- Add POST /api/lm/switch-channel endpoint with git init / ZIP fallback
- Add _backup_git/_restore_git helpers with safe rollback
- Version-info endpoint now returns has_git flag for auto-detection
- Check-updates always returns releases (changelog) regardless of channel
- Nightly mode shows 'N commits behind main' with commit hash and date
- View on GitHub link points to /commits/main in nightly mode
- Channel toggle UI with pill-style buttons in update modal
- Confirmation dialog with Esc / backdrop-dismiss support
- Channel derived from has_git on every page load, no localStorage
- i18n: 11 new keys translated across 9 non-English locales
- CSS: unified card-style sections in _base.css
- Tests: 8 new tests covering switch-channel, nightly response, init_git_repo
Two bugs prevented type-signature-based fallback from working:
- metadata_hook.py used getattr(obj.__class__, 'RETURN_TYPES')
which fails when _async_map_node_over_list is called with
a class (not instance) — obj.__class__ is the metaclass
'type', which has no RETURN_TYPES. Fixed: getattr(obj, ...).
- metadata_registry.py used type(extractor) is GenericNodeExtractor
to dispatch return_types. NODE_EXTRACTORS stores class
references, not instances; type(Class) is always 'type',
never the class. Fixed: extractor is GenericNodeExtractor.
GenericNodeExtractor (previously a no-op) now inspects
RETURN_TYPES to detect MODEL loaders and CONDITIONING
encoders in nodes not registered in NODE_EXTRACTORS.
- Propagate return_types from the hook layer through the
registry to GenericNodeExtractor.extract() and update().
- MODEL detection: scan ckpt_name/unet_name/model_path/
model_name/gguf_name fields, validate by extension.
- CONDITIONING detection: scan text/clip_l/t5xxl/prompt
fields, store prompt text and conditioning tensor.
- _fill_missing_metadata also checks node_cache, so
GenericNodeExtractor-handled nodes survive cache.
Users can now right-click nodes and assign meta hints
(primary_model, primary_sampler, positive_prompt,
negative_prompt) to override the metadata processor's
heuristic inference.
- Store extra_data from the API request so workflow node
properties (including lm_marker_role) are accessible
during metadata processing.
- _get_user_marks scans extra_data.extra_pnginfo.workflow
for meta_* marks, falling back to prompt.original_prompt.
- extract_generation_params checks user marks before
heuristic inference for sampler, model, and prompts.
- Warn on duplicate marks or invalid marked nodes.
Drop the SequenceMatcher-based fuzzy_match fallback that froze the server
when FTS returned empty results. FTS now returns empty set for zero results
(no fallback), and when the index is not yet ready, search returns empty
rather than scanning all items in Python.
The previous tooltip was misleading: users thought workflow embedding was
automatic. New wording explains this opt-in flag stores the complete
workflow inside images, allowing one-click restoration via drag-and-drop.
PNG and WebP only.
Add two new optional parameters to the Save Image node:
- webp_method (INT, 0-6, default 6): Controls WebP compression level.
0=fastest/largest, 6=slowest/smallest. Previously hardcoded to 0.
- jpeg_subsampling (INT, 0-2, default 0): Controls JPEG chroma
subsampling. 0=4:4:4 (best quality), 1=4:2:2, 2=4:2:0.
Frontend JS extension hides/disables each parameter when the
selected file_format doesn't apply (e.g., webp_method is hidden
when saving as PNG or JPEG). 7 new tests cover parameter plumbing
and default consistency across INPUT_TYPES, save_images(), and
process_image().
hydrate_model_data replaces model_data with .metadata.json content which
may lack sha256 (corrupted file, concurrent write, etc.). Restore the
cached sha256 after hydration and persist the fix back to disk so
subsequent lookups don't hit the same error.
Also improve error log to include file_path for debugging.
- Replace plain-text Lora hashes with Hashes JSON dict matching A1111 convention
- Add Civitai resources JSON array with AIR URNs for direct model version linking
- Add Clip skip, Version: ComfyUI fields to generation params line
- Build AIR strings from local scanner cache (no API calls needed)
- Add complete sampler name mapping (CIVITAI_SAMPLER_MAP) and base model → AIR slug mapping (BASE_MODEL_AIR_SLUG) sourced from civitai ecosystem constants
- Remove lora text prepending from prompt line; LoRA info now in structured JSON sections
LM Studio and some other OpenAI-compatible servers reject
response_format=json_object but accept json_schema. Switch to the
equivalent json_schema format and add a fallback that retries
without response_format when the provider rejects the format type.
widget.value is a getter/setter that returns a new array on every read,
so handleStrengthDrag with updateWidget=false mutated a discarded copy.
Introduce __dragActive flag to suppress renderLoras in setValue during
drag, allowing mutations to persist through widget.value without
destroying the DOM. Use try-finally to guarantee flag cleanup.
