To set up

• Copy template-.env to .env
• Get a Gemini API key and add it to .env as GEMINI_API_KEY.
• Get an ElevenLabs API key and add it to .env as ELEVENLABS_API_KEY.
• Run npm install to install packages.

To run

Backend:

npm run dev:server

Frontend:

npm run dev

Testing & Building

# Run tests
npm test

# Lint code
npm run lint

# Build for production
npm run build

# Start production server (serves built files)
npm start

Deployment

Both compose setups run their own Y-Sweet with auth on, and refuse to start without Y_SWEET_AUTH and Y_SWEET_SERVER_TOKEN in .env (generate them as a pair; see template-.env). Compose builds the app's connection string from the token, so YSWEET_CONNECTION_STRING in .env is ignored there.

Local dev with Docker

# Auto-merges compose.yaml + compose.override.yaml (uses localhost URLs)
docker compose up -d

Production

# On the server
./deploy-docker-compose.sh

It runs git pull, builds, and starts with compose.prod.yaml (which sets the public y-sweet URL). deploy-flyio.sh is the Fly.io equivalent.

Other commands

# View logs
docker compose logs -f

# Stop
docker compose down

Environment variables are loaded from .env automatically.

Project Details

This is a live translation application for presentations/talks. It provides:

• AI-powered translation (Google Gemini)
• Text-to-speech output (ElevenLabs)
• Collaborative editing (Y-Sweet/Yjs)
• Multiple layout configurations, chosen by URL — see docs/LAYOUT_URLS.md

See CLAUDE.md for detailed architecture and development information.