Bulk delete merged staged batches by physically moving each loser's
files into the winner's batch dir with os.rename. Cross-volume bulks
(winner and loser on different filesystems) always hit EXDEV, forcing a
rollback and degrading to the batch_ids array with per-batch undo.
Merge is now manifest-only: loser entries are appended to the winner's
manifest with their staged paths unchanged, so staged files keep living
in each model's own .lm-pending-delete/<batch_id> dir (no data IO, no
EXDEV). Loser dirs are recorded in the winner manifest's merged_sources
and each loser manifest is stamped merged_into so its own purge timer, a
post-restart sweep or a direct undo call no-op. A cross-volume bulk is
one undoable batch again, and undo/purge clean up the loser dirs once
the merged batch settles.
Stop resetting the whole listing after a successful download. The legacy
flow reloaded page 1, scrolled to the top and hijacked the sidebar's
active folder whenever a custom target folder was used, which made the
Updates view lose its place (and sometimes render as an empty page).
Downloads only flip the update flag for one model, so the listing is now
reconciled in place through the virtual scroller:
- Updates view: the model's cards are removed once its newest eligible
version is installed (the flag is model-level).
- Normal listing: the card stays; only update_available is cleared.
- Model not in the current view (different folder/filter/window):
no-op; the sidebar folder tree alone is refreshed.
- Falling back to the legacy reload only when no virtual scroller is
available (e.g. recipes page, duplicates mode, HF downloads).
- Add a 'when to use / when not to use' gate: UI behavior questions
default to Vitest/jsdom, E2E only for behavior spanning server +
browser; description updated so the skill triggers less eagerly
- Pin the browser driver to Chrome DevTools MCP and explain why
kimi-webbridge (user's real browser) is not a substitute
- Drop generic MCP pattern boilerplate duplicated by
references/mcp-cheatsheet.md (SKILL.md 385 -> 145 lines)
- Move recipe rematch fixture / fresh-state / cancel-gap notes to
references/recipe-rematch-fixtures.md
- Extract a shared BaseModelPicker (search, keyboard navigation,
filename-based suggestions, dynamic API models such as MiniMax H3
under 'Other (API)') used by both the single-model metadata modal
and the bulk base model modal
- Rework the bulk base model modal into a dedicated inline-list
layout: fixed modal size, sticky-free footer with app-standard
modal-actions/primary-btn/cancel-btn buttons, and an inline option
list that scrolls itself instead of an overlay dropdown covering
the footer
- Selecting an option in change mode now filters the list to the
selection instead of resetting and scroll-jumping to it
- Restore opaque sticky section headers in the bulk modal so scrolled
items no longer bleed through
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.
Address the rate-limit flood and secondary errors seen during large
recipe ingestion (example-images directory import):
- batch import: share one adaptive-concurrency semaphore across the whole
batch (previously each item got a fresh semaphore, so the min/max
concurrency bounds never applied and every item ran concurrently);
synchronize the shared semaphore capacity after each completed item.
- comfy parser: guard ckpt_name against list/None values so re.search no
longer raises TypeError and fails the whole image import.
- civarchive client: normalize empty-string failure payloads to
"Request failed" and treat a missing payload as an error, fixing the
"'NoneType' object has no attribute 'get'" crash.
- civarchive client: log connectivity-guard offline-cooldown
short-circuits at DEBUG instead of one ERROR per request.
Add LORA_MANAGER_SETTINGS_DIR env var and standalone --settings-path to pin
the settings location (settings.json, cache/, wildcards/, backups/, logs/,
stats/) to an arbitrary directory. The override takes precedence over
portable mode and the platform user config dir, and skips legacy migration,
so sandboxed dev/E2E runs no longer need to write settings.json in the repo
root or collide with the real instance.
standalone.py pre-scans argv for --settings-path at import time because the
settings location is resolved before main() parses arguments. SettingsManager
portable-switch migration is a no-op while the directory is pinned.
Update the lora-manager-e2e skill (prefer --settings-path sandboxing;
start_server.py passes it through) and the lora-manager-runtime-context
skill (document precedence; inspect script honors the override).