The validation in getApiEndpoints threw for page types not in
MODEL_TYPES (e.g. 'recipes'), crashing the recipes page initialization
when FilterManager calls it via createBaseModelTags(). The throw was
synchronous and outside the fetch().catch() chain, causing an uncaught
promise rejection that aborted the entire app initialization.
getApiEndpoints is a URL builder -- validation belongs to callers that
need strict type checking (they already use isValidModelType()). For
non-model-type pages like recipes, the generated URLs are correct
(the backend does have /api/lm/recipes/* routes).
Fixes regression from f53f859a (feat(filter): add debounced tag search).
- Move preset apply handler from span.preset-name to div.filter-preset so
clicking anywhere on the tile triggers the preset, not just the label text.
- Add whitespace heuristic in showToast() to skip translate() for plain
messages that are already translated at the call site. This prevents
i18next from logging 'Translation key not found' for pre-translated
strings like 'Preset "name" applied'.
- Use safe .get() in RecipeCache._resort_locked instead of itemgetter to prevent KeyError when recipe missing created_date; align sort key with _sort_cache_sync (prefer modified, fallback created_date, fallback 0)
- Add base_model to allowed_fields in persistence_service.update_recipe() so the field passes validation
- Route bulk base model updates through updateRecipeMetadata() on recipes page instead of generic saveModelMetadata(), matching existing isRecipesPage pattern used in setBulkFavorites and saveBulkTags
The drop indicator top position was calculated using only
getBoundingClientRect() offsets (post-CSS-transform viewport space)
without accounting for container.scrollTop (pre-transform layout space).
This caused the indicator to drift upward as the user scrolled down,
eventually disappearing entirely.
Fixed by adding container.scrollTop to the position calculation and
only dividing the GBCR visual-diff portion by scale, since scrollTop
is already in pre-transform coordinate space.
- Add enable_civarchive_api toggle (default on) to allow disabling
CivArchive to avoid its rate-limit windows entirely
- Add metadata_provider_order dropdown with two presets:
CivitAI → CivArchive → Archive DB (default) and
CivitAI → Archive DB → CivArchive
- Wire both settings through backend (metadata_service, settings_manager,
misc_handlers) and frontend (SettingsManager, state, settings modal)
- Reorder Metadata section in settings modal: toggles → status/management
→ fallback order, for natural top-down workflow
- Make update_metadata_providers() log the effective provider chain
using actually-registered providers rather than settings assumptions
- Add 5 test cases covering all provider-combination paths
- Complete i18n translations for 6 new keys across all 9 non-English locales
- Add PersistentModelCache.update_single_model() for lightweight targeted
SQL update (single row + incremental tag/hash deltas, no full table scan)
- Add ModelScanner.sync_cache_from_metadata() with compare-first logic:
skips entirely when cache is already in sync; when stale, updates the
entry in-place (O(1) instead of O(n) remove+append), incrementally
adjusts tag counts/hash index/version index, and resorts only when
sort-relevant fields changed
- Wire sync_cache_from_metadata() into BaseModelService.get_model_metadata()
via fire-and-forget asyncio.create_task — disk I/O is already paid for
- Include identity re-validation guard against concurrent cache replacement
- Add 16 tests covering _cache_entries_differ, sync_cache_from_metadata
(no-change, in-place, fallback, conditional resort), and
update_single_model (insert, tag delta, hash delta)
When a KSampler marked as 'Send Gen Params Target' has widget inputs
wired to Primitive nodes (PrimitiveNode, PrimitiveInt, PrimitiveFloat,
etc.), sending gen params from the Lora Manager UI now updates the
Primitive node's value instead of the KSampler widget. This is
necessary because ComfyUI's execution engine reads from the connected
input, ignoring the widget value when a wire is present.
Also fix two minor issues found during review:
- Remove unnecessary String() wrapping on numeric gen params (seed,
steps, cfg) to preserve native types through the JSON/WS path
- Correct misleading isNodeEnabled comment: LGraphEventMode values
are 0=Always, 2=Never, 4=Bypass (not 'Normal/Enabled')
When Civitai returns 404 for /models/{id} (e.g. due to Civitai API bug
where un-deleted models still get 404), the fallback to CivArchive
provides metadata. However CivArchive may return mirrors with every
entry marked deletedAt, while the file's downloadUrl is still valid.
Before this fix, _build_download_urls_from_file_info used an if/else
that skipped the downloadUrl fallback whenever the mirrors array was
non-empty, even when all mirrors were filtered out. Now downloadUrl
is always tried when no usable mirror remains.
Also deduplicated the inline mirror-processing code at the second call
site by replacing it with a call to the shared helper.
