Skip to content

Chunk Quality v3.4: retrieval simulator, coverage matrix, report export #12

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
willscott-v2 merged 2 commits into from
Jul 25, 2026
Merged

Chunk Quality v3.4: retrieval simulator, coverage matrix, report export #12

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2d1d6b8
Retrieval simulator + coverage matrix + report export (v3.4)
willscott-v2 Jul 25, 2026
14e1f7c
Correct SESSION_LOG branch-cleanup note
willscott-v2 Jul 25, 2026
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
16 changes: 16 additions & 0 deletions SESSION_LOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,22 @@

**Next:** Will reviews v3.3 → then Phase 3 (v3.4): BM25 retrieval simulator, query coverage matrix, markdown optimization-report export.

## 2026-07-25 (cont.) — Phase 3 (v3.4) executed; Chunk Quality plan COMPLETE

**Will said "get through all the outstanding work."** Built v3.4 on `feature/chunk-quality-v3.4` (stacked on v3.3), PR #12:
- Client-side BM25 (k1=1.2/b=0.75, no deps) over sections — heading indexed with text, overlap included, mirroring what the JSONL export feeds a real pipeline. All in-browser.
- "Test retrieval" panel: single query → top-5 ranked sections with scores/bars/matched terms; winner highlighted with its flag chips and a stands-alone verdict (Hayes: "Chunk 2 · section 1 wins but is flagged (Dangling reference)").
- Coverage check: one query per line → matrix (query / status / best section / score / terms), covered = ≥60% of query terms matched, weak, or gap. Summary counts + Copy CSV.
- "Copy optimization report": markdown deliverable — score/grade, top fixes, flags grouped Fix first / Then improve / Worth knowing, coverage table, content-gaps list, settings. Verified via clipboard read in Playwright (Hayes: 4k chars, "dog boarding rates" correctly a gap, "cost per month" correctly weak).
- State: coverageRows reset per extraction; all buttons delegated (view re-renders per run); Enter submits the single query.

**Backlog cleanup:** `feature/enhanced-chunking` turned out to be already deleted (backlog item was stale); pruned leftover local tracking refs for the merged `chore/remove-dead-css` + `feature/marketing-shell-parity` branches. Branch list is now just main + the three chunk-quality branches.

**Open at day's end:**
1. **Merge chain awaiting Will:** #10 (v3.2, base main) → #11 (v3.3, retarget to main after #10) → #12 (v3.4, retarget after #11). Merging main auto-deploys.
2. Fixture/snapshot tests (split/merge/overlap/analyzer) — deferred since v3.1, still the one real backlog item.
3. Phase 4 optional: template query generator, `?queries=` Ontologizer deep-link, compare mode, README rubric docs.

## 2026-07-24 — Chunk Quality plan (from Chuck's feedback) + repo moved out of archive

**Session ran from ontologizer-next; planning only, no code changes here yet.**
Expand Down
Loading
Loading