Resolve PsychSift visual-audit findings and preserve workflow accessibility

Merged
BI
1mo ago
+2926-1679

Summary

Resolve the settled PsychSift visual-audit findings across shared tokens and control roles, compact Clinical chrome, responsive navigation, catalogue pages, First Nations reference context, Work search, and accessible error recovery. Preserve the agreed typography, mode identities, 48px target contract, separate disclosures, reading widths, and clinician hold on the bedside calculator panel.

The implementation and evidence handoff accounts for all 32 original findings and eight assurance IDs, including explicit manual-acceptance limits.

Adversarial review repaired stale Services shortlist identifiers that blocked valid additions and Therapy zero-result announcements hidden behind an inert scenario-editor background. Services preserves synchronous URL/query state, replaces same-view shortlist history, and pushes genuine view transitions and explicit Clear. Therapy uses one accessible singleton, deduplicates delivered counts, announces initial zero/nonzero outcomes, and resumes pending announcements after loading interruptions. Independent Standards and Spec reviews and corrective reviews found no outstanding confirmed issue.

Areas touched: App experience, Design system, Knowledge and records, Personal practice, Clinical reference content, Data platform and access, Delivery and assurance, Prototypes

Verification

  • Hosted CI on 24337a5439c9ba71bba90db8cc2e7873c32d2239: PR required succeeded. Static checks, both coverage partitions and their aggregate, production builds, critical UI and all three production UI shards, safety/config checks, and Lighthouse passed. Secret Scan and SAST also passed. CI run.
  • Regression-first proofs reproduced the interaction defects before repair. Latest affected owning suites: 6 passed, 37 passed (6.66s), including real Sheet inert containment, stale/mixed shortlist recovery, native Back/Forward, and delivered versus interrupted announcement timing.
  • Fresh local production build, including complete TypeScript, on code commit 58ea6a9964e9429c8c94616dea1d11dc1a0d65b3; complete owning ui-tools browser spec: 104 passed (3.2m), including stale-bookmark recovery at 390px/1280px. Subsequent Services formatting emits identical minified JavaScript. Earlier full local suites and their exact evidence versions remain in the handoff.
  • Earlier full local npm run verify:ui stage on implementation commit 0022c8121463669f222de38f642089a87e2ea932: 765 passed, one iOS-only skip. This is versioned baseline evidence; current-head comprehensive proof is the hosted CI result above.
  • Whole-repository npm run format, final changed-file formatting, scoped ESLint, complete design-system contract and immutable-review checks passed. Adoption: 55 components / 235 roots; design sync: 55 components / 7 guidelines; diff integrity: 47 changed test files, 741 -> 797 cases, self-test 27/27.
  • Repository-awareness snapshot freshness: in step with data/repo-awareness-snapshot.json (320 pages, 638 documents, 2803 reviews). The generated catalogue includes the new audit decision/handoff and review records.
  • Lighthouse: Every graded route is within tolerance of the committed baseline. Desktop Documents-search LCP measured 944ms against a 961ms baseline; the budget and baseline are unchanged.

Remaining unobserved acceptance: physical Safari, actual installed PWA, virtual keyboards, comprehensive assistive-technology/zoom/contrast coverage, and remaining provider-dependent populated clinical outcomes. The release browser matrix was skipped for this event; no manual release-matrix workflow was dispatched. Original Windows screenshots and raw audit telemetry were not supplied. Provider-backed clinical evals were not run because this change alters presentation and interaction, not retrieval or clinical decision-support logic.

Risk and rollout

  • Risk: shared control and chrome changes affect many routes. Current required CI and local owning journeys pass; remaining manual/device acceptance stays explicit. The bedside calculator panel remains hidden pending clinician verification.
  • Rollback: revert the PR as a unit; token compatibility rollback declarations remain available. No database migration or provider configuration change is included.
  • Provider or production effects: publishing this feature branch and PR triggers configured GitHub checks. This task does not merge, deploy, change provider configuration, or arm auto-merge.

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 PsychSift Production (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

These checks concern the source diff: no clinical content, retrieval/ranking, permissions, credentials, database target, or clinical decision-support logic changes. New fixtures remain synthetic. The conditional deployment-classification item is satisfied by the absence of a clinical decision-support behavior change; this is not clinical, cultural, or regulatory approval. Existing review and verification requirements remain in force.

Notes

All 40 IDs are accounted for without unconditionally closing the manual assurance tasks. Immutable adversarial review records are included; historical issue and review ledgers are preserved. Current hosted results above supersede the publication-time pending statuses in the versioned handoff.