The reorder shortcut cannot be made reliable in this UI. `Alt + Arrow` is
the browser's tab-history / back-forward gesture on several platforms,
and the modal already binds bare `ArrowLeft`/`ArrowRight` to model
navigation, so the binding either did nothing — a keypress with nothing
focused never reaches a listener on the tag list — or fought the browser.
An affordance that occasionally navigates the page away is worse than
having no keyboard path at all, so drop it.
Reordering is pointer-only again: drag the chip (tags) or its `⠿` grip
(trigger words, whose chip body is click-to-edit). Everything that existed
only to serve the shortcut goes with it — the keydown listener, the hover
tracking used to resolve the target chip, the aria-live announcements, the
per-grip position labels and `moveItemWithinContainer`. The grip becomes a
decorative, non-focusable `<span>` (`aria-hidden`, behind a 5px drag
threshold) instead of a `<button>`, so it no longer promises a keyboard
action it cannot perform.
The tooltip and hint drop the shortcut mention in all 10 locales
(`common.reorder.dragHandle` = "Drag to reorder" and the localised
equivalents); `common.reorder.ariaLabel` and `common.reorder.announcement`
are pruned from every locale by the sync script. The i18n guidelines
record the decision so no shortcut is re-added without re-adding the keys.
Fill in the 4 `modals.download.progress.*` keys added by the previous
commit across all 9 locales, so no `[TODO: Translate]` placeholder remains
and the "no remaining placeholders" claim in the guidelines holds again.
No new terminology: `metadata` reuses the §5 row (fr métadonnées, de
Metadaten, es metadatos, ru метаданные, he מטא-נתונים, ja メタデータ,
ko 메타데이터, zh-CN 元数据, zh-TW 中繼資料) and the fetching phrasing
mirrors each locale's existing `download.fetchingRepoFiles` /
`fetchingVersions` (de passive "werden abgerufen", es "Obteniendo", fr
"Récupération des", ru "Получение", he "מביא", ja "取得中", ko "가져오는
중"). "model file" follows `errors.noModelFiles` in each file.
`{name}` and `{source}` are verbatim §1-R2 placeholders. `{source}` is
replaced at runtime with the *untranslated* platform name, so its
surrounding spacing follows each locale's `modelCard.actions.viewOnSource`
precedent — ja `{source} から`, ko `{source}에서`, zh `从 {source}` /
`從 {source}`, he `מ-{source}` (as in the existing `מ-CivitAI`), ru
`из {source}` (as in `из Workflow`) — and no brand ever appears inside the
translated text.
Punctuation: ASCII `:` for the Latin / Cyrillic / Hebrew locales and for
ja / ko, whose four sibling keys in the same `progress` block already use
ASCII; French keeps this file's ` : `; zh-CN / zh-TW use full-width `：`
like their siblings.
The guidelines gain a status block recording the pass and those spacing
precedents, so a future source added to the same slot does not have to
re-derive them.
Verified: `pytest tests/i18n/test_i18n.py` 20 passed,
`sync_translation_keys.py --dry-run` reports no drift, `npm test` exits 0
(1259 JS + 91 Vue). Each locale file gains exactly 4 lines — the values
were substituted as literals rather than re-serialising the JSON, so no
formatting churn.
A ModelScope or Hugging Face download landed as a bare filename, hash and
source link; the model card stayed empty until the user ran "Enrich
Metadata with AI" by hand. But everything that makes a CivitAI download
useful — the display name, the description, the tags, the trigger words,
the example images, the preview — is already published by those sites'
public APIs, so asking for it at download time is deterministic work, not
model work.
Add `py/services/model_sources/hydration.py`, called by
`_save_source_metadata()` once the sidecar exists and the file is in the
scanner cache. It fetches the model card plus the site's card extras and
hands them to the same `PostProcessor` the AI skill uses, with an empty
`llm_output`, so the two paths cannot drift apart. What lands:
* `model_name` from the site's own display name (ModelScope's `Name`), so
the card stops showing the local filename — written only while the value
still equals the file stem, since once a user renames a model that
choice is theirs to keep
* `civitai.name` from the matched version's label (`showName`), which the
card renders as the version chip
* `civitai.description` / `modelDescription` from the author summary plus
the README as HTML
* `civitai.images` / `preview_url` from the per-file example images
* `civitai.trainedWords` from the per-file trigger words
* `base_model`, `tags` and `usage_tips` as before
Provenance stays honest: the pass records
`metadata_source = "source:<platform>"` rather than the skill's
`agent:enrich_hf_metadata`, and — because no provider ran — it no longer
stamps `llm_enriched_at`; that stamp is now conditional on the LLM
actually answering, which is what the field means. The five hand-rolled
`civitai` dict merges in the post-processor collapse into one
`_merge_civitai()` helper.
