Also matches: rebuild the cross-mode panel and put it on every mode

[BigSimmo]

Summary

• Rebuilt the "Also matches" cross-mode panel so it reads as a distinct, lower layer beneath the results rather than a flat strip pinned under them, and tuned the layout for phone, tablet and wide desktop separately.
• Gave the panel to the seven modes that never had it: DSM, Specifiers, Calculators, Factsheets, Dictionary, Sources and On Call. Prescribing is no longer suppressed. Answer keeps its own CrossModeLinksSection and is the one recorded exemption.
• Added a coverage contract test that registers all seventeen modes, so a mode either mounts the panel or carries a written reason.

Design changes

• The section is a recessed tray on --surface-subtle holding raised mode cards on --surface-raised + --e1. Previously it was border-lux + --shadow-inset, which card-recipes.ts already documents as reading flat: two edge treatments on one surface.
• Each mode card is vertical now — identity row, full-width result rows, footer link. The old horizontal tile | titles | View all row squeezed titles into the middle third and truncated them at every width, which is why "Mental Health Hospital in the Home" arrived as "Mental Health Hospital in th…". Titles get the whole card and clamp at two lines.
• Result rows are real 48px tap rows with their own hover, focus and chevron affordance (visible on touch, hover-revealed from sm up). The footer reads "View all in <Mode>", matching the cross-mode typeahead's existing wording.
• Loading is a skeleton grid in the real card geometry, so the tray no longer collapses to a single line of text and snap-opens to a four-card grid when matches land.
• Columns are 1 / 2 / 4 across phone, tablet and xl. Four columns inside a 1024px content well leaves no room for a two-line clinical title, so the tablet band holds two columns all the way up to xl.
• The visible aria-live node is replaced by an sr-only role="status" announcer (SPEC §9.2 — a live region is never visible content). It reports pending / count / empty rather than one fixed string.
• Accent follows the recorded option-C study: tint on the icon tile and the short code chip, mode title in heading colour, no rail. also-matches-accent-mockups.tsx is annotated as shipped.

Prescribing — and the two defects found while verifying

Both were mine, both were caught by running the gates rather than by reasoning, and both are fixed in the commits on this branch. Recording them because the second one changes the stated reason for a decision this repo already had.

1. The announcer contradicted the screen (b2c5610). The role="status" node rendered one fixed string, so a screen reader was told "No additional matches in other modes." while four populated mode cards were on screen. It now reports the real state.

2. Prescribing mounted the panel twice (96d9749). Prescribing declares resultKind: "documents" deliberately — it searches the indexed sources, not a forms table — so ClinicalDashboard's documents arm matched it and mounted a second panel over the one MedicationPrescribingWorkspace already renders under the medication list. /?mode=prescribing shipped two.

   This is the half of the original modeId !== "prescribing" early return that the first commit read wrongly. That commit described the suppression as purely positional (the panel used to sit above the medication results). It was also the de-duplication for this shared result kind. The workspace now owns the single mount, because only it knows where the medication result list ends, and the exclusion names the mode rather than the result kind, because the kind is shared and the ownership is not.

   tests/ui-stress.spec.ts caught it by asserting the count is exactly one. The coverage test also carries an offline guard for the same duplicate, and that guard was mutation-verified: removing the exclusion fails it with expected '…' to contain 'searchMode !== "prescribing"'.

Verification

All results below are from the current head (96d9749).

• npm run lint — exit 0. [gate-receipts] recorded a pass for "lint:internal" (7028 input files).

• npm run typecheck (tsc --noEmit -p tsconfig.json) — exit 0, no diagnostics.

• npx vitest run — full offline suite: Test Files 3 failed | 1147 passed | 1 skipped (1151), Tests 3 failed | 16858 passed | 2 expected fail | 3 skipped (16866). The three failures (rag-plan-package-parity, privacy-readiness-contract, clinical-hazard-controls) were reproduced on a stashed clean tree and are pre-existing on main, not this PR's.

• Focused re-runs after each fix: tests/universal-also-matches-mode-coverage.test.ts 20/20, plus audit-navigation-auth-regressions and medication-prescribing-workspace.dom — 3 files / 42 tests passed.

• npm run check:design-system-contract — passed. It caught a real defect mid-change (a --text-soft decoration-alias consumer), corrected to --decoration-soft.

• npm run check:type-scale, check:icon-scale, check:mockups, check:diff-integrity — all passed. diff-integrity: PASS — 11 changed test file(s), 115 -> 119 test case(s), against base 6cfaa94c8.

• npx prettier --check . — whole tree: "All matched files use Prettier code style!", committed.

• Browser (focused, not the full gate). npm run plan:browser selected level: full for this diff, so the complete Chromium suite is CI's to run and was not run locally. What did run locally, against an isolated production build:

  • tests/ui-universal-search.spec.ts — 20 passed (40.0s). Covers the panel's own contracts: after the owning results in the DOM, phone-deferred fetch until expansion, absent on Answer, once on Favourites and the legacy Tools URL.
  • tests/ui-tools-search-mode-mockup.spec.ts — 15 passed, including no horizontal overflow at 320 / 390 / 639 / 768 / 1440 / 1920.
  • tests/ui-stress.spec.ts — 3 passed (12.1s) after the duplicate fix. This is the file whose medication assertion this PR rewrites, and the run before the fix failed it: expect(locator).toHaveCount(expected) failed … Expected: 1 Received: 2.

