Fix live multi-turn: pass the resolved drug, stop did-you-mean garbage

This commit is contained in:
2026-08-05 16:54:35 +07:00
parent ef08b4929e
commit 1e8cbdb586
29 changed files with 2013 additions and 83 deletions
+19
View File
@@ -81,6 +81,25 @@ self-hosted embedding/vLLM plan (assumption GĐ-3 in
## Active ownership
- Claude: **IN PROGRESS, 2026-08-05** — making the live demo path survive a
reviewer typing into the UI. Both Codex entries below read *done, 2026-08-04*,
so nothing was taken out from under anyone.
Claiming: `apps/ai-service/rag/{ports,service}.py`,
`apps/ai-service/adapters/embedding.py`, `apps/ai-service/config.py`,
`apps/ai-service/tests/*` (additions), `apps/web/**`,
`packages/api-client/src/*`. **Not touching** `ingestion/`, `cli.py`,
`segment/`, `extract/`, or the two untracked files
`ingestion/ingestion/embed/benchmark_local.py` and
`ingestion/tests/test_embed_benchmark_local.py`, which are Codex's and
still uncommitted.
Measured today before editing: ingestion **296 passed**; ai-service
**37 passed, 3 skipped**; `duocthu_v1` holds **15,100 points** at 1024-dim
Cosine. Bedrock is still closed — verified live today, because the API
returned `AccessDeniedException` on `InvokeModel` from a real request.
**No cloud call, no spend.**
- Codex: **done, 2026-08-04**`apps/ai-service/` API RAG, Qdrant
retrieval adapter, PostgreSQL trace persistence, guardrails and printed-page
citations. Claiming `apps/ai-service/{main.py,config.py,adapters/,routers/}`,