fix(dashboard): clearing the composer brought back two retired mode homes

[BigSimmo]

Summary

Follow-up to #2672. You asked me to root-cause seven items from that PR's audit and apply the smallest safe fix to each. Three were real, two were my own misreads, one is yours to rule on, and one was resolved by another session. All of that is set out below with evidence.

Issues 1 and 2 turned out to be a single root cause, and it is the same defect a clinician reported on Sources. Clearing the composer emptied the React query but left q and run=1 in the URL. showSharedHome reads run=1 straight off the URL (submittedUrlRunRequested), so it stayed suppressed while the mode branch beneath it, now seeing no query, fell back to the home its route had been consolidated away from. Reproduced in a browser against a running build before any change:

Route	Click	Rendered
/?mode=prescribing&q=…&run=1	composer clear	medication-home (the Dose/Safety/Monitoring tiles)
/documents/search?q=…&run=1	composer clear	document-search-empty-state (the "Start here" home)

Both are described as retired by the repository itself — medications/page.tsx says the tiles were retired, documents/page.tsx says the idle Documents view was, and tests/ui-tools.spec.ts asserts in a comment that medication-home is "retired and no longer reachable from any route". It was reachable, by that one click.

The fix is one handler. onClearQuery in ClinicalDashboard.tsx now also drops the submitted search from the URL, so the shared home renders. appModeSelectionHref rather than appModeHomeHref, because every mode reaching this handler is dashboard-owned so its home is the shared home at /, and appModeHomeHref("prescribing") still answers /medications, which only 307s back. /documents/search also has to leave its own route entirely, which the selection href does.

The pre-existing answer guard is kept and is load-bearing: a generated answer is durable state the URL does not own, so the clear button must never discard one. Verified separately in a browser that answer mode is untouched.

Issue 7 fixed. tests/privacy-readiness-contract.test.ts was the third governance suite failing on a shallow clone with reviewedCommit does not exist: <sha>, which reads as a governance breach rather than a missing object. #1M0J6D guarded the other two and left this one. It now mirrors them exactly. The gate is not weakened: on a full clone the commit resolves, checkGit stays true, and the real ancestry check runs.

Issue 6 resolved, but not here. A competing done request (4036f064) was already pending on #ZKR5YK, backed by an implementation of the mode-pill change. Two pending mutations on one row block reconciliation for everyone (#686WHW), so this branch cancels its own re-scope rather than fight over the row.

Three items deliberately not changed

• Issue 3, the duplicate /?mode=tools launcher. Confirmed live by browser probe: /?mode=tools renders tools-home from the legacy client-only launcher while canonical /tools renders ToolsSearchResultsPage. It is real duplication, but removing it is an owner ruling plus a sizeable refactor, not a small safe fix: the mode pill deliberately always returns to the shared home, so special-casing Tools trades one inconsistency for another, and deleting the launcher means re-pointing six assertions in tests/favourites-auth-gate.dom.test.tsx that render it directly. Left for a decision.
• Issue 4, the shared favourites-hub test id. Not a defect — my misread. The three nodes are mutually exclusive branches identifying one logical surface, and expectSingleSettledOwner (a general Next-streaming helper, not a workaround for this) asserts exactly one owner and passes.
• Issue 5, hardcoded /documents and /medications in tools-catalog.ts. Not a defect — also my misread. These render as plain <a href> links to a mode's canonical namespace root, which redirects server-side; that is normal, not the client-side push into a 307 that made the /sources case worth changing. Changing them would break six pinned expectations across ui-tools.spec.ts and search-pins-menu.dom.test.tsx for no user-visible gain.

Flagged, not actioned

appModeHomeHref("prescribing") returns /medications for the bare case while returning /?mode=prescribing for every case that carries a query — an inconsistency inside one function, and the reason the sidebar link, the medication back link and search pins all route through a 307. Fixing it is correct but changes two pinned hrefs for no visible gain, so it is recorded here rather than bundled in.

