A voice-first English speaking tutor you can install as a web app. You talk to it out loud, it answers with real speech, shows the live transcript, and gives you a short pronunciation + grammar score after every turn. Built on the Gemini Live API.
npm install # runtime dependencies: `ws` and `tsx` (runs server/*.ts, no build step)
cp .env.example .env # add your GOOGLE_API_KEY
npm start # http://127.0.0.1:8080Open that URL in a browser and tap the microphone. (Microphone access needs a
secure context — localhost or HTTPS. To try it on your phone, serve it over
HTTPS or use a tunnel.)
npm test # unit tests, no network
npm run e2e # real round trip against the Gemini Live API
npm run browser-check # headless Chrome driving the real page with a fake micMore detail: docs/DEVELOPING.md — environment variables, model fallbacks, testing, and the cache/versioning rules for shipping an update. Design system: DESIGN.md. Want to help? CONTRIBUTING.md.
