fix(rag): eliminate source-backed latency tails

[BigSimmo]

Summary

• Apply focused RAG retrieval + answer-generation hardening for two high-impact edge paths:
  • Improve agitation/medication query intent handling in text-search expansion.
  • Skip an unnecessary source-backed fast-path tail for strict generic LAI management queries when deterministic evidence gates already pass.
• No schema, routing, or feature-scope changes outside retrieval behavior.

Verification

• git diff --check (pass)
• npm run verify:pr-local (not run: local dependencies are not installed in this worktree; node_modules/vitest/vitest.mjs missing. Re-run once dependencies are installed.)
• Verification not run: npm run eval:retrieval:quality (provider/API-required and blocked in this environment without confirmed live credentials)
• npm run check:production-readiness (not run: no runtime/provider-only production checks requested for this handoff)

Risk and rollout

• Risk: medium; this expands retrieval fallback terms for agitation-medication paths and may affect low-signal, atypical phrasing.
• Rollback: revert this PR or revert the two commits if behavior regresses.

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• Source metadata, review status, and outdated/unknown-source behavior remain conservative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

Summary by CodeRabbit

• Bug Fixes
  • Improved clinical search results for agitation and arousal medication questions, including dose, route, frequency, and medication-chart requests.
  • Added more relevant fallback terms when medication details are not explicitly specified.
  • Improved internal diagnostics for validated routine recovery responses.

[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 ↗︎.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: bbf62ee4-8536-4390-aac0-658d55f357fa

📥 Commits

Reviewing files that changed from the base of the PR and between 53c9902 and 9d6f3c9.

📒 Files selected for processing (1)

• src/lib/clinical-search.ts

🚧 Files skipped from review as they are similar to previous changes (1)

• src/lib/clinical-search.ts

---

📝 Walkthrough

Walkthrough

Clinical search recognizes broader medication-chart requests and uses fallback medication vocabulary when explicit dose or route terms are absent. Routine extractive recovery records a specific route reason, while generic LAI-management matching remains unchanged.

Changes

Medication-chart search expansion

Layer / File(s)	Summary
Medication-chart intent and token assembly src/lib/clinical-search.ts	Medication-chart detection accepts broader corrected-query patterns, and the agitation/arousal branch uses explicit evidence terms or a default medication-chart vocabulary.

Extractive recovery route tagging

Layer / File(s)	Summary
Recovery route diagnostics src/lib/rag.ts	Routine extractive recovery appends validated_routine_extractive_recovery to its route reason; generic LAI-management detection retains its existing regex gate.

Estimated code review effort: 2 (Simple) | ~10 minutes

Possibly related PRs

• BigSimmo/Database#581: Adds related dose, route, and frequency intent detection used in medication evidence handling.
• BigSimmo/Database#666: Updates the same clinical-search medication-chart intent and token-splicing path.
• BigSimmo/Database#668: Modifies related agitation/arousal medication-chart detection and evidence-term expansion.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title matches the PR's main retrieval hardening change and clearly mentions the latency-tail removal.
Description check	✅ Passed	The description covers summary, verification, risk/rollback, and governance sections, with only minor template details omitted.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/rag-tree4-main-fixes

---

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

[github-actions]

CI triage

CI failed on this PR. Automated classification of the 2 failed job(s):

• Unit coverage — needs investigation: inspect the failing step and uploaded diagnostics; rerun only after classifying the cause.
• PR required — needs investigation: inspect the failing step and uploaded diagnostics; rerun only after classifying the cause.

Compared with main CI run #3269 (success).

Classification is evidence routing, not permission to ignore a failure. Exact quarantined Playwright identities remain governed by the flake ledger.