The browser extension's apiFetch treats any 404 as a missing endpoint and
retries the legacy non-/api/lm URL, producing two spurious
'error_middleware - WARNING - API GET ... 404' log lines per occurrence.
- complete_download_in_queue / update_download_queue_status /
retry_download_from_history: 'not found' is a normal business outcome,
return 200 + success:false instead of 404 (extension behavior unchanged;
apiGet ignores the HTTP status)
- error_middleware: downgrade /api/lm/download-progress/ 404s to debug like
previews - the 404 status itself stays (extension uses it for failure
detection), only the log level is lowered
Old workflows (saved before the control_after_generate feature) carry a
shorter widgets_values array. The frontend's index-based widget restore
then shifts the old weight_dtype value into the hidden control widget
(leaving an invalid value like 'default') and silently resets
weight_dtype to its default. On graph load, hand the shifted value back
to weight_dtype when it still sits at its default, then reset the
control mode to 'fixed' so old workflows keep loading deterministically.
- Log expected "GID not found" tellStatus probes at DEBUG, and treat a
forgotten GID as permanent so the poll loop recovers immediately
instead of burning 4 retries x 3s of ERROR lines per cycle
- cancel_download tolerates a forgotten GID and always pops the
in-memory transfer so concurrent polls cannot re-register a
cancelled download
- Restore sweep deletes aria2 state records with no resolvable target
path instead of skipping them forever
- Clearing the download queue now also cancels in-memory tasks, removes
live aria2 transfers and drops persisted state for the cleared ids
(partial files on disk are preserved)
After a drag move empties the selected folder, refresh() resets the
stale activeFolder to root but the grid kept showing the old filtered
(empty) view until a manual reload. Trigger resetAndReload when the
fallback happens post-initialization; the initial page load is untouched
because it picks up the cleared filter on its own.
initialization.js falls back to polling /api/lm/init-status when the
/ws/init-progress WebSocket cannot be established, but no route ever
registered that path — each poll 404'd and the page never reloaded after
the scan completed. Report the aggregate status of all four scanners and
omit pageType so every initialization page accepts the update.
restoreSelectedFolder trusted localStorage blindly: a stale activeFolder
(moved/deleted, or saved while the tree was still empty) left the grid
filtered to a nonexistent folder with a phantom breadcrumb and no way to
recover short of clicking the root breadcrumb. Validate the persisted
path against the freshly loaded tree and reset to root when it is gone;
skip validation when the tree load failed so transient errors don't wipe
the saved location.
The recipes page always rendered with is_initializing=False, so a cold
start displayed an empty grid that never updated until a manual refresh.
Mirror the model pages: gate render_page on the scanner state, broadcast
init progress from RecipeScanner (including a completion message, and a
failure fallback so the page never stalls), and teach initialization.js
to detect the /loras/recipes page before the generic /loras match.
get_cached_data() claimed to wait for a running initialization but
actually returned the placeholder empty cache, so API requests during
startup saw zero recipes. The initializing flag was also set only after
the LoRA scanner wait, leaving an unguarded window. Mark initialization
before the first await and have callers await the in-flight task.
- Add Tag Autocomplete ON/OFF entry to the Prompt (LoraManager) node
right-click menu, cross-referencing the slash commands
- Show the current autocomplete state (/autocomplete or /noautocomplete
hint) below the slash command list
- Show a one-time dismissible tip in the suggestion dropdown on first use
- Clarify toggle command labels (Turn autocomplete ON/OFF) and cross-link
all three entry points in the settings tooltip
- Share the setting write path via setLoraManagerSettingValue()
With trust_env=True, aiohttp auto-loads credentials from ~/.netrc (e.g. a
'machine civitai.red' or 'default' entry) and refuses to combine them with
the explicit Authorization: Bearer header, aborting every authenticated
CivitAI request with 'Cannot combine AUTHORIZATION header with AUTH
argument or credentials encoded in URL'.
Will Miao