Two guards keep it safe. Only a model whose stored
`source_platform`/`source_url` match the repository being downloaded is
updated, so a local file that merely shares a name never receives another
model's card; and a file already on disk is topped up too, which
back-fills models downloaded before this existed. READMEs and detail
payloads describe the repository rather than the file, so a short-lived
process-wide `ModelSourceCache` (300 s, 32 entries) keeps a batch over one
repository to two HTTP requests. Every failure is logged and swallowed:
hydration can never fail a download.
Fix the hash policy while here. `_save_source_metadata()` went straight to
`MetadataManager.create_default_metadata()`, bypassing the per-type
factory on the owning scanner, so a checkpoint paid a full SHA256 inside
the download request — `CheckpointScanner`/`OtherScanner` deliberately
record `hash_status="pending"` with an empty `sha256` for their multi-GB
files. Metadata is now created through `scanner._create_default_metadata()`.
Hydration copes with the empty hash: `_matching_versions()` falls back to
the repository basename, which is exactly what the download just wrote.
Report both post-transfer stages, which advance no byte counter and so
read as a stall: the bar sat at 100% showing `0 B/s` for the seconds spent
hashing and fetching. `_report_phase()` broadcasts
`{"status": "metadata", "stage": "indexing" | "source", "platform": ...}`,
and `LoadingManager` names the stage in the status line (keeping the batch
position), retitles the item line, replaces the dead speed figure and runs
a sheen over the bar. `stage`/`platform` are machine-readable; the wording
is localised in the frontend.
Finally, `modelscope.ai` is its own catalogue rather than an alias of
`modelscope.cn` — `referall13/EM1` exists only on `.ai` and
`jj3550945163/Krea-2-LORA` only on `.cn` — so its URLs were rejected with
"Invalid model URL format". Register it as `ModelScopeIntlSource`
(`platform="modelscope-ai"`, `msai:` group prefix, its own default
download directory) and derive every URL either deployment builds from a
per-class `base_url`. `modelscope.com` stays an alias of `.cn`, which is
what it redirects to. The frontend source table, the link dialog hints and
the docs mirror the split.
Verified against the live APIs: both reported `.ai` repositories list
their files, read their READMEs and yield name / version / base model /
trigger words / example images. Backend 3092 passed; frontend 1259 JS +
91 Vue passed. The nine locales carry the new progress copy in the next
commit.
The three `common.reorder.*` keys (grip tooltip and hint, the per-grip
aria label, and the aria-live announcement) are rendered in all 9
locales. They sit under `common` rather than in a feature namespace
because both the tag editor and the trigger-word editor render them, and
only `dragHandle` is visible copy — the other two are screen-reader text.
`Alt` and the `↑/↓` glyphs stay verbatim everywhere, the same precedent
as `Shift+Enter` in `modals.model.metadata.notesHint`, because they name
the keys rather than an action. "position X of Y" reuses each locale's
existing counting phrasing (ja `{total} 件中 … 番目`, ko
`총 {total}개 중 …번째`, fr `sur {total}`, ru `из {total}`), and
parentheses follow each file's own convention: full-width in zh-CN /
zh-TW / ja, ASCII in ko and the Latin/Cyrillic locales. No
`[TODO: Translate]` placeholder is left in any locale.
docs/i18n-translation-guidelines.md gains the matching §2 subsection and
status note so a later terminology sweep preserves these renderings; the
leaf-key count in its header is corrected to 2025 at the same time.
Fills in the 35 `sidebar.*` placeholders the folder-sidebar feature series
left behind — view options (tree/list, empty folders), folder creation,
the delete confirmation modal and its result toasts, and folder
renaming, including the undo copy and the "deleting a folder never
cascades over model files" rule the backend enforces.
All nine locales are translated, so no `[TODO: Translate]` placeholder
remains anywhere: the state §7 of the translation guidelines describes
holds again. Terminology reuses the existing §2 maps (folder, model
root, sidebar) with tree/list view added, recorded in a new §2
subsection; the stale leaf-count in the header is refreshed too.
A file was matched to its published version by comparing basenames against
each version's `stats.fileList`. Renaming the weights — routine once a
model is filed away, and the reason the scanner records a sha256 at all —
made the match fail silently, so the file lost its example images and its
preview with no indication why.
The detail payload's `ModelInfos.safetensor.files[]` carries a real sha256
per published file, and the local hash is already on disk, so match on that
first: it is the one identifier a rename cannot invalidate. Exact basename
and `showName` matching remain as fallbacks, and an unknown hash falls
through to them rather than giving up, so a re-encoded file still resolves.
Verified against the live repository: a renamed `c1-st1000` file with its
hash yields the c1-st1000 image, the same rename without a hash yields
nothing, and supplying c1-st2000's hash resolves to the c1-st2000 image even
when the filename claims otherwise.