Refactor _create_session() to make-before-break: snapshot old session,
assign new one first, then close old. Previously, concurrent download
retries called _create_session() without the session lock (violating its
docstring contract) and closed the old session while other coroutines
held active references — causing aiohttp to raise "NoneType has no
attribute connect" when dereferencing the torn-down connector.
Also wrap the two _create_session() calls in the integrity-retry and
network-retry paths with self._session_lock to match the locking
discipline used by the session property and refresh_session().
- Add Notes/Description tab switching with tab state persistence in widget value
- Lazy-load model description and version description from /lm/loras/metadata
- Render CivitAI HTML descriptions inline via v-html
- Auto-fetch description when LoRA selection changes while on Description tab
- Fix Vue mode height containment via contain:layout size (lm-vue-node class)
- Fix scroll wheel isolation: widget scroll vs canvas zoom in both render modes
- Add docs/comfyui-dual-mode-widgets.md with widget rendering patterns
In ComfyUI Vue render mode, WidgetDOM.vue reuses its component instance
during undo/redo without re-calling mountWidgetElement(), leaving newly
created widget containers detached from the DOM.
- AutocompleteTextWidget: scan for empty containers by ID prefix and reuse
- Loras widget: scan for empty .lm-loras-container elements and reuse
- Prevent duplicate event listeners by guarding listener setup on new
containers only
- Keep container in DOM on cleanup (clearChildren instead of remove)
so it can be found and reused by the next factory invocation
When using full path lora syntax, the context menu (single/bulk)
and bulk copy actions were passing only the file basename to
buildLoraSyntax(), ignoring the folder prefix. This caused the
output to look like legacy A1111 format even when full path mode
was enabled.
Aligns all entry points with ModelCard.handleSendToWorkflow(),
which correctly includes the folder prefix.
Also fixes selectAllVisibleModels() to cache the folder field,
preventing missing prefix on select-all-then-send flows.
Add GET variants of the two POST endpoints used by the send-to-workflow
feature. Parameters are read from query string instead of JSON body,
supporting both simple repeated node_id params and JSON-encoded node_ids
for complex graph references.
In add_to_queue, check download_history before INSERT OR IGNORE. Without
this check, a fire-and-forget /queue/complete failure on the extension side
would allow the same download_id to be re-inserted after complete_download()
deleted it from the queue — creating phantom queued entries for already-
finished downloads.
Add a pure frontend node that shows filename and editable notes for
a selected LoRA. Connect any output from a LoRA Loader/Stacker/Randomizer/
WanVideoSelect to the lora_source input — selecting a LoRA in the source
widget updates the info display automatically.
- Python node (LoraInfoLM): display-only, no workflow execution
- Vue widget: filename label, auto-sizing notes textarea, save button
with ComfyUI toast feedback on save
- Frontend extension: wire-based selection propagation with stale-response
race guard; clears display on wire disconnect
- Backend: get-notes endpoint now returns file_path alongside notes;
matching supports full-path lora syntax; fix NoneType crash in
trigger words endpoint; document cache file_name invariant
- Wired into all four lora widget nodes (Loader, Stacker, Randomizer,
WanVideoSelect)
- workflow_registry.js: add force param to refreshRegistry(), bypass fingerprint
dedup when responding to lora_registry_refresh WS message. Without this, the
backend's wait_for_all() times out after 0.5s because the frontend skips the
register-nodes POST when the workflow fingerprint hasn't changed (common after
ComfyUI restart with an empty or unchanged workflow).
- misc_handlers.py: demote 'No nodes registered after refresh' from WARNING to
DEBUG — empty workflows are a normal operational state, not a warning-worthy
condition.
- Handle compound node IDs (e.g. "252:0") from expanded group subgraphs
to fix 400 Bad Request on workflows with group nodes
- Frontend proactively pushes node data via afterConfigureGraph and
LiteGraph hooks (onNodeAdded/onNodeRemoved/graphChanged), eliminating
WebSocket round-trip latency for most "Send to Workflow" operations
- Add content-fingerprint dedup to skip duplicate register-nodes POSTs
- Fast-path cache returns immediately when tabs are registered (including
0-node registrations), avoiding unnecessary WS refresh cycles
- Distinguish "Empty Registry" from other errors in standalone UI toast
- Reduce WS refresh timeout 2s→0.5s, add cooldown and lock to prevent
concurrent refresh storms
- All [LM:Registry] logs at DEBUG level
Changes:
- Backend: _validate_folder_paths() now checks checkpoints↔unet overlap
within the same library using os.path.realpath() for symlink resolution
- Backend: set() calls _validate_folder_paths() for both folder_paths and
extra_folder_paths before writing
- Backend: extracted _normalize_path_set() helper to eliminate duplicated
normalization logic
- Frontend: inline error display with red border + error message below the
conflicting input, no save triggered
- Frontend: path normalization (strip trailing slash, lowercase) in pre-check
to reduce false negatives vs backend realpath
- Frontend: asymmetric error UX — message only on the user-edited side,
red border on the pre-existing conflict side
- CSS: has-error styles with hardcoded rgba fallback for older browsers
- i18n: checkpointUnetOverlap + checkpointUnetOverlapInline keys added to
all 10 locale files
Remove dynamic height calculation that auto-resized the node when
LoRAs are added or removed. The widget now stays at the size the user
sets via the node resize handle, scrolling when content overflows.
