Specifiers builder: correct the one-of mapping, and show each option's review status

[BigSimmo]

Summary

Three gaps left by the catalogue-driven builder in PR #2676. None is a computation bug — all three are about what the screen implies versus what the data supports.

• The one-of mapping was wrong on four group labels, and could block a valid combination. singleSelectGroupLabels decides whether a specifier group renders as radios. The dataset carries no exclusivity field — a group is a label and a list — so that mapping is authored in this repo, and a wrong entry makes a combination unreachable with nothing on screen to say why. Four labels were wrong against the manual and are now multi-select: Classes (concurrent alcohol, cannabis and tobacco use disorders are ordinary and each is coded in its own right), Clusters (personality disorders across clusters are common and coded as both), Aetiology (the neurocognitive list pairs a certainty term with the aetiology itself, so probable Alzheimer's disease needs two picks, and delirium has to allow more than one contributor), and Attraction (limited to incest is a separate axis, applied alongside the attraction type). Type stays one-of, but specific phobia opts out of it, since the manual asks for every applicable phobia type to be coded — so builderGroupSelection now takes the disorder as well as the label.
• More importantly, the mapping can no longer block anything. Every single-select group carries a control that reopens it as a checkbox list. A mapping error now costs a default rather than a combination the clinician cannot record, which is the failure direction this surface should have. The step copy states that one-of is the builder's default, not a verified manual rule.
• Review status was invisible where it is acted on. The builder shows a catalogue option at the moment the clinician selects it and copies it into a note, but showed nothing about that item's review state. Of the 585 catalogued specifiers, 71 have verified sources and all 585 remain pending qualified clinician review. SpecifierReferencePage and the detail routes already render ReviewStatusBadge for exactly this reason, so the option rows now carry it too. Generated definitions stay withheld, unchanged.

The curated mood path is untouched throughout: it renders no badges (those are hand-authored records, not catalogue rows), keeps its own fit and exclusion rules, and its wording output is identical. Deep links and the copied phrase are unchanged.

Scope: one library file, one component file, one unit test, plus two issues-inbox request files.

Verification

• npm run verify:pr-local

Verification not run: verify:pr-local selects heavy scope for an executable change (lint, typecheck, full test, build and the static suite), which duplicates the verdict CI reaches on this PR. The gates actually run locally, with their real output:

• node ./node_modules/typescript/bin/tsc -p tsconfig.typecheck.json --noEmit — clean, no output.

• eslint on all three changed source/test files --max-warnings 0 — clean.

• npx prettier --check on every changed file, the two inbox requests included — All matched files use Prettier code style!

• npx vitest related --run across the specifier suites — Test Files 3 passed (3) / Tests 25 passed (25). Four new cases pin the corrected mapping: that the four relaxed labels are multi-select, that specific phobia opts out of the label-wide Type rule while delusional disorder does not, that two substance classes can be held at once, and that relaxBuilderGroups reopens a named group without mutating the original or touching its siblings.

• npm run check:design-drift-ratchet — ✓ inline style={{ }} attributes: 233 (ceiling 233). No new inline styles; this gate caught a regression on Specifiers builder: offer every catalogued base diagnosis, not five hardcoded mood presets #2676 so it is run deliberately.

• node scripts/ledger-inbox.mjs check — Ledger inbox check passed: 36 pending request(s), 1080 applied.

• npm run verify:ui when UI, routing, styling, browser behavior, reduced-motion, or forced-colors behavior changed

Focused browser proof, not the full UI gate. npm run plan:browser selected level focused and attributed the changed component to its owning specs:

  28 passed (43.5s)

[browser-test-plan] PASSED at level "focused" — this is focused browser proof, not the full UI gate.

Every existing builder journey passed unchanged, including blocks incompatible specifiers and preserves severe psychotic-features wording (the curated mood path's own exclusion rules, a separate mechanism from the catalogue mapping), keeps the guided builder readable without horizontal overflow on phone, and the reduced-motion / forced-colors journeys. The full verify:ui suite is left to CI.

• npm run verify:release before release or handoff confidence claims

Risk and rollout

• Risk: Low, confined to the Specifiers Build route's catalogue path. No wording output, deep-link resolution, or data access changes. The clinical direction is conservative in both changes: the mapping now permits combinations it used to block, and it surfaces an unreviewed state that was previously silent.
• Rollback: Single-commit revert on each commit. No data, schema, or configuration changes.
• Provider or production effects: None. No Supabase, OpenAI, or Railway surface is touched, and no provider-backed command was run.
• RAG impact: none

Clinical Governance Preflight

• Source-backed claims still require linked source verification before clinical use — each applied specifier still links to its /specifiers/<slug> record, and its review status is now visible before selection rather than only after opening that record.
• No patient-identifiable document workflow was introduced or expanded without explicit governance approval
• Supabase target remains Clinical KB Database (sjrfecxgysukkwxsowpy) — untouched.
• Service-role keys and private document access remain server-only — untouched.
• Demo/synthetic content remains clearly separated from real clinical sources
• Source metadata, review status, and outdated/unknown-source behavior remain conservative — this PR strengthens that: 514 of 585 catalogue items read Review due at the point of selection, and generated definitions remain withheld exactly as on the reference pages.
• Deployment classification/TGA SaMD impact was checked when clinical decision-support behavior changed — the builder remains a documentation aide-memoire producing a wording string, not decision support. Both changes move it further from implying decision support: it withdraws an implied manual rule and stops enforcing an unsourced one.

What this PR does not close, and where it is now recorded

Both remaining gaps need a qualified clinician rather than a code change, so both are queued as issues-inbox requests in this PR (merge-safe request files, not ledger rows; npm run issues:reconcile applies them after it lands):

• P2 — the catalogue has no clinician sign-off. 585 items, 71 source-verified, none reviewed, dataset lastUpdated 2026-05-09. This PR makes that state honest at the point of selection, which is as far as code reaches.
• P3 — the fifteen remaining one-of labels are authored here rather than reviewed. The reopen control bounds the consequence to a default the clinician can override rather than an unrecordable combination, so this is a confirmation pass and not a blocker.

🤖 Generated with Claude Code

https://claude.ai/code/session_011RJFStNCUPhDrEdAJe2r4i

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Team

Run ID: 98427c80-4f98-405d-9215-0b4e2b00f4a4

---

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.}

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_b8cead28-f187-4a09-8a50-672d84d49ad2)

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-07T10:48:19.310004Z	d1e882a	PR opened

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[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 ↗︎.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d1e882a533

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".