A collection repository publishes many model files under a single source id,
but enrichment re-read the README and the model-detail payload for every one
of them: eight checkpoints meant sixteen HTTP requests, each detail payload
being 10-22 KB of JSON.
Add `ModelSourceCache`, created by `execute_skill()` for the duration of a
run and passed to the provider through a new optional `cache` argument on
`fetch_model_card_context()`. The agent caches the README (repository-wide
and provider-agnostic), and ModelScope caches its detail payload under a
provider-namespaced key.
Only successful reads are memoised, so a transient failure is still retried
for the next file, and the per-file selection is redone from the cached
payload so a checkpoint never inherits a sibling's example images. Nothing
is retained across runs — a model card can change at any time — and download
URLs are not routed through the cache.
Measured over the eight checkpoints of one ModelScope repository: 16
requests before, 2 after.
To keep the two concerns separable, `_build_card_context()` now turns a
detail payload into a `ModelCardContext` as a pure function.
`_build_prompt_context()` was only reached when the LLM was configured,
so a user with no provider got nothing at all from a linked model source
— no preview, no example images, no author summary, no tags — even
though all of that is deterministic data from a public API.
Split the model-card fetch into `_load_source_card()`, which runs for
every source-backed enrichment, and have the post-processor apply its
result whether or not the LLM runs. The prompt is then built from the
already-fetched card rather than re-fetching it.
Invoking "Enrich Metadata with AI" still always calls the provider; a
model source supplying a description, images and tags is not treated as
a reason to skip it, since the LLM's summary and notes are richer and an
action that silently does not call out to the provider would be
unpredictable. The site data acts as a fallback for the gaps the LLM
leaves.
Add `base_model_resolver.resolve_base_model()` to map the site's own
names (`krea/Krea-2-Turbo`, `KREA_2_TURBO`) onto the canonical
vocabulary, used only when the LLM returns no base model. It is strictly
conservative — exact normalised matching plus a bounded set of variant
suffixes, and it only ever returns a name that is already in the
vocabulary — so an uncertain hint defers to the LLM instead of writing a
plausible-looking wrong value.
ModelScope became a linkable source, but downloading from it was impossible:
the URL picker only recognised huggingface.co, the file listing hit a
huggingface-only endpoint, the resolve URL was hardcoded, and the default
path template always wrote into a `huggingface/` directory.
Move the download knowledge into the providers so the handlers stay generic:
- `ModelSource` gains `list_files()`, `file_download_url()`,
`default_revision` and `default_subdir`. `HuggingFaceSource` keeps the Hub
tree API (`/api/models/{id}/tree/{rev}`, LFS-aware sizes, `main`).
`ModelScopeSource` uses `/api/v1/models/{id}/repo/files?Revision=master`
— which reports real byte sizes for LFS files, so no HEAD probe is needed,
and which only accepts `master` (an HF-imported repo still 404s on `main`)
— and downloads through `/models/{id}/resolve/{rev}/{path}`. That URL
redirects to a CDN target carrying a time-limited `auth_key`, so it is
rebuilt on every request and never cached, which is also what keeps
resumable Range requests working.
- `hf_handlers.py`/`HfHandler` become `model_source_handlers.py`/
`ModelSourceHandler` with `list_model_source_files` and
`download_model_source`. New routes `/api/lm/model-source-files` and
`/api/lm/download-model-source`; the old `/api/lm/hf-repo-files` and
`/api/lm/download-hf-model` paths stay as aliases, and a payload without
`platform` still means Hugging Face, so existing callers are unaffected.
- A downloaded sidecar now records `source_platform` + `source_url` (with the
`hf_url` alias only for Hugging Face) instead of always writing `hf_url`,
and `use_default_paths` files ModelScope downloads under
`modelscope/<owner>/<repo>`. The now-unused shared HF aiohttp session and
its shutdown hook are gone; providers open short-lived sessions.
- Frontend: `detectUrlType` returns the platform-neutral
`model-source-repo` / `model-source-file` plus an explicit `platform`, the
DownloadManager's `hf*` state and methods are renamed to `source*`, every
`source === 'huggingface'` check becomes `isExternalModelSource()`, and
batch groups are keyed by `platform:repo` so the same `owner/name` on two
sites renders as two groups. A bare `owner/name` still means Hugging Face.
- `is_valid_source_id()` centralises repo-id validation (exactly
`owner/name`, no traversal, no leading dot). This also fixes the old HF
download check that rejected any dot in the name, i.e. legitimate repos
such as `black-forest-labs/FLUX.1-dev`.
