). The portable mode check (settings.json in repo root
with use_portable_settings: true) is preserved unchanged.
Build a local_cache from the scanner cache before calling the metadata
parser. When a resource hash is found in the cache, populate the entry
directly from cached civitai metadata instead of calling CivitAI's
/model-versions/by-hash endpoint.
This eliminates redundant API calls and retries for the common case
where the example image only uses the parent model plus a checkpoint.
- refreshRecipes() now accepts fullRebuild param and passes it to scan endpoint
- Use consistent toast.api.refreshComplete / toast.api.refreshFailed keys
- Use loadingManager.show() with progress bar (matching models page style)
- Both Refresh and Rebuild Cache now hit the real /api/lm/recipes/scan endpoint
- Add sidebarManager.refresh() after recipe scan completes
- Backend scan_recipes handler reads full_rebuild query param
Bug: when scrolling down on recipes page, any operation with
preserveScroll: true would fetch only page 1 data then restore
scroll position to beyond the loaded items, leaving the grid empty.
Fix:
- Remove preserveScroll: true from all 7 must-refresh trigger
paths (filter, search, sort, import, settings reload, sync,
rebuild cache, sidebar folder nav)
- Replace full list refresh with updateSingleItem() for repair
and bulk missing-LoRA download operations
- Update tests to match new scroll-free behavior
- Apply CivitaiApiMetadataParser's base_model result to metadata in
_do_import_remote_recipe and _do_import_from_url (was previously discarded)
- Extract baseModel from raw civitai_info before populate_checkpoint_from_civitai
so it's not lost when the type check rejects non-checkpoint model versions
- Only format and save checkpoint entry when it has real data (modelId, versionId,
name, or version), preventing empty {'type': 'checkpoint'} stubs
- Add wildcards and backups to skip_files in all three ZIP upgrade
skip locations: _clean_plugin_folder, copy loop, .tracking generation
- Remove logs from skip_files (logs are transient and rotate automatically)
- Add _prune_old_logs() to session_logging.py: keeps only the 3 newest
session log files, deletes older ones on each standalone startup
On Windows, shutil.rmtree() fails when deleting a directory that contains
an open SQLite database file. The ZIP update path in _download_and_replace_zip()
calls _clean_plugin_folder() which tries to delete the cache/ directory,
but downloaded_versions.sqlite is held open by DownloadedVersionHistoryService.
Fix:
- Add close() method to DownloadedVersionHistoryService to release
the persistent SQLite connection
- Call close() before _clean_plugin_folder() in the ZIP update flow
- Add 'cache' to the skip_files list so the runtime cache directory is
never deleted during plugin updates
When certifi is available, pass its CA bundle path as --ca-certificate
to the aria2c subprocess so that aria2 downloads use the same
certificate store as Python aiohttp downloads. Graceful fallback when
certifi is not installed.
Adds a new bulk operation in the recipes page that allows users to select
multiple recipes and repair their metadata in batch.
Backend:
- New POST /api/lm/recipes/repair-bulk endpoint accepting recipe_ids array
- repair_recipes_bulk handler iterates repair_recipe_by_id for each recipe
- Response includes per-recipe updated data for frontend card refresh
Frontend:
- Bulk context menu: new 'Repair Metadata for Selected' item in Metadata section
- BulkManager.repairSelectedRecipes() with loading/toast flow
- Uses VirtualScroller.updateSingleItem() per repaired recipe (no full reload)
- Visibility controlled via repairMetadata actionConfig flag
Locales:
- Added repairMetadata, repairBulkComplete, repairBulkSkipped, repairBulkFailed
- Translated across all 9 supported languages
Add corruption detection to _repair_single_recipe: if checkpoint.modelVersionId matches any LoRA's modelVersionId, the checkpoint is corrupted (a LoRA was saved as checkpoint). Clear the checkpoint and remove the matching LoRA entry, then let enrichment re-resolve the correct checkpoint from CivitAI metadata.
This fixes the retroactive repair path for the modelVersionIds[0] fallback bug.
When importing a CivitAI image as a recipe, modelVersionIds[0] was blindly used as the checkpoint version ID. This array mixes checkpoints and LoRAs without ordering guarantees, causing LoRAs to be saved as the recipe checkpoint.
Fix by:
1. Removing the modelVersionIds[0] fallback in _download_remote_media
2. Parsing resources entries with type:"model" as the checkpoint
3. Adding model type validation in populate_checkpoint_from_civitai
Also add 2 tests for the new behavior and fix 3 tests whose mocks lacked the required model.type field.
Previously check_pending_models() only skipped models already in
processed_models, so models that had permanently failed (no CivitAI
images available, download errors) were forever reported as "pending".
This caused repeated auto-download cycles with no actual work to do.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
When CivitAI returns 404 (ResourceNotFoundError) and a fallback provider
like CivArchive subsequently rate-limits, the ChainedMetadataProvider
now suppresses the RateLimitError instead of propagating it. Previously,
the rate-limit error would bubble up through _refresh_single_model and
cause the outer retry loop to re-process the same model repeatedly,
producing dozens of duplicate "Model X is no longer available" log
messages and wasting API quota.
The model is NOT permanently marked as ignored — its last_checked_at
timestamp is preserved, so it will be retried on the next refresh cycle
when the rate limit has cleared and CivArchive may still have the data.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
