- add POST /model_resolver/local-matches-by-hash backend endpoint
- expand local hash lookup sources to lora_manager_archive, popular, and model_list
- deduplicate backend hash matches by SHA256 value instead of source-hash tuple
- add frontend sync for remote hash matches and hash-label disambiguation
- propagate sha256, file_info, selected_file, and selected_version in model selection/download flows
- render local match status with consistent hash labels across search panel and resolvers
Replace duplicated filename-to-query logic with a single
\uild_filename_search_queries\ helper in matcher/civarchive/huggingface,
standardize Civitai URL generation in the frontend via \getCivitaiModelUrl\,
unify browser user-agent and response file-size extraction in the downloader
path, and derive workflow node-type category hints from widget categories
rather than maintaining a large static mapping.
- Add clear_all_search_caches() in core/sources/__init__.py and replace 4
individual cache-clearing call sites in __init__.py with the new helper.
- Add shared parse_civitai_model_path() in core/type_utils.py and use it
from both CivitAI and CivArchive URL parsers to remove duplicated regex logic.
- Rename parse_huggingface_url() to get_huggingface_repo_and_file_from_url()
in core/resolver.py for clarity.
- Add stripModelExtension() and hasModelExtension() utilities (with a single
MODEL_EXTENSIONS_REGEX) in web/resolver/utils/render_format_methods.js
and replace repeated inline regex patterns across search_panel.js,
model_info_methods.js, tabs_loaded_methods.js, and queue_methods.js.
Extract MODEL_EXTENSIONS and format_size_bytes into core/type_utils, removing duplicates from workflow_analyzer and downloader. Add getFilenameFromPath to render_format_methods and replace inline path-splitting logic across JS resolver actions, search, and views.
- Introduce JobProgressTracker in __init__.py to consolidate the duplicated
thread-safe progress/cancellation state and cleanup logic previously
implemented separately for search and hash jobs.
- Extract shared pollHashCalculation method in model_info_methods.js to
centralize hash polling, abort checking, and error handling across all
call sites.
- Use getMissingModelKey in search_panel.js to eliminate inline key
construction duplication.
Implement end-to-end hash comparison for local models and local matches.
Backend:
- Add /model_resolver/local-model-hashes API endpoint
- Add get_local_model_hash_metadata() to read sidecar SHA256 metadata
Frontend:
- Add Compare Hashes context menu item and CSS
- Pass missing_key/missing_search_key through resolveDownloadMethods
- Enhance local match rendering with hash status and context data
- Add hash normalization, formatting, and candidate collection helpers
- Wire compareLocalModelHashesWithCurrentFinding action
Also update HuggingFace author index metadata.
- Propagate missing_search_key through resolution queue and download
resolution paths so cache lookups remain consistent after workflow
re-analysis or active searches.
- Set preserveSearchCacheAcrossNextWorkflowSync before reloading
workflow data following batch queue/resolve operations.
- Remember applied resolu tions (rememberAppliedResolvedSelections)
after both queue and direct downloads so alias tracking is not lost.
- getMissingSearchKey now short-circuits on missing_search_key and
search_key fields instead of always deriving from node identity,
fixing spurious cache misses.
- Guard settleInactiveSearchProgress so non-active workflow states
only update the cache when the active workflow hosts a matching
background search job.
- After workflow analysis, call applyResolvedSelectionAliases
so analysis data reflects user selections.
- When switching workflows, clone and merge the previous workflow
search-result cache and migrate background search jobs that are
still running, so active searches are not interrupted or dropped.
Integrates aria2c as an alternative download backend alongside the built-in
Python downloader. Adds settings for selecting the backend and configuring
the aria2c executable path, plus a status check button in the options panel.
Backend changes:
- core/downloader.py: aria2 daemon lifecycle, JSON-RPC client, pause/resume
state machine, and transfer tracking
- core/settings.py: download_backend normalization and aria2c_path setting
- __init__.py: new API routes for pause, resume, and aria2 status checks
Frontend changes:
- options_methods.js: backend selector, aria2c path input, check button,
and local storage persistence
- resolve_download_methods.js: pause/resume handlers, pending status
transitions, and progress rendering for aria2 downloads
- queue_methods.js: queue panel pause/resume buttons, backend badge, and
debounced panel updates
- search_panel.js: actionsHtml support for composite action buttons
Apply normalizeDownloadCategory to the category token before checking
color mappings, preventing unknown-category tokens from falling through
to the default class. Also add clip_gguf to the clip color mapping.