Verified against the live APIs: the example repo lists 8 weight files with
correct sizes, and a ranged GET of the built resolve URL returns 206 after
following the redirect to the CDN. Backend 2853 passed; frontend 1143 JS +
91 Vue passed. The nine locales carry the refreshed download copy in the
next commit.
Complete the 15 [TODO: Translate] keys the model-source feature left behind
(modelCard.actions.viewOnSource, loras.contextMenu.linkModelSource,
modals.linkModelSource.*, modals.model.versions.sourceGroupInfo,
toast.contextMenu.enrichNeedsSource, toast.contextMenu.enrichUnsupportedSource),
and refresh the two enrichment labels that feature made stale.
- Brands stay Latin per R3: Hugging Face / ModelScope / TensorArt appear
verbatim, and {source} is substituted by the caller at runtime, so no locale
embeds a transliterated platform name. The placeholder-URL value
(modals.linkModelSource.urlPlaceholder) stays byte-identical to en.json per
the §6 URL exception.
- "model source" / "model page" / "model card" are new nouns and each locale
gets exactly one rendering; "AI enrichment" reuses the noun already in each
file from the previous enrichHfAgent copy. All of it is recorded in §2.
- modelCard.actions.viewOnSource follows each locale's existing
viewOnHuggingFace pattern rather than the neighbouring viewOnCivitai one, so
de/ru/he/ja/ko do not gain a third "View on ..." shape.
- loras.contextMenu.enrichHfAgent and loras.bulkOperations.enrichHfAgent read
"AI HF metadata" in all nine locales. The feature invalidated that by also
covering ModelScope, so both values drop the HF qualifier (the key names keep
the historical Hf, and the guidelines now say so).
- Script conventions: fr keeps ASCII apostrophes and a space before ':' (the
file is 351 ASCII vs 26 U+2019 and the modal being replaced was ASCII); ko
keeps ASCII ':' and '()' (188 vs 6); CJK locales keep full-width punctuation;
every ellipsis is ASCII '...'. Placeholders are verbatim per R2.
- modals.linkModelSource.enrichNote is phrased as a rule with the current
exception in parentheses, so the guidelines call that out for whoever adds
the next link-only source.
pytest tests/i18n: 20 passed, and scripts/sync_translation_keys.py --dry-run is
a no-op (no missing and no stale keys). Frontend: 1130 JS + 91 Vue passed.
Backend: 2815 passed. en.json is untouched by this commit.
A model file could only ever be linked to huggingface.co: `set_hf_url`
validated the URL with a huggingface-only regex, the agent fetched the card
from a hardcoded HF URL, and the readme processor built every relative image
path off `https://huggingface.co/{repo}/resolve/main`. ModelScope publishes the
same model-card convention (README.md + YAML frontmatter, often carrying
`base_model:` and `trigger_words:`) behind a public, key-less API, so the
enrichment pipeline could already serve it - it was the plumbing that was
HF-shaped, not the idea.
Make the external source a first-class, provider-driven concept:
- New `py/services/model_sources/` registry. A `ModelSource` owns URL
recognition (lenient for stored values, strict for user input), the
canonical page URL, model-card fetching, the asset base URL and the
capability flags. `HuggingFaceSource` is the previous logic relocated;
`ModelScopeSource` reads `/models/{o}/{n}/resolve/{master|main}/README.md`
and falls back to `/api/v1/models/{o}/{n}/repo`. `TensorArtSource` is
link-only on purpose: tensor.art answers plain HTTP clients with a
Cloudflare challenge and its internal API (ap-east-1.tensorart.cloud /
cn.tensorart.net) rejects every /v1/model/* route with "invalid
authorization header", so it declares supports_enrichment=False rather than
failing silently later.
- Metadata gains `source_platform` + `source_url`; `hf_url` stays as a
read/write alias, written only for Hugging Face, so existing sidecars,
cached rows and third-party consumers keep working. Normalisation runs at
the scanner, the persistent cache (both directions, plus two new columns
behind an ALTER migration) and the linking handler - which is what stops a
user who switches sources from leaving a stale `hf_url` on a ModelScope
model.
- The agent pipeline keys off the provider instead of `hf_url`: the fast-fail
gate now explains *why* a model is skipped (no source / unknown source /
source without a reachable card), the prompt context exposes
source_url/source_id/source_label/asset_base_url while still filling the
legacy hf_url/repo aliases, and the four README image extractors take a
base_url (defaulting to HF) so relative paths resolve against the right
site. Version grouping generalises to hf: / ms: / ta: keys.
- `POST /api/lm/set-hf-url` keeps its path and its legacy payload keys but
accepts `source_url`, validates against every provider and returns the
platform. `GET /api/lm/model-sources` lets the UI render the supported-site
list from the server.