  Caveat that matters: this container has Playwright chromium-1194 against the repo's pinned 1234, so the runner's preflight fails closed (feat(ui): compact phone bottom search bar on search/result views #255/chore(deps-dev): bump prettier from 3.9.1 to 3.9.4 #312). These runs used PLAYWRIGHT_CHROMIUM_EXECUTABLE_PATH pointed at the 1194 headless shell — a different browser build from the pinned one. Treat them as real evidence for behaviour, not as a substitute for CI's pinned run.

• npm run verify:pr-local — Verification not run: the local heavy-run coordinator was held by the browser runs, and it mirrors what pr-required runs on this PR. Every gate it aggregates for this scope is listed above, run directly.

Visual proof was also taken directly: the panel screenshotted at 390 / 834 / 1440 px in light and dark, plus the loading state, against the dev server with a mocked /api/search/universal. That is what surfaced the title truncation and a 2px row/footer misalignment.

Note on the red PR required on b2c5610: that run reported CANCELLED with no failing job: coverage production-ui. Pushing 96d9749 cancelled those two mid-flight (cancel-in-progress), and the aggregate stays red on a cancelled job because a cancelled job verified nothing. Nothing failed on it — static, safety, build, production-ui-critical, lighthouse-budget and caring-contacts-db all reported success. The workflow's own message says to look for the newer run on the current head SHA, which is the one that counts.

Risk and rollout

• Risk: presentation, plus eight mount points (seven new, one unsuppressed). The panel fetches /api/search/universal only after submission, and on phones only after the reader expands it, so the added mounts add no speculative requests on narrow viewports. The realistic failure mode is cosmetic, or a duplicate mount — which is now covered offline and in the browser.
• Rollback: revert any of the three commits independently, or the branch. No schema, no migration, no config.
• Provider or production effects: None. No Supabase, OpenAI or Railway calls were made. The Supabase integration confirmed on this PR that it ignored it — no supabase/ changes.
• RAG impact: none — no retrieval, ranking, selection or scoring surface is touched. universal-search-mode-context.ts and the retrieval libraries are unchanged.

Clinical Governance Preflight

Not applicable in substance — presentation and mount points only, not ingestion, answer generation, ranking, source rendering, document access or clinical output. For completeness:

• Source-backed claims still require linked source verification before clinical use — unchanged.
• No patient-identifiable document workflow was introduced or expanded.
• Supabase target remains Clinical KB Database (sjrfecxgysukkwxsowpy) — no Supabase config touched.
• Service-role keys and private document access remain server-only — unchanged.
• Demo/synthetic content remains clearly separated from real clinical sources — unchanged.
• Source metadata, review status, and outdated/unknown-source behavior remain conservative — unchanged. The Favourites access gate that hides gated hrefs from guests is preserved verbatim.
• Deployment classification/TGA SaMD impact — no clinical decision-support behaviour changed.

Notes

• One thing worth a reviewer's eye: prescribing showing cross-mode suggestions at all is a judgement call. It is below the medication results and below the interaction verdicts, and collapsed on a phone, but if the preference is that a dosing surface stays free of them entirely, reverting 96d9749's mount is a one-line change and the coverage test's exemption map is where the reason would be recorded.
• tests/universal-also-matches-mode-coverage.test.ts exists because this mount is the part that goes missing: a merge resolution silently dropped the Forms mount once before (recorded in the merge-loss audit for Remove cross-mode “Also matches in other modes” panel and related mounts/tests #1804), and nothing failed when it did.

🤖 Generated with Claude Code

https://claude.ai/code/session_01JQ7yPfVVVh62WursxM6mbF

---

Note

Low Risk
Changes are presentation and mount wiring on search result pages; universal search still runs only after submission (and on phones after expand), with offline and browser tests guarding duplicate prescribing mounts.

Overview
Rebuilds the cross-mode “Also matches” panel so it reads as a recessed tray under primary results: new vertical mode cards (full-width titles, identity row as “view all”, skeleton loading), responsive 1 / 2 / 4 columns, and sr-only status text that matches pending vs populated vs empty states.

Rolls the panel out on DSM, Specifiers, Calculators, Factsheets, Dictionary, Sources, and On Call, and re-enables it for prescribing below the medication list in MedicationPrescribingWorkspace instead of suppressing fetches/render in the component. Prevents a duplicate panel on /?mode=prescribing by excluding prescribing from ClinicalDashboard's universal mount (prescribing shares the documents result kind). Answer stays exempt via its own cross-mode links.

Adds tests/universal-also-matches-mode-coverage.test.ts so every app mode either mounts UniversalSearchAlsoMatches or documents why not; page unit tests mock the panel to avoid auth coupling.

^{Reviewed by Cursor Bugbot for commit 33a8532. Configure here.}

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.

[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: 7cde6380-fd9a-4a57-9d25-0e2c5fe59d59

---

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_47c5de5e-5e77-4286-8c8e-e3e08261506b)

[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 ↗︎.

[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_9c99190b-089e-49f0-a770-5e50b9273ca1)

[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_cb093334-6dbf-40da-87c8-eee70e7211e2)