- Dual sign-off is now the default (thesis_required_approvals defaults to 2).
- Entity-merge review queue (migration 0003): the fuzzy/Qwen tier no longer
auto-merges — it writes CANDIDATES (entity_merge_candidates) with a same/different
suggestion + confidence + reason for a human to approve (merge) or reject (keep
separate). entity_merge.py applies/rejects (durable via entity_merges, soft-delete,
repoint links+edges); decided pairs aren't re-surfaced.
- entity_jobs.py: UI-triggered background index jobs (rebuild/update/find-duplicates)
as subprocesses with a one-at-a-time lock; status in /api/system/status.
- server.py: /api/index/{rebuild,update}, /api/entities/find-duplicates,
/api/entities/merge-candidates [+ /{id} decide] — admin-gated.
- docs/thesis-seed-v2.md: concrete, plain-English rewrite per Grant's feedback.
Backend verified end-to-end on synthetic data (candidate gen -> approve/reject).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Fuzzy tier (backend/ingest/fuzzy_resolve.py + llm.py): local Qwen adjudicates
the deterministic resolver's flagged name-variant candidates; merges are
durable via entity_merges (deterministic re-runs respect them), losers
soft-deleted, logged. Idempotent.
- Incremental sync (backend/ingest/sync.py): re-embeds only rows changed since a
watermark (ingest_sync_state); first run / --recreate = full. Tested full→0→1.
- Start9 packaging (start9/0.4): Dockerfile bundles ingest+mcp + fastembed/mcp;
"Build search index" action runs the init in a subcontainer; MCP shipped as a
manual stdio server (not a daemon); version 0.1.0:44. INGEST_PACKAGING.md.
- backfill.py: factored embed_and_upsert() shared with sync.
Verified end-to-end on synthetic data + live Sparks/Qwen/Qdrant.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Keysat