- Drop updateWidgetHeight() and hardcoded entry-count height math
- Set --comfy-widget-min-height once (200px) instead of recalculating
- In Vue mode: add contain:layout+size to break the ResizeObserver
feedback loop that forced node growth with content (CSS via
.lm-loras-container.lm-vue-node scoped to vueNodesMode only)
- Remove unused "Node 2.0: Maximum visible LoRA entries" setting
- Bulk refresh filter now excludes models with hf_url
- Individual refresh for HF models only checks CivitAI API
- CivArchive client validates model IDs before querying
- Add onerror handler on <img> previews to fallback to no-preview.png
- Fire async cache cleanup when preview file returns 404
- Add ModelCache.clear_preview_by_path() for safe stale-url removal
- Downgrade /api/lm/previews 404 log from warning to debug
- Merge Relink to Civitai and new Link to HuggingFace into a single
'Link Model' submenu with sub-options for each source
- Add POST /api/lm/set-hf-url endpoint to associate a model with a
HuggingFace repo URL, saving hf_url to .metadata.json
- Add link_hf_modal.html for URL input, following relink-civitai pattern
- Use update_single_model_cache instead of add_model_to_cache to
prevent duplicate cache entries after linking
- Remove os.path.realpath usage for consistency with relink-civitai
- Raise errors instead of silently falling back to LoRA scanner when
model root cannot be determined
- Scope .input-group CSS rules to modal IDs to fix style conflicts
with download-modal.css
- Add i18n keys across all 10 locales with translations for
zh-CN, zh-TW, ja, ko, de, es, fr, he, ru
When Civitai returns 429 (Too Many Requests) during example image
downloads, the previous behavior treated all failures identically and
permanently removed the corresponding images from model metadata —
making them impossible to retry.
This commit adds:
- 429 detection + Retry-After header parsing in download_to_memory
- Exponential backoff retry (up to 3 attempts) in
download_model_images_with_tracking
- Separate tracking of rate-limited vs permanently failed URLs
- rate_limited_models progress tracking persisted to disk
- Rate-limited models are NOT added to failed_models/processed_models
so they are automatically retried on subsequent download runs
- Force mode clears failed_models when rate-limited images exist
- Parse limit.output from model catalog alongside model IDs
for per-model max output token limits
- Use catalog lookup in chat_completion_json() to set max_tokens;
fall back to 4096 for unknown models (e.g. local Ollama)
- Remove the JSON retry (response_format → plain text fallback);
keep _try_salvage_json as last-resort for truncated responses
- Reduce Ollama num_ctx from 32768 to 8192 (sufficient for
metadata enrichment, saves VRAM)
- Fix stale test comment referencing removed retry
Remove tests/enrich_hf_validation/baselines/ from git tracking
(.gitignore entry + git rm --cached). These contain README snapshots
from community HF repos that may include NSFW/sensitive content.
Local files are preserved on disk for offline reference.
Commit 9a0d866b changed _strip_fenced_code_blocks to preserve bash/shell
code blocks (they carry CLI setup and trigger-word metadata signal).
Update the two affected tests to expect bash content in the output
instead of asserting it is stripped.
- Rename test_bash_code_block_stripped → test_bash_code_block_preserved
- Update assertions: expect 'pip install' in result
The bare call inside _build_prompt_context
would raise NameError because class methods don't close over class-level
scope. Use instead to trigger attribute lookup.
Update enrich_hf_metadata prompt.md clue locations for better LLM accuracy.
Update baseline report to v2 (mean 69.0, 46 models, +2.2pp vs baseline 71.1%).
Consolidate README snapshots into baselines/readmes/.
_Previous_ _find_scanner_for_model and identify_model_type contained ~25 lines
of identical scanner-iteration + path-matching logic. Factor it into
_find_model_entry() so a new scanner type or edge-case fix can't drift apart.
- Rename py/agent_cli/ -> py/metadata_ops/ (module was never agent-related)
- Rename tests/agent_cli/ -> tests/metadata_ops/
- Remove 9 low-value/debug INFO log points across agent_handlers.py,
agent_service.py, llm_service.py, and metadata_ops/__init__.py
- Keep LLM raw response at DEBUG level for diagnostics
- Consolidate per-model progress + LLM result into single concise
log line with basename instead of full path
- Update package/class/method docstrings to clarify this is a
pipeline infrastructure, not a true agent loop
Will Miao