docs(rag): record the S1b canary pair, add packet S1d, mark S4/T4 merged - #2048
Merged
Conversation
…k S4/T4 merged - HANDOVER: S1b canary evidence (baseline 32025082010 -> post 32039841070, first post run 32038751592 red on one non-golden case, root-caused); S1c/S5/S6 Ready; S4 and #212 T4 Merged; new packet S1d (final-gate gap recovery: hedged cited low-confidence fast answers must recover extractively, not collapse to a citation-free provider_source_gap) with §3 body and §7 prompt. - COORDINATION §3 wave 1 re-cut (S1c + S1d + G1 + S5 + S6 parallel; RAG behaviour PRs merged one at a time with own canary pairs); §7 state. - One immutable inbox request for S1d (P2 task). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
This pull request has been ignored for the connected project Preview Branches by Supabase. |
Contributor
Caution Review failedAn error occurred during the review process. Please try again later. Comment |
Uh oh!
There was an error while loading. Please reload this page.
This was referenced Aug 17, 2026
BigSimmo pushed a commit
that referenced
this pull request
Aug 17, 2026
…e HANDOVER status-table conflict) Conflict was the §2 status table: PR #2048 added the S1d packet row and recorded the S1b canary pair on main while this branch updated the S5 row. Resolution keeps main's table (S1d + S1b/S4/T4 updates) and re-applies the S5 row (#2056) and the S5-reconciled 44/44 answer-gate denominator on the S1 row's evidence. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019JLzXeRzE8zze3SoCoo5SH
This was referenced Aug 17, 2026
BigSimmo pushed a commit
that referenced
this pull request
Aug 17, 2026
…ket S1d) A hedged, cited, low-confidence fast answer whose body matches the finalizer's broad gap-like regex used to collapse into a citation-free evidence_gap (final_quality_gate:provider_source_gap) with no recovery, while the identical failure shape caught in the generation loop or the outer generation-fallback catch rebuilds a source-backed extractive answer — model phrasing alone chose the branch (canary 32038751592 red vs 32039841070 green on identical retrieval). finalizeRagAnswerQualityCore's gap conversion now first attempts the same source-backed extractive recovery (recoverFinalGateGapExtractively) when the route was fast + strong_routine_retrieval and answer-input results exist, validated by the existing isSafeExtractiveFallbackCandidate battery and stamped with the established generation_fallback / source_backed_extractive_fallback markers plus final_quality_gate_source_backed_recovery:<reason>. Empty-retrieval, strong-route, comparison, and dose/threshold gaps stay terminal and are now pinned by tests. No rag.ts change (4362-line budget untouched). Also queues an inbox cancellation for the pending S1d ledger add from PR #2048 (a done request cannot fingerprint a not-yet-reconciled row) and documents the recovery in docs/rag-behaviour/behaviour-map.md §7. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01S5QKovS1ZhbZA5kETfQoaR
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
docs/rag-improvement/HANDOVER.md: S1b row now carries its canary pair (baseline run 32025082010 on2bd146eed→ post run 32039841070 one6ad0d5db: recall 1.0/1.0, zero per-case rr regressions, answer gate 45/45; the first post run 32038751592 was red onsummary-discharge-guidanceonly and is root-caused, not attributed to S1b); S1c, S5, S6 marked Ready; S4 (feat(rag): adversarial fixture contract, validator, baseline record and data-flow register (packet S4 / B0) #2036,f5b093291) and fix(supabase): make GUC-set migration hosted-safe; neutralize obsolete N+1 migration #212 T4 (harden(worker): validate claimed job rows and enrichment read-backs against Zod row contracts (#212 T4) #2037,1726537b7) marked Merged; new packet S1d — final-gate gap recovery (§2 row, §3 body, §7 prompt): a hedged, cited,confidence:"low"fast answer whose body matches the finalizer's broad gap-like regex is converted to a citation-freeprovider_source_gapwith no extractive recovery, whereas the same shape caught in-loop recovers extractively; lands before S2.docs/rag-improvement/COORDINATION.md: §3 wave 1 re-cut (S1c + S1d + G1 + S5 + S6 in parallel; RAG-behaviour PRs merged one at a time with their own canary pairs; R0 reconcile landed as docs(issues): reconcile 28 inbox requests into the outstanding-issues ledger #2045); §7 current state.RAG impact: no retrieval behaviour change — documentation and one ledger inbox request only.
Verification
npm run verify:pr-local -- --files <all changed>docs scope:completed: check:runtime, check:installed-lock-parity, format:changed, sitemap:check, docs:check-index, docs:check-inventory, docs:check-scripts, docs:check-links, check:branch-review-ledger, check:outstanding-issues, check:ledger-write-discipline — failed: (none).Ledger inbox check passed: 1 pending request(s), 217 applied.Ledger write discipline passed for 14421fba90d7..HEAD.src/,tests/,scripts/change. UI and provider-backed checks not applicable.Risk and rollout
Notes
e0db503cbunderdocs/branch-review-records/.Opened by the RAG programme coordinator chat (Claude Code).