RAG impact: none — no file under src/lib/rag/, no retrieval RPC, no ranking config, no eval harness or golden fixture is touched.

Verification

• npm run verify:pr-local — 40 gates completed, 0 failed, 0 not reached, including lint, typecheck, test (the full offline suite, which now passes outright because this PR fixes the third shallow-clone suite) and build.
• Mutation-verified. With the fix disabled (if (false && submittedUrlRunRequested)) both new browser cases fail; restored, both pass. The guards are real, not vacuous.
• Browser proof on a production build: tests/ui-tools.spec.ts 99 passed (2.2m) and tests/ui-smoke.spec.ts 108 passed (3.5m) — 207 tests across the two specs this change touches, the two most likely to catch a dashboard regression.
• Reproduction and confirmation done in a real browser, not inferred: counts moved medication-home 1 → 0 and shared-home-empty-state 0 → 1, with the URL landing on /?mode=prescribing&focus=1; the Documents pair moved the same way to /?mode=documents&focus=1.
• Answer mode re-checked in a browser after the change: URL unchanged, 22 answer nodes preserved.

UI verification not run: npm run plan:browser returns level: full, escalating because src/components/ClinicalDashboard.tsx is a shared foundation, so no narrowed run of the whole suite was available and a local full verify:ui would only re-derive the verdict CI reaches. The focused browser proof above covers the changed surface; the complete Chromium suite is left to the Production UI job, which runs because this PR is opened ready for review rather than as a draft.

Risk and rollout

• Risk: low, and confined to one event handler. The only behaviour change is that clearing the composer on a submitted dashboard search now also clears the URL, which is what makes the shared home appear instead of a retired one. Answer mode is explicitly excluded by the pre-existing guard so no generated answer can be discarded. Route-owned modes (/dsm/search, /services/search and the rest) never reach this handler and are untouched, which the 207 passing browser tests bear out.
• Rollback: revert 9b4533a and d8367c6. The ledger commit d3e1b24 is an immutable inbox record and should be left alone. No data, schema, or configuration change, and nothing to unwind on the provider side.
• Provider or production effects: None. No Supabase, OpenAI, migration, environment, or deployment change, and no provider-backed command was run.
• RAG impact: none — no ranking surface is touched.

Clinical Governance Preflight

This change touches one composer event handler and one test guard. It does not alter ingestion, answer generation, search ranking, source rendering, document access, privacy, the production environment, or clinical output. What it removes is a retired page appearing where a clinician expected their search cleared, which is a small improvement to the honesty of the interface rather than a change to any clinical content. Each box below is confirmed unchanged rather than newly assured.

• 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

No clinical decision-support behaviour changed, so the SaMD classification is unaffected.

Notes

Branch naming: this work belongs on claude/vigilant-mayer-60ftd5, whose previous PR (#2672) was merged. The branch was correctly restarted from latest main, but publishing it needed either a force-push or a remote branch deletion, and this environment denies both. With the owner's approval it was published under claude/mode-home-clear-fix instead. The stale claude/vigilant-mayer-60ftd5 still points at ccc6a7d, whose content is already on main via #2672, and can be deleted safely.

🤖 Generated with Claude Code

https://claude.ai/code/session_01QaizDqAQZWHgfGLitWU8xf

---

Generated by Claude Code

---

Note

Low Risk
Single composer clear handler plus test-only shallow-clone guards; answer-present guard unchanged, so generated answers are not cleared by this URL sync.

Overview
Clearing the search composer no longer leaves run=1 (and related query params) in the URL after the React query is emptied. That mismatch had kept showSharedHome off while prescribing and documents modes fell back to retired idle UIs (medication-home, document-search-empty-state) instead of the consolidated shared home. The onClearQuery path in ClinicalDashboard now calls router.replace with appModeSelectionHref when a submitted run was URL-driven, while still bailing out when an answer is on screen.

Tests and CI: Playwright cases lock in prescribing and documents clear-to-shared-home behavior. The privacy readiness contract test gains the same shallow-clone / missing reviewedCommit handling as other governance suites so cloud CI does not fail spuriously without weakening checks on full clones.

