Phase 3: Add AI submenu Playwright tests

[workcontrolgit]

Summary

• 4 new page objects under page-objects/:
  • ai-assistant.page.ts — chat UI locators for /ai/assistant
  • ai-hr-insight.page.ts — suggestion buttons, question input for /ai/hr-insight
  • ai-nl-search.page.ts — search input, parsed expression, results table for /ai/nl-search
  • ai-vector-search.page.ts — search input, score badges, results table for /ai/vector-search
• 5 new test files under tests/ai/:
  • ai-navigation.spec.ts — routes accessible, /ai → /ai/assistant redirect, /ai-chat backward-compat redirect
  • ai-assistant.spec.ts — page loads, disabled banner, chat UI, send/clear
  • ai-hr-insight.spec.ts — page loads, suggestion buttons, ask a question
  • ai-nl-search.spec.ts — page loads, search input, parsed expression after query, clear
  • ai-vector-search.spec.ts — page loads, search input, score badges, clear
• Updated tests/screenshots/blog-screenshots.spec.ts — replaced stale /ai-chat tab-based screenshots with new per-page screenshots for /ai/assistant, /ai/hr-insight, /ai/nl-search, /ai/vector-search

Test design notes

• All tests use test.skip() gracefully when aiEnabled is false — CI passes without Ollama
• AI-dependent tests (send message, get response) use extended timeouts (30–90s) for Ollama inference
• Navigation tests verify both new routes and backward-compat redirect

Test plan

• Run npx playwright test tests/ai/ with full stack running — navigation tests should pass
• With aiEnabled: false — all AI page tests skip gracefully (no failures)
• With aiEnabled: true + Ollama — chat/search tests execute and verify responses

[github-actions]

🎭 Playwright Test Results

Browser	Status
chromium	❌ Failed
firefox	❌ Failed
webkit	❌ Failed

📊 View detailed report