Switch screenshot workflow to direct Playwright script with mock API
Replace @playwright/test spec with a standalone take-screenshots.mjs
script that uses the playwright package directly. Add mock-api.mjs to
provide a fake OpenAI-compatible /models endpoint so the app can boot
without real API keys. Simplify the workflow to use dev server instead
of Docker build.
Run locally: node e2e/mock-api.mjs & OPENAI_BASE_URL=http://localhost:18080/v1 npm run dev & node e2e/take-screenshots.mjs
https://claude.ai/code/session_01Tt4hLetfAZaAUNpXk21QT3