- Frontend: a `modelSourceHelpers` mirror of the registry drives the link
dialog, the card/modal globe (branded "View on ModelScope/TensorArt"), the
version-group key and the enrichment gate; the versions tab no longer sends
ms:/ta: keys to the CivitAI API.
TensorArt stays in the list because provenance is worth keeping even when the
card is unreadable - the dialog says so plainly ("Sites that don't expose one
(currently TensorArt) can only be linked") and the context menu disables
enrichment with a matching tooltip, instead of the user getting
"Unsupported URL".
Verified against the real ModelScope API: jj3550945163/Krea-2-LORA returns a
1882-byte card whose frontmatter carries base_model/tags/trigger_words, and
relative images resolve to .../resolve/master/....
Tests: backend 2815 passed; frontend 1130 JS + 91 Vue passed; pytest
tests/i18n and a Jinja compile pass over templates/. The nine locales carry
[TODO: Translate] for the new strings, completed in the next commit.
Other Models management is opt-in and its folders come from
folder_paths.get_folder_paths(). In plugin mode ComfyUI registers vae,
upscale_models, text_encoders, clip_vision and controlnet out of the box, so
enabling the feature works immediately. Standalone only knows the keys present
in settings.json.folder_paths, and that file is edited by hand - there is no UI
for those keys - so a standalone user who followed the announcement banner
reached "Enable Other Models" and then an empty page.
Gate the announcement on the capability instead of on how the process was
started:
- Config.get_other_models_availability() probes every canonical other key
(legacy clip collapses into text_encoders where the host exposes
map_legacy) and reports which sub_types resolve to a folder that exists on
disk. It deliberately ignores enable_other_models: the question is "could
this work here at all?". An empty folder counts, because CivitAI downloads
can target it.
- /api/lm/settings exposes it as the derived, non-persisted
other_models_paths_available flag; a probe failure yields null and the
banner fails open.
- BannerService only registers the announcement when the flag is not false.
`=== false` (not falsy) keeps a cached/older payload working, and nothing is
written to dismissed_banners, so the banner can return once folders exist.
- The Other page grows an "enabled but nothing to scan" empty state driven by
config.other_roots, showing the settings.json snippet for standalone and a
pointer to ComfyUI model paths otherwise, plus an Open Settings action. It
also covers the corner where only a non-default sub_type has a folder.
Translate the six other.noPaths.* keys into all nine locales and record the
new "folder key" / "on disk" terminology in the i18n guidelines.
Backend tests and pytest tests/i18n could not run in this environment (no
pytest/platformdirs); the probe was exercised against a stubbed folder_paths.
Frontend: 120 files / 1101 JS tests passed.
Three pre-enable strings listed exactly the old default set (VAE, upscaler,
text encoder, CLIP vision), so they read as "these are what enabling
manages" - now wrong twice over, since clip_vision became opt-in and
ControlNet was never named.
Point them at the capability instead: other.disabled.description and
banners.otherModels.content enumerate all five sub_types, and
settings.folderSettings.enableOtherModelsHelp names all five folder
categories the master switch gates. Model-type names stay in Latin per the
model-type rule; de compounds as CLIP-Vision- und ControlNet-Ordner and the
slash-list locales keep their existing VAE / Upscaler / Text Encoder / ...
casing. No placeholders or HTML are involved.
Editing en.json leaves the nine locales stale, and the sync script only adds
missing keys, so each locale is updated in the same pass by exact-literal
replacement of the one line - no JSON round-trip, no formatting churn (three
changed lines per file). Record the refreshed strings and the
"capability, not defaults" rule in the i18n guidelines.
DEFAULT_ENABLED_OTHER_SUB_TYPES managed vae, upscaler, text_encoder and
clip_vision while controlnet was the sole opt-in type. That split was not
defensible on demand breadth: ControlNet is the broader category by install
base, and clip_vision is the narrower one (IPAdapter/SVD image conditioning,
usually one to three files) whose CivitAI type is retired upstream.
Keep the default set to the dependency-style assets every pipeline needs and
where "which one am I actually using" is the real problem - VAE, upscalers
and text encoders - and treat clip_vision and controlnet symmetrically as
opt-in. The feature is still unreleased, so the change needs no migration.
- Sync all five surfaces holding a default: DEFAULT_ENABLED_OTHER_SUB_TYPES,
DEFAULT_SETTINGS, both DEFAULT_SETTINGS_BASE/createDefaultSettings lists,
updateOtherModelsControls()'s fallback and the Jinja fallback.
- The selection is persisted per user, so only the untouched default moves;
existing default_other_roots entries for a disabled sub_type are preserved.
- Fix the Jinja fallback using `or`, which treated an all-unchecked empty
allow-list as "unset" and re-checked every box on render; `is none` keeps
the empty list empty.
- Document the revised defaults and rationale in the plan.
