mirror of
https://github.com/lukaszraczylo/claude-mnemonic.git
synced 2026-10-14 05:30:11 +00:00
0845aeb58a
* fix(db): keep archived notes out of search, lists, counts and the vector rebuild An archived note is meant to be kept but unused, yet most read paths did not filter on is_archived: full-text and LIKE search, vector-hit hydration, context injection, the paginated lists, counts and the vector rebuild all still returned it. A vector hit for an archived note also still pointed at a live embedding, because archiving never dropped it. - add notArchivedFilter/visibleScope and apply them to every read path - the paginated lists take includeArchived; the API reads ?include_archived=true, and exports ask for them so a backup is complete - archiving drops the note's vectors, unarchiving syncs them back - add //go:build fts5 to observation_update_test.go so `go vet` without tags builds (it referenced a helper from a tagged file) Adds a store test that archives one of three notes and checks every path, and that fails if the filter is removed. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> * fix(worker): drop the vectors of notes archived through bulk-status too POST /api/observations/bulk-status with action "archive" archived notes but left their vectors in the index; only the dedicated archive handler dropped them. Both paths now call dropArchivedFromVectors. Co-Authored-By: Claude Sonnet 5.5 <[email protected]> --------- Co-authored-by: Claude Sonnet 5.5 <[email protected]>