docs: council Phase C lean-out plan + safety audit (rebased)#154
Conversation
…night) Source: claude --print --model claude-opus-4-7 ran the lean-out audit 2026-04-30 02:00-02:12. Agent prepared the plan but couldn't Write the file directly (permission gate). Plan extracted from stdout. Headlines: - Baseline: SDK 68,717 LOC / 249 files. Target: 43,861 (-36%) - Scoring: keep _confidence.py (FSRS/Bayesian); pragmatist disagrees, prefers Beta - Modules: only 1 of 3 'duplicate pairs' is real (inspection.py) - Correction APIs: correct() canonical; rest private/renamed - Cloud: REMOVE _cloud_sync.py, MOVE cloud/* to extra - MCP/daemon/contrib/adapters/mining/sidecar/22 hooks → extras Architect (90 days) vs Pragmatist (60 days) disagreement flagged. Wildcard lens failed (deepseek 500); 6 of 7 voices in synthesis. Full council #2 launch report at ~/.hermes/council_reports/council_2026-04-30T02-02-45.md.
Per-candidate verification using rg/search across ~/.claude/, SpritesWork/, ~/.hermes/, and Gradata's own src+tests. Findings invalidate several 'DELETE' recommendations from LEANOUT_PLAN: - _cloud_sync.py: NOT dead. Imported by tests, doctor, migrations, hooks. Move to [cloud] extra, do not delete. - scoring/memory_extraction.py: NOT 'broken import'. Has dedicated test + brain.py reference. Audit before deletion. - daemon.py: KEEP as extra. Phase B just wired BrainLockedError into it. - hooks/implicit_feedback.py: KEEP. Phase B just hardened it. - correction_detector privatize: DEPRECATE-rename across releases, not immediate. Confirmed safe to delete: graduation/scoring.py (0 imports verified). Confirmed safe rename: events_bus.py, _config.py (deprecation alias needed). Live Brain snapshot at ~/.hermes/brain-snapshot-20260430-0845/critical/. Phase B (242c408) is independently mergeable.
…etes, Phase B 3970/3970 pytest pass Pragmatist's DELETE list (notifications/onboard/safety) verified UNSAFE: - notifications.py: PUBLIC API (__init__.py exports Notification), Brain.subscribe() - onboard.py: Brain.init() canonical bootstrap (README headline) - safety.py: PII redaction used by _core.py Phase B test sweep: 3970 passed, 0 failed, 4:34 wall. Phase B branch is provably non-regressive — safe to push.
…elete failed (test_graduation_scoring imports it) Lesson: '0 production import sites' != 'safe to delete'. Test coverage counts. Opt-in features with dedicated test files = real surface area. After full verification pass, NONE of the 14 LEANOUT_PLAN deletions should ship without first migrating their tests + verifying again.
There was a problem hiding this comment.
Your free trial has ended. If you'd like to continue receiving code reviews, you can add a payment method here.
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (2)
📝 Walkthrough
WalkthroughIntroduces two new planning and verification documentation files for a lean-out initiative: one establishing baseline metrics, reduction targets, and a phased execution strategy; the other documenting safety audit results and verified deletion/deprecation candidates for the Gradata repository. Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~12 minutes Suggested labels
✨ Finishing Touches🧪 Generate unit tests (beta)
Review rate limit: 2/5 reviews remaining, refill in 30 minutes and 48 seconds. Comment |
Clean rebase of #147. Cherry-picked Phase C commit(s) only, drops stale m1 base.