Tests assert the new default trio, the normalize fallback, that both opt-in
types stay out of the default scan, and the auto-set iteration test now
enables clip_vision explicitly since it exercises the loop, not the default.
get_download_path_template() fell back to "{base_model}/{first_tag}" for any
unconfigured model type, so other-model downloads were silently nested under an
arbitrary CivitAI tag even though the settings UI exposes no template row for
"other" and priority_tags has no "other" entry (making {first_tag} resolve to
tags[0]).
Add DEFAULT_DOWNLOAD_PATH_TEMPLATES with other -> "" so unconfigured and
unknown types resolve to a flat layout under the already sub_type-scoped
default_other_roots; explicit settings.json values still win. Mirror the flat
default in the frontend DEFAULT_PATH_TEMPLATES and stop the download/move
default-path previews from rendering "/undefined" or a dangling slash.
Complete the 36 keys left as [TODO: Translate] by the Other Models
feature (VAE / Upscaler / Text Encoder / CLIP Vision / ControlNet
management page and its opt-in toggles): settings.folderSettings.*,
other.*, initialization.other.*, toast.settings.otherRootsFailed and
banners.otherModels.*.
Model-type names (VAE, Upscaler, Text Encoder, CLIP Vision, ControlNet)
stay in Latin per the model-type rule, so the five subType* values are
intentionally identical to en.json; "Other Models" is a page/feature
name and is translated. Document the new terminology in the i18n
translation guidelines and note the completed i18n phase in the plan.
Documents the settings keys and defaults, the enabled/disabled behaviour
matrix, the backend and frontend touch points, cache consistency, the
discoverability surfaces (hidden nav + announcement banner + download CTA)
and the minimal settings.json.example policy.
The Windows-only case-insensitive match in ModelScanner._reconcile_cache
is the only pass left unverified by the recent realpath cleanup: realpath
may already cover case differences on Windows, and if the branch is ever
reachable it is O(files x cache entries). Records the reachability
question, the verification steps for a Windows run, and the two possible
fixes.
Recipes imported from CivitAI image URLs can contain 0 LoRAs: the backend
only sees the REST image API + EXIF, while the complete generation data
lives in the image page's internal trpc payload (see
docs/recipe-civitai-image-no-metadata.md). When the companion
lm-civitai-extension is installed with a valid license, re-import (single
and bulk) of CivitAI-image-sourced recipes is now delegated to the
extension via DOM CustomEvents; the extension scrapes the image page with
the user's session and calls back into the reimport endpoint with the
full metadata payload. Without the extension (or with an invalid license)
the native path runs unchanged.
- POST /api/lm/recipe/{id}/reimport accepts optional payload params
(image_url/name/resources/gen_params/base_model/tags); the payload path
reuses the import-remote engine with reimport semantics (user-edit
carryover, delete-after-save), and malformed/failed payloads fall back
to the legacy URL import. Response gains loras_count.
- The endpoint also accepts GET: the extension is GET-only by convention
(documented in AGENTS.md).
- New static/js/utils/extensionReimportBridge.js (probeExtension /
delegateReimport / getCivitaiImageInfo) wired into RecipeContextMenu
and BulkManager with silent native fallback.
- i18n: toast.recipes.reimportingViaExtension added and translated in
all 9 locales.
The recipe "Repair Metadata" action has been marked Deprecated in the UI
for a while and cannot reliably recover recipes imported from CivitAI URLs
whose REST meta has no resources/hashes and whose image has no embedded
metadata (e.g. CivitAI-only generation data). Drop the feature end to end.
Backend:
- remove repair routes (repair, cancel-repair, recipe/{id}/repair,
repair-bulk, repair-progress) and their handler mappings/methods
- remove RecipeScanner repair_all_recipes / repair_recipe_by_id /
_repair_single_recipe and REPAIR_VERSION
- remove WebSocketManager recipe-repair progress channel
- drop repair_version column from the persistent recipe cache
- rematch mutual-exclusion now only checks rematch
Frontend:
- remove repair entries from per-recipe, bulk and global context menus
- remove repairRecipe / repairSelectedRecipes / repairRecipes + cancelRepair
and the repairBulk API client method/endpoint
- drop recipe-repair i18n keys (synced across locales; doctor keys kept)
Tests/docs: delete test_recipe_repair.py, update scaffolding/routes/ws/
persistent-cache/integration tests and i18n guideline examples.
R1 instructed agents to "translate the newly added keys in every locale"
right after syncing, while R8 and §7 make [TODO: Translate] placeholders
the sanctioned end state during feature development until the feature
owner explicitly asks for translations. Reword R1 and the AGENTS.md
Localization section to say stop after syncing and never translate
proactively.
