From f9fc38bc0ffd5925e20c655bf0f329efe7247c4d Mon Sep 17 00:00:00 2001 From: BigSimmo <87357024+BigSimmo@users.noreply.github.com> Date: Tue, 11 Aug 2026 23:36:21 +0800 Subject: [PATCH 1/4] Improve mobile evidence review clarity --- docs/branch-review-ledger.md | 1 + .../answer-result-surface.tsx | 2 +- .../clinical-dashboard/evidence-panels.tsx | 9 +++-- .../clinical-dashboard/visual-evidence.tsx | 38 +++++++++++++++---- tests/visual-evidence-tabs.dom.test.tsx | 9 +++++ 5 files changed, 47 insertions(+), 12 deletions(-) diff --git a/docs/branch-review-ledger.md b/docs/branch-review-ledger.md index e14e3b4c94..db2eb1cfd3 100644 --- a/docs/branch-review-ledger.md +++ b/docs/branch-review-ledger.md @@ -882,3 +882,4 @@ Records before 2026-07-28 were written by hand and had drifted: 146 lines carrie | 2026-08-10 | PR #1803 / claude/codex-m4b-shadow-tight-migration-53a8kn | b778a56e9c3fa7642a783dde85e1130559d71e24 | shadow-tight token migration onto the e1 elevation tier and alias retirement (#262 part 1) | Migrated all 150 var(--shadow-tight) occurrences across 71 files to var(--e1) (90 gated production sites across 48 files, 60 mockup); deleted all three alias declarations (:root, .dark, forced-colors); pinned legacyShadowAliases 220 to 127 with exact per-path counts, closing 3 aliases of re-accumulated slack; added a whole-stylesheet absence assertion (mutation-verified); updated GATES.md section 3 plus a new section 6, TOKENS.md section 6, design-system.md, both redesign direction docs, .design-sync/conventions.md and ledger #262. Verified in Chromium that the ckb-v2 tier override is picked up by the alias substitution, so the change is value-preserving; that check is recorded as a prerequisite for the remaining six aliases. | npm run verify:cheap (30 static gates plus lint plus typecheck green; design-system contract passed, legacy shadow aliases 127; unit suite 553/554 files, 6024 tests passed, 1 pre-existing root-permission failure in tests/pr-handoff-stop.test.ts reproduced on untouched base a16dd26); npm run format:check whole tree; targeted Chromium computed-style measurement. verify:ui not run, Playwright browser revision drift #255, delegated to CI Production UI. No provider-backed gates. | | 2026-08-11 | codex/answer-loading-ui-20260811 | 6758f8156f9d1b3e893981dfd7a1f6563aa90da0 | answer creation loading UI | No high-confidence findings | UI 3 passed; unit 8 passed; lint, typecheck, build, design-system and offline RAG passed; full suite 6022 passed with 16 unchanged baseline failures | | 2026-08-11 | claude/codex-m4c-retire-shadow-nliak3 | 448a0d084c4cd2cda6153dd7f03dcb67c43a8df0 | DS Track A2 (#261): retire --shadow-focus; composer focus onto sanctioned outline; contract guard; baseline ratchet; design-system docs + ledger | Approved — PR #1807. Token deleted in both themes; .chat-composer-shell-delta:focus-within uses outline 2px var(--focus) at offset 2px and no longer overrides box-shadow. Reach premise corrected: 0 of 37 production routes render the class (only /mockups/calculators-search). legacyShadowAliases 127->125, globals.css pin 3->1. | check:design-system-contract PASS; design-token-contract.test.ts PASS + mutation-verified both ways; verify:pr-local PASS except pre-existing tests/pr-handoff-stop.test.ts failure baselined on untouched base e8b61d8; build PASS; check:rag:fixtures PASS (36 cases); Chromium look both themes on the mockup route (inspection only, rev 1194 vs pinned 1234 #255); verify:ui/verify:phone-chrome NOT run — delegated to CI | +| 2026-08-11 | work | 45fd05c8c3947835c0368666ff576c7a38b33ee4 | mobile evidence sheet UX, accessibility, and feedback logic | fixed unexplained claim marker, excess panel reserve, unclear purpose and feedback copy; no remaining high-confidence defects | focused DOM 7/7; Chromium evidence journey 1/1; offline RAG 23 suites/574 tests | diff --git a/src/components/clinical-dashboard/answer-result-surface.tsx b/src/components/clinical-dashboard/answer-result-surface.tsx index e8576a4d1a..c485a3a8e5 100644 --- a/src/components/clinical-dashboard/answer-result-surface.tsx +++ b/src/components/clinical-dashboard/answer-result-surface.tsx @@ -393,7 +393,7 @@ function StagedAnswerResultSurfaceImpl({ open={evidenceOpen} onClose={closeEvidenceReview} title="Evidence" - description="Review by evidence type." + description="Check how well sources support this answer." titleAccessory={{evidenceTrustLabel}} closeLabel="Close evidence" headerLeading={ diff --git a/src/components/clinical-dashboard/evidence-panels.tsx b/src/components/clinical-dashboard/evidence-panels.tsx index a180b1927c..897d53ae11 100644 --- a/src/components/clinical-dashboard/evidence-panels.tsx +++ b/src/components/clinical-dashboard/evidence-panels.tsx @@ -1152,9 +1152,10 @@ export function AnswerFeedbackPanel({ >
Answer review
+Is the answer supported?
- Capture misses for retrieval and RAG evals without changing the answer. + Record whether the linked evidence supports the answer. This sends feedback for review; it does not change + the answer.
Claims checked
+ Check which claims are backed by the retrieved sources, open the supporting passages, and flag evidence problems + before relying on the answer. +