Skip to content

docs: record the HUB-043 deploy verification #13

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
funsaized merged 1 commit into from
Aug 13, 2026
Merged

docs: record the HUB-043 deploy verification #13

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 22 additions & 0 deletions docs/CURRENT_STATE.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,6 +57,28 @@ a `document_id` and every one resolves in SQLite, so corpus-wide retrieval
drops nothing. (33 retained documents have no Qdrant chunks — deduplicated
sources — and remain lexically reachable.)

**Deployed and verified on the stack.** `research-hub` and `research-worker`
rebuilt from `63aacc6`; `app/{retrieval,query,document_store,clients,main,
observability}.py` SHA-256 verified identical between the repo and both
running containers. `/readyz` reports `capability: all` with all six services
true.

Live on the deployed API:

| Call | Result |
|---|---|
| `/query` "how does reciprocal rank fusion combine rankings" | 6 chunks from **4 sources** across different jobs; top 3 dual-channel `dense,lexical` |
| `/query` + `topic_filter` "Kubernetes pod autoscaling…" | 4 chunks from 3 sources, all on-topic |
| `/query` + an unmatched `topic_filter` | **0 chunks** — the filter narrows, it does not silently widen |
| `/rag` "microservices versus monoliths tradeoffs" | answered with citations over 4 distinct sources |

Persisted state after the deploy is unchanged: the attempt-11 report and
source registry are still byte-identical (`068d60b2…`, `d6748d76…`), the v4
seal is `762e7a19…`, Qdrant holds 68,072 points, and the store holds 679
documents / 758 observations / 71,125 lexical rows. `hub_retrieval_duration_seconds`
is live and `hub_retrieval_score` recorded 18 observations across the runs
above.

## Live report generated through the deployed judge gate (2026-08-13)

The judge pivot had been proven structurally but no report had ever been
Expand Down
Loading