fix(ui): fix transcript-strip pill gap in WebKit

[mariusvniekerk]

Summary

• Fix transcript-strip pill backgrounds not filling the container height in WebKit (Tauri app)
• Add WebKit as a Playwright e2e test target since the app ships as a Tauri app
• Fix virtualizer measurement test flake in WebKit (row height coincidentally matching estimate)

Root cause

WebKit doesn't resolve height: 100% on flex children when the parent only has min-height.
The pills rendered at 16px inside a 26px container, leaving an 8px gap.
Fix: change min-height: 26px to height: 26px on .transcript-strip.

Before / After (WebKit, Playwright)

Before — pills don't fill container height:

After — pills fill full container height:

Test plan

• Playwright e2e passes in Chromium (66/66)
• Playwright e2e passes in WebKit (66/66)
• CI installs webkit browser (npx playwright install --with-deps chromium webkit)
• Verify in Tauri app visually

🤖 Generated with Claude Code

[roborev-ci]

roborev: Combined Review (8dd0674)

Verdict: No medium-or-higher findings; the changes look good to merge.

---

Synthesized from 3 reviews (agents: codex, gemini | types: default, security)

[roborev-ci]

roborev: Combined Review (e932a3b)

Verdict: Changes introduce one test-breaking issue and one meaningful test coverage regression risk.

High

• frontend/e2e/transcript-strip.spec.ts:1
  Invalid leading space in the Playwright import path (" @playwright/test") will cause module resolution failure and break the E2E test file.
  Fix: Change it to import { test, expect } from "@playwright/test";

Medium

• frontend/e2e/virtualizer.spec.ts:156
  The revised assertion now passes when any single rendered row differs from ESTIMATE_PX, which weakens the original guarantee that rows were actually remeasured. Because it still depends on exact float equality, a minor rounding difference can make the test pass even if most rows remain stuck at the estimate.
  Fix: Assert remeasurement directly with a small epsilon and fail when too many rows remain at the estimated height.

---

Synthesized from 3 reviews (agents: codex, gemini | types: default, security)

[mariusvniekerk]

Given that the test runs the high priority review is clearly imvalid

[wesm]

agreed. merging this, thank you!