Docs inbox: Ledger entries update and withdraw a competing outstanding-issue mutation in favor of another close request; findings are preserved in this PR narrative.

^{Reviewed by Cursor Bugbot for commit d8367c6. Configure here.}

[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: 80ff22d3-789a-45c2-bc77-250ace47c080

---

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 ↗︎.

[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 2026-09-07T11:12:46.510183Z	d8367c6	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.

[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_009cdec3-35ec-4549-a5c5-8d07f370c72f)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d8367c6a71

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".

[BigSimmo]

@codex resolve actionable Codex review findings for this pull request and current head using the repository instructions. This is the pull request's single automatic repair pass: do not perform a fresh review, create new standalone findings, or request another review. Work only the existing unresolved Codex threads on the current head. The only repository destination is BigSimmo/Database, and the only branch destination is the pull request head branch claude/mode-home-clear-fix at starting commit d8367c6; never publish fixes to a detached or synthetic work branch and never create a stacked pull request. Use the authenticated GitHub connector to commit each approved fix to BigSimmo/Database:claude/mode-home-clear-fix, then verify that the pull request head contains the published commit before reporting success. Always fix P0 and P1 findings. For P2 and lower findings, fix only clear, scoped, low-risk issues; otherwise disposition them with a concise reason. For a fixed thread, reply with as the first line and as the second line. For a no-code disposition, use followed by . These result markers authorize the workflow to close that exact thread only after it verifies a fixed commit is the pull request head; a local-only commit is not a fix. If publication or verification fails, do not use either result marker, do not claim success, and leave the thread open with the blocker. If human input or new authorization is required, do the same. Finish only after every actionable thread is fixed or dispositioned and closed, or explicitly left open for a human decision. Do not update the branch from main, address unrelated reviews, broaden scope, or create more than one scoped fix commit. Do not use external APIs, paid services, credentials, dependency changes, or broad refactors unless explicitly authorized. Add targeted tests where behavior changes and run the narrowest relevant validation.

[chatgpt-codex-connector]

Summary

• Updated the composer clear path to abort and invalidate an active Answer request before navigating home, preventing a delayed response from restoring the submitted URL and answer. src/components/ClinicalDashboard.tsxL3318-L3337
• Removed automatic git fetch --deepen behavior from the privacy readiness test. Shallow clones now skip only the unavailable ancestry assertion without network access, while full clones retain validation. tests/privacy-readiness-contract.test.tsL25-L58
• Added a focused delayed-response browser regression proving that clearing an in-flight Answer request remains on the shared home after the response would have settled. tests/ui-universal-search.spec.tsL581-L607
• Committed the changes as 4d85c42ab3e0f213452886a4437aa67fc581786c, pushed them directly to claude/mode-home-clear-fix, and verified that this commit is the head of PR #2730.
• Replied to and resolved both existing Codex review threads using the required fixed-head markers. No stacked or replacement pull request was created.

Testing

• ✅ npm run test -- --run tests/privacy-readiness-contract.test.ts — 8 tests passed.
• ✅ npm run ensure
• ✅ npm run test:e2e -- tests/ui-universal-search.spec.ts --project=chromium --grep "clearing a pending Answer request" — focused Playwright regression passed.
• ✅ npm run format
• ✅ npm run lint:changed:internal -- src/components/ClinicalDashboard.tsx tests/privacy-readiness-contract.test.ts tests/ui-universal-search.spec.ts
• ✅ npm run typecheck:source
• ✅ git diff --check
• ✅ git ls-remote origin refs/heads/claude/mode-home-clear-fix — confirmed remote head 4d85c42ab3e0f213452886a4437aa67fc581786c.
• ✅ gh api graphql -f query='query { repository(owner:"BigSimmo", name:"Database") { pullRequest(number:2730) { url headRefName headRefOid reviewThreads(first:50) { nodes { id isResolved } } } } }' — confirmed the expected PR head and both review threads resolved.

View task →