Share auto-import planning between preview and write paths, expose a machine-readable CLI result envelope, and preserve explicit parser options during analysis.
Route Desktop JSON dry-runs through a shared read-only push analyzer so validation, deduplication, and member inference match actual writes. Keep JSONL on the file analyzer path and cover new and incremental predictions without database mutations.
Use the long-running worker timeout and refresh Desktop caches after successful push imports.
Align auto-created member metadata and counts across initial and incremental imports, exclude SYSTEM from member updates, and keep existing-session writes validated and atomic.
Thread ownerId and member aliases through the TS and native parsers. Persist aliases across full and incremental imports so exported ChatLab JSON retains user identity metadata.
Remove persisted annual summary snapshots whenever a session is deleted through shared or desktop cleanup paths. Drop cached in-memory snapshots when their backing files disappear so stale responses cannot expose deleted session data.
The UI stayed on "parsing messages" during the whole post-parse phase:
the importer emitted an 'importing' stage the frontend had no text for,
and on CLI Web the synchronous DB writes / index / FTS build starved the
event loop so buffered SSE progress events only arrived after the import
finished.
- Replace 'importing' with 'saving' (message writes) and a new 'indexing'
stage (indexes / FTS / session index) in ParseProgress, with UI texts
for all four locales.
- Yield a macrotask between native parser batches and before heavy
synchronous steps so SSE progress events flush while importing.
Co-authored-by: Cursor <cursoragent@cursor.com>
getJieba() invalidated its cache whenever no dict file existed on disk,
so users without a custom dict rebuilt a Jieba instance per tokenize call.
FTS indexing tokenizes every message, making imports seconds slower
(43MB sample: 9.2s -> 3.5s after the fix). Cache now only invalidates
when the on-disk dict state actually changes (added or removed).
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(semantic-index): stabilize embedding warmup
Improve warmup job reliability and prevent raw API response body from
reaching app.log. Status code alone is sufficient for diagnosing failures;
stripping the response detail eliminates the risk of chat content being
written to the log via OpenAI-compatible services that echo submitted input.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(semantic-index): clear preload error after warmup retry
When local model preload fails, a later warmup can successfully recreate the extractor through embedDocuments but leave modelPreloadStatus stuck on error.
Mark the local model ready after a completed warmup writes chunks, and cover the retry path with a regression test.
* fix(semantic-index): guard warmup model status updates
Capture the model configuration used by a warmup job before it starts, and only clear local preload errors when the current local model still matches that job.
Add a regression test for switching local models while an older warmup is still completing.
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Wait for initial contact rows before scrolling to groupmates so tab navigation does not land on temporary rows. Avoid writing contacts session facts or latest facts when a session DB fingerprint changes during computation.
digua
