Cue helps students first understand what they want to say, then practice saying it well.

• frontend/: Vite + React mobile-in-web app.
• backend/: Express API with Gemini, Google STT/TTS, and mock fallbacks.

Backend:

cd backend
cp .env.example .env
npm install
npm run dev

Frontend:

cd frontend
cp .env.example .env
npm install
npm run dev

Open the frontend dev URL and keep VITE_API_BASE_URL pointed at http://localhost:8787/api.

Cue V0.4 runs without external credentials. Leave GEMINI_API_KEY and GOOGLE_APPLICATION_CREDENTIALS empty in backend/.env.

Mock mode keeps the full demo clickable:

• Learn start/message
• Bridge recap generation
• Speak preparation
• Practice submission
• Review generation
• Take 2
• Audio preview data URLs

Put real backend credentials in backend/.env:

GEMINI_API_KEY=your_key
GEMINI_MODEL=gemini-3-flash-preview
GOOGLE_APPLICATION_CREDENTIALS=/absolute/path/to/service-account.json

Do not commit .env or service account files.

Visible modes:

• Learn
• Speak

Hidden transition:

• Bridge

Core flows:

Home -> Learn -> Bridge -> Speak Prep -> Speak Practice -> Speak Review
Home -> Speak Prep -> Speak Practice -> Speak Review

Language concepts are separate:

• uiLanguage: app interface language.
• sourceLanguage: learner thinking/learning language.
• targetLanguage: speaking practice language.