Enhance the deleted-LoRA reconnect flow in the recipe modal:
- Suggest local reconnect candidates when the panel opens, ranked by
identity (same hash / same CivitAI version) then filename/name
similarity, with a hard filter on confident base-model mismatches;
the input gets a Combobox backed by the same endpoint as you type.
- Snapshot the pre-reconnect entry and offer a permanent restore:
reconnected entries show an undo icon at the right end of the info
row, with the original filename in the tooltip.
- Relax the manual reconnect base-model guard to a three-tier check:
exact/unknown labels pass silently, same-architecture families
(e.g. Pony <-> Illustrious) pass with a warning toast, and only
cross-architecture mismatches stay hard-rejected.
[TODO: Translate] placeholders are now the sanctioned intermediate state
during feature development; translate all pending keys in one pass only
when the feature owner asks. R8 notes the exemption so placeholders are
not 'fixed' prematurely.
§3/§5/§6 now describe the resolved state (regression watch-list instead of a
to-do list), §4 documents the single intentional placeholder deviation
(mappingsUpdated drops {plural} where '<noun>s' cannot be appended). de
help.updateVlogs.playlistTitle translated.
Audit of all 10 locale files found recipe/checkpoint mistranslations,
inverted ko tag logic, stale help texts, placeholder contract deviations,
and untranslated feature blocks. Document the conventions (R1-R9), per-
language term maps, confusion hot-spots, and the translation workflow so
future agents and translators follow the established decisions (e.g. keep
'Recipe' untranslated in French, use 配方 in Chinese).
Phase 2 of docs/plans/issue-1085-rate-limit-design.md:
- Batch import: items that fail due to vendor rate limiting are now
SKIPPED with a "re-run the import later" hint instead of FAILED, so a
transient 429 no longer pollutes failure accounting; the progress
broadcast carries a rate_limited flag.
- Batch import UI: show a one-time "rate limited — slowing down" toast
and swap the running status text while rate_limited; i18n keys synced
to all locales.
- Downloader: download_file / download_to_memory / get_response_headers
register 429 cooldowns with the RateLimitCoordinator, so subsequent
API calls queue behind a download-triggered rate-limit window.
Implement Phase 1 of docs/plans/issue-1085-rate-limit-design.md:
- New RateLimitCoordinator: per-host shared Retry-After gate with
exponential backoff (30s base, 1800s cap), minimum inter-request pacing
(default 0.75s), herd-free waiter serialization via per-destination
locks, and a bounded wait (default 300s) that raises instead of parking.
- Downloader.make_request: connectivity-guard fail-fast first, then gate
pacing; on 429 register the cooldown and wait-and-resend (bounded);
errors that passed through the gate are marked gate_handled.
- FallbackMetadataProvider / MetadataSyncService: a network provider 429
no longer fails over to other network providers (stops the CivArchive
flood); sqlite stays as local last resort. Rate-limited lookups now
report "Rate limited" instead of "Model not found", so transient 429s
no longer mark models civitai_deleted.
- _RateLimitRetryHelper skips its own sleep for gate_handled errors,
removing the double wait.
- New settings: rate_limit_gate_enabled, rate_limit_max_wait_seconds,
rate_limit_min_interval_seconds.
- Three-state autov3 field (not-checked / checked-unavailable / 12-hex value)
in .metadata.json sidecars, in-memory ModelHashIndex, and SQLite
(models.autov3 column + autov3_index table) with column-presence migration
- Background self-terminating backfill for legacy rows: per-model-type
concurrency guard, executor-offloaded I/O, Civitai-first resolution
(SHA256-matched version file) falling back to the embedded safetensors
header hash
- Civitai-first propagation on metadata refresh, scan, and download paths;
reject the empty-string SHA256 placeholder and strip OneTrainer 0x prefix
- List API hash filters and hash index lookups accept 12-char AutoV3
- Cap safetensors header reads at 64 MiB to prevent crafted-file allocation
- Prevent stale AutoV3 mappings on file replacement while preserving them on
same-file re-registration (lazy-hash completion)
- 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
Introduce an agent skill framework for LLM-driven metadata enrichment:
- AgentCLI (py/agent_cli/): in-process wrappers around internal services
using standard relative imports, eliminating the need for sys.path hacks
- LLMService: centralized BYOK (bring-your-own-key) LLM client supporting
OpenAI, Ollama, and custom OpenAI-compatible endpoints
- PostProcessor: deterministic engine that applies LLM output via AgentCLI
(replaces old handler.py + _BASE_MODEL_ALIASES approach)
- SkillRegistry: filesystem-based skill discovery (skill.yaml + prompt.md)
- AgentService: orchestrates skill execution with WebSocket progress
- Frontend AgentManager: WebSocket listeners, skill execution, config UI
- Context menu entries (single + bulk) for "Enrich Metadata (Agent)"
- Settings UI for AI Provider configuration (BYOK)
- Full i18n support across 9 locales
Bug fixes found during review:
- aiohttp.web.json_response: status_code= -> status=
- settings_modal cancelEditApiKey: wrong argument position
- AgentManager.isLlmConfigured: allow Ollama without API key
- PostProcessor._merge_tags: lowercase all tags to match TagUpdateService
Moved wiki-images to the wiki repo (willmiao/ComfyUI-Lora-Manager.wiki). Updated README.md image reference to use wiki raw URL. Removed docs/LM-Extension-Wiki.md (superseded by wiki pages).