- Expand preferred categories to include clip_vision, ipadapter, sams, ultralytics
- Add clip_gguf alias for text_encoders in both backend and frontend
- Add explicit diffusion_models/text_encoders mappings in CATEGORY_MAP
- Refactor get_model_roots to skip custom_nodes/configs, use template key
aliases, accept new folder keys from known categories, and normalize
ultralytics paths (yolo/bbox, yolo/segm, yolo -> parent)
- Pass missing object through resolve_download_methods and search_panel
so category selection can use widget hints
- Refactor frontend download target resolution:
- getMissingSupportedDownloadCategories respects widget/type hints
- getNodeTypeDownloadCategory splits exact-node map and adds
normalizeNodeCategoryToken / getNodeTypeDownloadCategoryFromTokens
for GGUF and future node families
- getSourceResultDownloadCategory and _sourceSupportLevel accept
missing for unified logic
- Prioritize dynamic widget category hints in workflow analyzer and
include category_hints/folder_key_hints on all model refs
- Add folder count badges and path normalization in missing browser
(normalizeMissingFolderRootPath handles category suffixes / bbox/segm)
Add a /model_resolver/search-cancel/{progress_id} endpoint with a
SearchCancelled exception so in-flight source searches abort cleanly on
batch cancel. Track sourceProgressIds alongside sourceControllers so the
frontend can cancel backend progress polls. Rework CivArchive extraction
with robust __NEXT_DATA__ parsing, hash-page file resolution, model-link
scraping, confidence-based candidate filtering, and expanded type maps.
Add conftest.py for pytest path setup, update .gitignore for scratch/,
and refresh huggingface-author-index.json.
Replace scattered api.fetchApi(...) + response.ok + response.json() boilerplate
across all web/modules with a unified this.fetchJson() helper defined on the
base dialog class.
Changes:
- dialog.fetchJson(url, options, context) handles headers, status checks, and
JSON parsing in one place
- Extracts server-side error messages (errData.error) for better UX
- Optional silent flag suppresses notifications for background/refresh requests
- All callers simplified: async return, shorter bodies, consistent error handling
- BOM removed from workflow_update_methods.js and tabs_loaded_methods.js
- Moved formatting helpers (formatBytes, formatDuration, getDownloadEtaText,
formatDownloadProgressMeta) and getLocalMatchDisplayInfo from
resolve_download_methods.js and missing_browser_methods.js into
render_format_methods.js for shared reuse.
- Relocated settings-related methods (getStoredTokens,
applyFrontendLoggingPreference, loadSettingsFromServer) and their
log_system imports from download_target_methods.js into
options_methods.js, where they are actually consumed by the options UI.
- Simplified missing_browser_methods.js to use
getLocalMatchDisplayInfo instead of inline local-match code.
Remove duplicated refreshMissingListRowForMissing and
refreshMissingListRowLocalMatch implementations, replacing them
with a single refreshMissingListRow method in missingBrowserMethods.
Also strip unnecessary BOM from imports.
Replace scattered inline encodeURIComponent(JSON.stringify(...)) and
manual oncontextmenu/dataset assignments with unified helpers added to
renderFormatMethods:
- encodeContextMenuModel(context): safe JSON serialization for data-model
- getContextMenuAttrs(context, tooltip): HTML attribute string for
template literals (used in 5 files)
- getContextMenuProps(context, tooltip): props object for imperative
element assignment (used in missingBrowserMethods)
- setContextMenuElement(element, context, tooltip): sets or clears
data-model, data-tooltip, and oncontextmenu on a DOM element
Reduces duplication across 5 modules, adds try/catch around JSON
serialization, and removes 14 lines of repeated oncontextmenu wiring.
- Add right-click 'Open Folder' context menu to folder browser rows and
root group headers, with left-click guards to prevent context actions on
non-primary clicks
- Deduplicate local model entries before filtering using path-based identity
and category ranking to prefer the active download category
- core/workflow_analyzer.py: add dynamic widget category hint extraction via threaded cache and sentinel-prefix matching; get_model_widget_category_hints now returns all inferred categories including dynamic ones, improving model folder detection for workflows with encoded category metadata.
- web: add ? help icon with context-aware tooltip to download folder selector explaining why a category was auto-selected (preserved manual choice, node type hint, workflow-detected category, default fallback, or best available). Tooltip refreshes live as the user changes the folder selection.
- web/css/resolver-main.css: style the new category help badge.
- metadata/huggingface-author-index.json: update index entry.
- Update HuggingFace author index metadata (file count, repo entries)
- Style empty folder browser state with grid layout and monospace path display
- When a non-existent subfolder is typed, inform user the path will be created during download
Switch download target dropdowns from absolute-positioned elements clipped by
scroll containers to body-ported fixed-position lists managed by
downloadTargetMethods. Adds showDropdownList, hideDropdownList,
portalDropdownList, positionFloatingDropdownList, and related helpers so the
UI can keep dropdowns visible and correctly placed near their anchors.
Add bindDropdownOutsideDismiss and pointer-guard utilities to downloadTargetMethods,
then use them in resolveDownloadMethods to remove duplicated hide-on-blur and
setTimeout-based hide patterns. This centralizes dropdown dismissal behavior and
handles pointer/mouse/touch events consistently across dropdown lists.
