Backend:
- Move SHA256 normalization logic from path_utils._normalize_header_sha256
into core/type_utils.normalize_sha256, adding support for 0x-prefixed
hashes and unifying the normalization path used by
extract_safetensors_header_sha256.
- Add tests for the 0x prefix normalization cases.
Frontend:
- Introduce copyTextWithFeedback in web/resolver/utils/html_utils.js
as a shared clipboard helper with configurable feedback text, CSS class,
and duration, replacing repeated inline clipboard copy logic in
globals.js, model_info_methods.js, and the notification utility.
- Centralize showNotification on ResolverManagerDialog and remove the
duplicate implementation from missingBrowserMethods.
- Replace local const localStorage = safeStorage aliases in
model_info_methods.js, missing_browser_methods.js, and
options_methods.js with direct safeStorage usage.
Replace inline regex path/URL normalization in resolver.py and
missing_browser_methods.js with the existing shared helpers:
- core/resolver.py: normalize_workflow_download_url now uses
prepare_remote_size_probe_url from type_utils instead of a local
huggingface-specific regex substitution.
- web/resolver/views/missing_browser_methods.js: path identity
generation now uses normalizePathIdentity from html_utils.js rather
than ad hoc replace/trim/lowercase logic.
This keeps normalization behavior consistent across the codebase and
removes duplicated regex patterns.
- Add run_in_background_thread() in __init__.py to unify success/cancel/error
handling for hash calculation and local metadata build progress
- Extract match_model_by_title_generic() in core/matcher.py and use it from
both CivitAI and CivArchive providers to remove duplicated version/file
selection logic
- Enable safetensors header SHA256 extraction by default and report hash source
through progress updates
- Update options_methods.js to match backend progress behavior
- Add tests/test_title_matching_hashing.py covering title matching and hashing
- Minor metadata updates
- Replace remaining direct requests.get calls with request_public_url,
request_source_response, and request_source_json in downloader,
civarchive, civitai, and popular sources
- Use normalize_model_file_info in CivitAI source instead of inline logic
- Update huggingface-author-index.json metadata
- Add tests for civarchive page/search requests and popular base model updates
- Adjust downloader redirect test for new request_public_url behavior
- Add core/network_utils.py with URL validation, public-IP enforcement,
redirect validation, and safe request_public_url/validate_public_http_url helpers
- Replace direct requests calls in downloader/type_utils/sources with
request_public_url so redirects are validated before follow-up
- Replace netloc string checks with host_matches_domain for CivitAI,
CivArchive, and HuggingFace host matching
- Add HuggingFace Xet download backend with progress reporting and
cancellation support via hf_xet/huggingface_hub
- Harden aria2 installer: reject symlinks/special files in archives,
restrict custom aria2c paths to the managed install, and update UI copy
- Normalize paths with realpath and enforce configured model roots in
local hash/metadata routes, plus remove caller-selected metadata_path
injection in sidecar writing
- Sanitize download/network errors to strip signed query strings before
logging or showing them to the user
core/path_templates.py: allow model alias tokens to match when suffixed
with a version string or plural form, reducing false negatives in
single-folder path resolution.
metadata/huggingface-author-index.json: update Comfy-Org index timestamp
and repository ordering.
This change introduces a complete Local Metadata Builder feature in the resolver
options panel. It allows users to scan local model files, create or update
.metadata.json sidecars, fill missing SHA256 hashes from safetensors headers or
file hashing, and review run history with cancel/error handling.
Changes include:
- backend API: /model_resolver/metadata-build/{capabilities,start,progress/{id},cancel}
- backend worker: core/metadata_builder.py with parallel metadata generation
- frontend section: Local Metadata Builder tab in Options
- frontend polling UI: progress, worker breakdown, history table, status/notifications
The URL parser previously required the hostname to contain "civitai.com",
so CivitAI download links using the "civitai.red" domain were treated as
non-CivitAI URLs and rejected.
This extracts the host check into _is_civitai_host() and accepts both
"civitai.com" and "civitai.red" base domains, so resolve/parse flows
handle all CivitAI download links correctly.
Previously, identify_missing_models grouped missing entries only by filename,
which could merge unrelated models from different categories that happen to
share a filename. This change deduplicates by (category, filename) and tracks
all_node_refs with a reference_count so the frontend can preserve one
resolution per node reference instead of collapsing duplicates.
Frontend changes in resolve_download_methods.js iterate over all_node_refs,
and missing_browser_methods.js now shows '+ N refs' in the missing-model
label and clarifies the locate tooltip when multiple workflow references
share the same missing model.
- Merge GET and POST aria2 status handlers into one route
- Remove redundant calculate_file_sha256 wrapper and use path_utils directly
- Remove redundant _normalize_archive_image, _normalize_filename, and _similarity wrappers
- Replace urllib with requests for remote JSON fetching in model_list
- Update metadata timestamps
Add an early guard in core/sources/civitai.py to return None for non-string or empty URL inputs before calling urlparse. This prevents exceptions and unexpected behavior when callers pass None or empty values, making the parser more robust.
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.
Replace duplicated SHA256 normalization, metadata sidecar discovery, trained-words
extraction, type coercion, and cache helpers into shared modules. Clarify that the
_civitai.py/_civarchive.py/_huggingface.py/_workflow_analyzer.py refactorings are
behavior-preserving so future contributors keep behavior aligned in one place.
Dariusz L