kefu/deploy/review-assistant-20260917/UI-VALIDATION.md

Medical assistant reminder UI verification

The Qt/HTML files and recipient helper are complete in both directories, preserving protocol capability differences.

• Production: 173 tests, 0 failures/errors/skips, 0 blocked external action attempts.
• Source: 172 tests, 0 failures/errors, 1 expected skip (no protocol transport), 0 blocked external action attempts.
• New UI/contact/settings tests: 27 per directory.
• Reports: production-ui-tests.json/.log, source-ui-tests.json/.log.
• Runner: C:/kefu/deploy/loading-performance-20260917/run_isolated.py. It uses temporary application data and blocks network, clipboard, screenshot, and WeChat interaction.
• Syntax checks: Python AST and node --check on both UI scripts.
• Actual Chrome headless rendering of original local HTML with synthetic state: assistant-settings-production-chrome.png, assistant-preview-production-chrome.png, assistant-settings-source-chrome.png, assistant-preview-source-chrome.png. No JavaScript page errors. The source-capable screenshot intentionally uses synthetic state to test rendering; actual source native capability is disabled.
• Qt offscreen texture screenshots are unreliable in this environment (100x30 initial texture); the Chrome previews above are the visual evidence.

Behavior checked: disabled by default, exact account/recipient validation, duplicate contact identities, no nickname routing, no alternate-account fallback, only existing M: private WeChat conversations, read-only database access, stale worker suppression, cancel safety, account-switch rejection, per-account persistence, rollback on disk error, recipient replacement disables reminder, no real test-send action, static example, user-triggered read-only notification summary.

Owned files: wechat_gui_qt.py, assets/ui/console-app.js, review_assistant_contacts.py, test_review_assistant_ui.py. Shared coordinator and visual/protocol engines are owned by other agents.