Add conservative typo-tolerant fuzzy matching to catalogue searches - #1800
Conversation
|
This pull request has been ignored for the connected project Preview Branches by Supabase. |
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (1)
📝 WalkthroughWalkthroughAdded conservative typo matching with bounded Damerau-Levenshtein distance. Integrated fuzzy fallback scoring into catalogue, therapy, formulation, specifier, and factsheet searches. Added regression tests for supported and rejected matches. ChangesFuzzy search behavior
Estimated code review effort: 3 (Moderate) | ~20 minutes Sequence Diagram(s)sequenceDiagram
participant SearchEntry
participant fuzzySearchTokenCount
participant SearchRecord
participant SearchResult
SearchEntry->>SearchRecord: evaluate literal query matches
SearchEntry->>fuzzySearchTokenCount: evaluate fuzzy tokens when literal score is zero
fuzzySearchTokenCount-->>SearchEntry: return fuzzy match count
SearchEntry->>SearchResult: include and rank matching records
Possibly related PRs
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src/lib/catalog-search.ts`:
- Around line 27-30: Update typoDistanceLimit so four-character terms receive no
fuzzy-matching allowance, requiring at least five characters before returning a
nonzero distance. Preserve the existing thresholds for terms of length eight or
more and add a regression asserting fuzzySearchTokenCount("SSRI", "SNRI")
returns 0.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: a9e629da-060f-4d98-9353-a01fc6b9239b
📒 Files selected for processing (12)
docs/branch-review-ledger.mdsrc/components/factsheets/factsheets-data.tssrc/components/therapy-compass/data/select.tssrc/lib/catalog-search.tssrc/lib/formulation.tssrc/lib/specifiers.tssrc/lib/therapies.tstests/catalog-search.test.tstests/factsheets-data.test.tstests/formulation.test.tstests/specifiers.test.tstests/therapy-card-preview.test.ts
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a553d65c49
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".
Sync stale PR head; merge-tree was clean (GitHub DIRTY was behind-main only).
There was a problem hiding this comment.
Pull request overview
This PR adds conservative, typo-tolerant fuzzy matching to non-answer catalogue-style searches (therapies, formulation mechanisms, specifiers, factsheets, and shared catalogue ranker) while keeping answer/RAG behavior unchanged.
Changes:
- Introduces a bounded Damerau–Levenshtein helper (
fuzzySearchTokenCount) and integrates it as a fallback signal inrankCatalogRecords. - Applies fuzzy fallback to several catalogue consumers (therapy compass search, server-side therapy catalogue, specifiers, formulation mechanisms, factsheets filtering).
- Adds focused regression tests covering common misspellings, insertions/omissions, and transpositions.
Reviewed changes
Copilot reviewed 12 out of 12 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
| tests/therapy-card-preview.test.ts | Adds a regression test ensuring therapy search recovers a close typo. |
| tests/specifiers.test.ts | Adds a regression test for typo recovery in specifier search. |
| tests/formulation.test.ts | Adds a regression test for typo recovery in mechanism search. |
| tests/factsheets-data.test.ts | Adds a regression test for typo recovery in factsheets filtering. |
| tests/catalog-search.test.ts | Adds tests for fuzzy token counting + guardrails (short abbreviations, unrelated words). |
| src/lib/therapies.ts | Adds fuzzy fallback to server-side therapy record search scoring. |
| src/lib/specifiers.ts | Adds fuzzy fallback to specifier scoring when literal score is absent. |
| src/lib/formulation.ts | Adds fuzzy fallback to mechanism scoring when literal score is absent. |
| src/lib/catalog-search.ts | Adds bounded typo-distance logic and fuzzy evidence integration/signals in shared ranker. |
| src/components/therapy-compass/data/select.ts | Adds fuzzy fallback to therapy compass UI-side scoring when literal matches are absent. |
| src/components/factsheets/factsheets-data.ts | Adds fuzzy fallback to factsheets search-page filter logic. |
| docs/branch-review-ledger.md | Appends a ledger row related to this work. |
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
SSRI and SNRI are one Damerau-Levenshtein edit apart; raise the fuzzy floor to five characters and pin a regression so medication-class labels do not cross-match in catalogue fallback consumers.
Babysit / unblock progress
Merge left until PR policy unblocked. |
Cloud agent tokens cannot PATCH pull request bodies; the Sync PR policy body job applies this scratch template with pull-requests:write. Remove before merge so it does not land on main (#230).
Keeps the scratch template off the merge tip (#230) and retriggers PR policy against the already-synced Clinical Governance Preflight body.
CI triageCI failed on this PR. Automated classification of the 2 failed job(s):
Compared with main CI run #9614 (cancelled). Classification is evidence routing, not permission to ignore a failure. Exact quarantined Playwright identities remain governed by the flake ledger. |
Production UI shard 1 failed locator.check({ force }) on the controlled
sr-only CBT cycle radio. Click the visible label instead.
|
Unblock snapshot (2026-08-10): PR head is b680dee. Current state is merge-tree clean, mergeable=true, mergeStateStatus=BEHIND, required workflows passing (Gitleaks + PR policy + PR required). I found 1 unresolved review thread on src/lib/catalog-search.ts (line 198) from copilot about fuzzy-signal ordering. I agree this is a valid behavior concern but it is currently advisory only, not a required-check blocker; no code change made during unblock pass. Mergeable blockers remain clear; merge left to you, with follow-up fix noted. |
Four merge-safe intake requests, no canonical ledger edit: - P1 rec: no gate detects a merged PR whose content is silently reverted by a later merge resolution. Blob sweep over origin/main found seven such PRs since 2026-08-06, six first wiped at one commit in the PR #1815 unblock chain. The reverts took each PR's tests in the same stroke, which is why nothing went red. - P2 task: re-land PR #1800 (fuzzy catalogue search) with the #310 one-edit cap applied in the same commit. - P2 task: re-land PR #1803 (--shadow-tight retirement); 67 files on main still reference the retired alias. - update #310: re-scoped. The defect is not reachable on main because the matcher is absent, not because it was fixed, so the row must not be closed and the fix belongs in the #1800 re-land. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DuYJz8hauCsCdx8r4fXiZU
Summary
Verification
node scripts/run-vitest.mjs run tests/catalog-search.test.ts tests/specifiers.test.ts tests/formulation.test.ts tests/factsheets-data.test.ts tests/therapy-card-preview.test.ts --reporter=dot— 49 passednpm run typecheckand changed-file ESLint checks passed for the edited filesnpm run formatcompleted cleanlynpm run verify:pr-localfull gate deferred to hosted CI after branch syncRisk and rollout
Clinical Governance Preflight
Clinical KB Database(sjrfecxgysukkwxsowpy)Notes
origin/main(merge-tree was clean; GitHub DIRTY was staleness) and raised the fuzzy floor to five characters for the SSRI/SNRI CodeRabbit finding.Summary by CodeRabbit