quantmodel — Enrich completed-pipeline changelog entries from task and commit context

Objective

Results / Changelog records at docs/results.html and docs/site-data/results.json showed generic placeholder titles and summaries like “Complete pipeline task (#824)” for 11 pipeline-completed tasks. Reviewers could not tell what shipped. The task required each generic entry’s displayed title to identify the actual delivered work and its description to concisely explain the material change and why it matters using both the linked Beads task and the associated commit as evidence, handling missing/sparse fields without fabricating details, fixing the changelog producer rather than hand-editing generated docs, preserving existing nongeneric entries, and verifying via focused producer and executable page coverage.

Approach

Classified as maintenance bounded producer fix (no StrategySpec, indicator, or data-source). No child tasks required.
Inspected research/reports/results.py:_load_entries — it rendered front-matter title/summary verbatim, so reconstructed research/results/qm-*.md files with generic placeholders flowed straight to the payload and HTML.
Added producer enrichment in research/reports/results.py: _is_generic_entry detection, _try_load_beads_task via bd show --json, _try_load_git_commit via git log/show parsing Commit \sha\` from the markdown body (fallback git log --grep=eid), and _enrich_generic_entry` that prefers the Beads task title, falls back to a cleaned commit subject, and builds a concise summary from the Beads description + commit file list/AC without inventing thresholds, handling sparse cases with an explicit “sparse — see commit and task” fallback and truncating safely.
Wired enrichment into _load_entries so future generic markdowns are auto-corrected at build time while nongeneric entries pass through unchanged.
Backfilled the 11 source markdowns with grounded titles/summaries derived from bd show and git show --stat:
qm-p0m.2 → “Extend Signals history with availability-bound aggregate” (commit 8a59458a, 170k-line signals/history.json rebuild)
qm-p0m.3 → “Replace indicator dropdown with sortable signal tiles and quality metrics” (fa1a6878, risk_off_history/signals_history)
qm-p0m.4 → “Make the aggregate risk gauge interactive from signal tiles” (5922b397, site.js 208 insertions)
qm-p0m.5 → “Standardize risk-on trigger markers and underlying indicator charts” (6682b536, charts.js 257 insertions)
qm-xz0.2 → “Extract SPYST charts and parse a strict signal schema” (73cd7d3b, 635-line extract.py)
qm-xz0.3 → “Validate SPYST parser on history and add a manual-review fallback” (5263cc97, validate.py 437 lines)
qm-xz0.4 → “Keep SPYST allocation logic separate and at parity…” (65223265, allocation parity)
qm-xz0.5 → “Hook SPYST into refresh with late-mail handling and parse alerts” (e50e4cb1, refresh.py 411 lines)
qm-xz0.6 → “Add offline SPYST email-to-allocation regression fixtures” (f99650c2, 386-line regression)
qm-7ad → “Implement max-CAGR-at-vol kernel and wrappers” (6048fd68, 601 insertions)
qm-rjo.4 → “Remove the Compare Strategy surface” (efa1b9e1, 428-line comparison.js deletion)

Files / code / data changed

research/reports/results.py — added import json, subprocess, _GENERIC_TITLE_RE, _is_generic_entry, _try_load_beads_task, _try_load_git_commit, _enrich_generic_entry, and enrichment wire-in _load_entries (future generic placeholders auto-enriched at build time, sparse handling without fabrication)
research/results/qm-p0m.2.md, qm-p0m.3.md, qm-p0m.4.md, qm-p0m.5.md, qm-xz0.2.md, qm-xz0.3.md, qm-xz0.4.md, qm-xz0.5.md, qm-xz0.6.md, qm-7ad.md, qm-rjo.4.md — front-matter title/summary replaced with grounded copy (titles = Beads task titles verbatim; summaries = 1–2 sentences: task why + commit file evidence with SHA)
tests/test_changelog_enrichment.py — new focused producer + executable page coverage (7 tests): generic enriched, summaries grounded in SHAs, nongeneric retained, sparse fallback, producer contains helpers, rendered HTML contains enriched titles with zero generic link placeholders, source markdowns searchable
Generated via producer (not hand-edited): docs/site-data/results.json (n_entries 483, 0 generic titles), docs/results.html (tbody links enriched), docs/results/qm-*.html (11 detail pages)

Validation

python3 -m research.reports.results — rebuilt deterministically; before 11 generic titles, after 0; docs/results.html shows enriched titles, no <td><a>Complete pipeline task remains
python3 -m pytest -q tests/test_results.py tests/test_changelog_enrichment.py — 14 passed (7 existing + 7 new); nongeneric qm-9kwp/qm-avfa titles unchanged, sparse case qm-test-generic-tmp correctly falls back to eid + “sparse — see commit and task” without fabricating
python3 research/scripts/check_static_reports.py docs — local bundle OK: 1207 files; 177 canary routes
python3 -m research.pipeline verify --offline — status passed (all offline scenarios, including changelog_entry artifact contract requiring research/results/*.md + docs/site-data/results.json + docs/results.html + docs/results/<id>.html via python3 -m research.reports.results)
python3 -m research.pipeline validate — status passed, nodes 28, edges 38, hash sha256:6802…
git diff --name-only shows no data/market.sqlite mutation, no private data, no hand-edited docs/site-data/results.json outside producer

Results

Changelog index now reads as specific delivered work rather than placeholders: each former “Complete pipeline task (#NNN)” row links with its real Beads title (e.g. “Extend Signals history with availability-bound aggregate”, “Remove the Compare Strategy surface”) and the payload summary concisely explains what changed and why (e.g. contemporaneous max, sortable tiles, SPYST refresh cadence) with commit SHA and file evidence. Nongeneric entries (qm-9kwp, qm-avfa) retain intended copy; synthetic generic with missing beads/commit degrades gracefully to a sparse notice. The fix lives in research/reports/results.py and regenerated artifacts, preserving deterministic site-data ownership and enabling git grep over research/results/qm-*.md.

Conclusions

Completed the maintenance fix end-to-end: 11 historical entries enriched, producer hardened for future generics, searchable source markdowns updated, deterministic docs/site-data/results.json/docs/results.html regenerated, and executable coverage ensures no generic placeholders return.

Problems / follow-ups

research/results/qm-*.md bodies still contain the original reconstructed “Complete pipeline task (#NNN) (commit …)” objective line — front-matter is what the payload renders, so no further change needed; a future pass could rewrite body objectives for completeness.
tests/test_reports_scaffold.py expects 11 meta strategies but now sees 13 after qm-7ad’s two “Max CAGR at vol” metas, and research.artifacts reports drift — pre-existing from qm-7ad, not introduced by this task; tracked as independent follow-up.

Enrich completed-pipeline changelog entries from task and commit context