- Delete examples/metadata/ directory and all example files
- Real metadata.json files in model roots are better examples
- Examples were artificial and could become outdated
- Maintenance burden outweighs benefit
- Remove 'Complete Examples' section from docs/metadata-json-schema.md
- Remove reference to example files in 'See Also' section
Rationale:
Users have access to real-world metadata.json files in their actual
model directories, which contain complete Civitai API responses with
authentic data structures (images arrays with prompts, files with hashes,
creator information, etc.). These are more valuable than simplified
artificial examples.
- Create docs/metadata-json-schema.md with complete field reference
- All base fields for LoRA, Checkpoint, and Embedding models
- Complete civitai object structure with Used vs Stored field classification
- Model-level fields (allowCommercialUse, allowDerivatives, etc.)
- Creator fields (username, image)
- customImages structure with actual field names and types
- Field behavior categories (Auto-Updated, Set Once, User-Editable)
- Add .specs/metadata.schema.json for programmatic validation
- JSON Schema draft-07 format
- oneOf schemas for each model type
- Definitions for civitaiObject and usageTips
- Add example metadata files for each model type
- lora-civitai.json: LoRA with full Civitai data
- lora-custom.json: User-defined LoRA with trigger words
- lora-no-triggerwords.json: LoRA without trigger words
- checkpoint-civitai.json: Checkpoint from Civitai
- embedding-custom.json: Custom embedding
Key clarifications:
- modified: Import timestamp (Set Once, never changes after import)
- size: File size at import time (Set Once)
- base_model: Optional with actual values (SDXL 1.0, Flux.1 D, etc.)
- model_type: Used in metadata.json (not sub_type which is internal)
- allowCommercialUse: ["Image", "Video", "RentCivit", "Rent"]
- civitai.files/images: Marked as Used by Lora Manager
- User-editable fields clearly documented (model_name, tags, etc.)
- Reframe supporter access section to emphasize sustainability and gratitude
- Add CivArchive support announcement and image
- Document new dedicated download button and hide models feature in v0.4.8
- Improve readability and flow of the overview and supporter sections
- Remove backward compatibility code for `model_type` in `ModelScanner._build_cache_entry()`
- Update `CheckpointScanner` to only handle `sub_type` in `adjust_metadata()` and `adjust_cached_entry()`
- Delete deprecated aliases `resolve_civitai_model_type` and `normalize_civitai_model_type` from `model_query.py`
- Update frontend components (`RecipeModal.js`, `ModelCard.js`, etc.) to use `sub_type` instead of `model_type`
- Update API response format to return only `sub_type`, removing `model_type` from service responses
- Revise technical documentation to mark Phase 5 as completed and remove outdated TODO items
All cleanup tasks for the model type refactoring are now complete, ensuring consistent use of `sub_type` across the codebase.
This commit resolves the semantic confusion around the model_type field by
clearly distinguishing between:
- scanner_type: architecture-level (lora/checkpoint/embedding)
- sub_type: business-level subtype (lora/locon/dora/checkpoint/diffusion_model/embedding)
Backend Changes:
- Rename model_type to sub_type in CheckpointMetadata and EmbeddingMetadata
- Add resolve_sub_type() and normalize_sub_type() in model_query.py
- Update checkpoint_scanner to use _resolve_sub_type()
- Update service format_response to include both sub_type and model_type
- Add VALID_*_SUB_TYPES constants with backward compatible aliases
Frontend Changes:
- Add MODEL_SUBTYPE_DISPLAY_NAMES constants
- Keep MODEL_TYPE_DISPLAY_NAMES as backward compatible alias
Testing:
- Add 43 new tests covering sub_type resolution and API response
Documentation:
- Add refactoring todo document to docs/technical/
BREAKING CHANGE: None - full backward compatibility maintained
- Restructure document to clearly separate simple vs complex widget patterns
- Add detailed explanation of ComfyUI's built-in callback mechanism
- Provide complete implementation examples for both patterns
- Remove outdated sync chain diagrams and replace with practical guidance
- Emphasize using DOM element as source of truth for simple widgets
- Document proper use of internal state with widget.callback for complex widgets
Will Miao