Document rejected RAG canary findings

[BigSimmo]

Summary

• correct Bump @types/react from 19.2.16 to 19.2.17 #18: lithium, ADHD, metabolic and admission misses have distinct mechanisms
• record the rejected lithium subject-evidence experiment and its archive ref
• add Refactor monolithic front‑end and decouple domain logic #51 for two different answer-gate failures observed on byte-identical trees
• explicitly defer further live retries and RAG tuning until the answer-quality canary is stable

RAG impact: no retrieval behaviour change — candidate fd3131803 was reverted by 9236ce981; the final PR tree differs from origin/main only in docs/outstanding-issues.md.

Verification

• npm run format:changed
• git diff --check
• final code tree equality was proven immediately after the revert
• candidate run 30007833352: golden retrieval recall 1.0, zero reciprocal-rank regressions; answer gate failed unsupported-route p95
• reverted confirmation 30009207429: golden retrieval recall 1.0; answer gate failed differently on citation failure

Code tests/build not rerun after the final docs-only commit: the code diff is empty and the candidate had already passed verify:cheap and verify:pr-local before being reverted.
UI verification not run: documentation only.

Risk and rollout

• Risk: documentation-only; no runtime behavior.
• Rollback: revert the docs commit.
• Provider or production effects: two explicitly authorized read-only canary runs were completed; no migration, mutation, deployment or additional retry.

Clinical Governance Preflight

• Source-backed claims still require linked source verification before clinical use
• No patient-identifiable document workflow was introduced or expanded without explicit governance approval
• Supabase target remains Clinical KB Database (sjrfecxgysukkwxsowpy)
• Service-role keys and private document access remain server-only
• Demo/synthetic content remains clearly separated from real clinical sources
• Source metadata, review status, and outdated/unknown-source behavior remain conservative
• Deployment classification/TGA SaMD impact was checked when clinical decision-support behavior changed

Notes

• Rejected commit remains recoverable at local ref refs/archive/rejected-rag/20260723/monitoring-subject-gate.
• No further live RAG run is recommended now.

---

Note

Low Risk
Documentation-only; no application code, deployment, or RAG behavior changes.

Overview
Updates docs/outstanding-issues.md only—no runtime or retrieval changes (reverted RAG experiment; tree matches main aside from this doc).

Adds a 2026-07-23 RAG reconciliation callout that supersedes the single-composer story in #18: lithium, ADHD, metabolic, and admission misses are documented as separate mechanisms (e.g. lithium fast-path, ADHD extractive budget, metabolic prose selection; #19 stays post-retrieval comparison). It notes the reverted lithium subject-evidence guard, archive ref refs/archive/rejected-rag/20260723/monitoring-subject-gate, keeps #29 for remaining fallback stubs, and defers further RAG tuning without a reproducer + passing canary pair.

Introduces P2 #51 to stabilise the live answer-quality canary before more RAG work: two byte-identical-tree runs passed golden retrieval but failed the answer gate for different reasons (unsupported-route p95 vs citation failure). Next action is gate/diagnostic work, not immediate retries.

Bumps the issues tracker next-id to 052 and surfaces #51 at the top of the open-items table.

^{Reviewed by Cursor Bugbot for commit 3694141. Configure here.}

[coderabbitai]

Important

Review skipped

Review was skipped as selected files did not have any reviewable changes.

💤 Files selected but had no reviewable changes (1)

• docs/outstanding-issues.md

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 6a09e039-939e-420c-b4fa-02563074bb92

📥 Commits

Reviewing files that changed from the base of the PR and between 5df8c96 and 4001006.

📒 Files selected for processing (1)

• docs/outstanding-issues.md

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/rag-answer-composition

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[supabase]

This pull request has been ignored for the connected project sjrfecxgysukkwxsowpy because there are no changes detected in supabase directory. You can change this behaviour in Project Integrations Settings ↗︎.

---

Preview Branches by Supabase.
Learn more about Supabase Branching ↗︎.

[BigSimmo]

Post-canary 30007833352 kept golden document recall at 1.0 with zero per-case reciprocal-rank regressions and improved the lithium targeting result, but the workflow failed its unsupported-route p95 gate (6,227 ms > 4,000 ms) and one additional answer case entered a review fallback. Per the protected RAG rollback rule, commit fd3131803 was reverted in 9236ce981. The reverted branch tree is byte-identical to current origin/main; confirmation canary pending. The rejected candidate remains recoverable at local archive ref refs/archive/rejected-rag/20260723/monitoring-subject-gate.

[BigSimmo]

Reverted confirmation run 30009207429 failed the answer gate for citation-failure rate 0.0227, different from candidate run 30007833352 (unsupported-route p95 6,227 ms). Both byte-identical trees passed golden retrieval with recall 1.0. Final PR diff is documentation-only; no further live retry is recommended.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 369414187a

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".