Make expandable folder rows clearly distinguishable and allow clicking the
left-side toggle zone to expand/collapse folders instead of selecting them.
- Add is-expandable class and data-state-key to expandable rows
- Extract toggleFolderBrowserState helper for reuse
- Extract isFolderBrowserToggleZoneClick to detect toggle-zone clicks
- Adjust toggle button width/positioning via CSS for expandable rows
- Apply box-sizing: border-box to folder browser toggle
- Remove fixed max-height constraints from folder browser scroll container
- Calculate available height dynamically based on viewport space
- Account for container chrome (borders, padding) and child elements when setting scroll max-height
- Use clampNumber for popup positioning to keep it within viewport bounds
- Ignore scroll events originating from within the list to prevent jittery repositioning
- Add box-sizing: border-box to folder browser for consistent sizing
- Remove 'Use model root' and 'Use typed folder' manual rows from the folder browser
- Remove category badge and folder browser header elements
- Add nested scroll element detection for scroll delta handling
- Preserve scroll position when re-rendering the folder browser on toggle
Refactor populateSubfolderOptions into renderDownloadFolderBrowser to show a
persistent floating panel with an expandable folder tree grouped by base
directory. Includes live preview of the selected target path, manual
typed-folder rows, and window/viewport-aware positioning. Prevents duplicate
portal mounts and cleans up resize/scroll bindings on hide.
Frontend now polls a new backend endpoint (/model_resolver/search-progress/{progress_id})
to display real-time, per-source progress messages and percentages during model searches,
replacing static 'Searching...' indicators with granular status updates.
- Centralize model key logic: replaced inline key construction across 5 files with
getMissingModelKey() and new getMissingModelDomKey() helpers in lifecycle_graph_methods.
- Extend getMissingModelKey(): now incorporates nested_key, category, and an identity
field (original_lora_name, original_path, expected_filename, name, filename, urn_string)
so that distinct models within the same subgraph produce unique keys.
- Fix duplicate DOM ID collisions: getMissingModelDomKey() encodes the full key as a
URL-safe base64 string for all element IDs (local-matches-body, selected-bar, combo,
locate, local-matches-refresh, selected-apply, selected-remove), preventing collisions
between subgraph instances that share a node_id/widget_index.
- Store missing_key on resolution objects in queue_methods to keep the canonical key in
sync between queue and browser views.
- Add settleInactiveSearchProgress to properly finalize pending/running
sources when background jobs are interrupted or removed
- Hook window focus and visibilitychange in ResolverManagerDialog to
trigger syncSearchProgressAfterResume so UI recovers after tab hide
- Refactor reconnectActiveSearchProgress to reuse sync logic and pass
workflowKey through startEstimatedSearchProgress and refreshSearchUiForMissing
- Add grid layout and CSS for a new 'Suggest' column in the download target
panel, including a lightbulb SVG icon button with a flash animation
- Update download_target_methods.js to render the suggest button in the
dedicated grid area and apply a ‘mr-is-suggesting’ animation on click
- Adjust search button label logic so it shows 'Search Again' when there is
a renderable search state in the panel, otherwise show 'Search Online'
- Remove readonly from base model and download target category inputs to allow typing
- Add input event listeners that filter dropdown options in real-time
- Normalize base model tokens for fuzzy search matching
- Introduce getSearchBaseModelInputValue to resolve typed input to known options
- Add optional { sync } parameter to setSearchBaseModel to avoid redundant UI sync during input
- Show/hide dropdown lists based on filtered result count
- Add version-trimmed alias generation to handle trailing version zeros (e.g. '4.0' -> '4')
- Track newly added base model names in 'new_models_added_list' field
- Include Ideogram 4.0, Boogu, and PolyGen in base-models registry
- Update web resolver to match aliases using both original and version-trimmed tokens
- Enhance base models update UI to show which new models were added
Make the Auto base model label context-aware by resolving the correct base
model for each individual missing/search entry instead of relying solely on
the dominant workflow base model. Search metadata, saved download targets,
local file matches, and path hints are now consulted before falling back
to the workflow-wide default.
When a model search returns no results for a specific base_model, retry the
same query with base_model_context=None so the provider can match any model.
Fallback matches are marked with any_model_match and base_model_fallback on
the result payload. The UI reports 'Found (Any Model)' when a result came
from this fallback path, and the cancel/progress button styling is tightened
to 16x16px with border-radius.
Implement AbortController-based cancellation for individual search sources
during background resolution. Users can now cancel a single pending/running
source search via a new cancel button in the progress UI, while other sources
continue uninterrupted. Add cancelled status tracking, source-level AbortController
management, CSS styling for the cancelled state, and cleanup on job teardown.
Also refactor getBackgroundSearchJob for consistent job lookup across callers.
Dariusz L