LiveKit Voice Agent with Next.js

A full-stack voice AI application built with LiveKit Agents, Next.js, and Rime TTS. Real-time voice conversations powered by GPT-5.2, with high-quality TTS and advanced speech recognition.

Deploy to production on Render with auto-scaling. Develop locally with Docker.

• Deploy to Production (Render)
• Features
• Repository Structure
• Local Development with Docker
• Local Docker Commands
• Customization
• Environment Variables
• Troubleshooting
• Additional Documentation
• Contributing
• License

🚀 Deploy to Production (Render)

This project is pre-configured for production deployment to Render using the included render.yaml blueprint.

What you get:

• ✅ Next.js web app with voice UI
• ✅ LiveKit voice agent service
• ✅ Auto-scaling and health checks
• ✅ Separate app and agent services
• ✅ Environment variable management

Deployment steps:

1. Fork this repository to your GitHub account

2. Set up LiveKit Cloud:

   • Sign up at LiveKit Cloud
   • Create a new project
   • Get your API key, secret, and WebSocket URL

3. Create environment variable group on Render:

   • Go to Render Dashboard
   • Create an environment group with these variables:
     • LIVEKIT_API_KEY - Your LiveKit API key
     • LIVEKIT_API_SECRET - Your LiveKit API secret
     • LIVEKIT_URL - Your LiveKit WebSocket URL
     • RIME_API_KEY - Get from Rime.ai
     • (LLM and STT use LiveKit Inference—billed via LiveKit Cloud—so OpenAI and AssemblyAI keys are not required.)

4. Create a new Blueprint Instance on Render:

   • Click "New" → "Blueprint Instance"
   • Connect your forked repository
   • Select branch (usually main)
   • Select the environment group you created

5. Deploy:

   • Click "Apply" to create all services
   • Render will:
     • Build and deploy Next.js app
     • Build and deploy LiveKit agent
     • Link services together

6. Access your deployed app:

   • Web App: https://your-app-name.onrender.com
   • Agent will auto-connect via LiveKit Cloud

Important: Agent Resource Requirements

⚠️ The agent service requires at least 8 GB of RAM to load AI models (VAD, turn detection).

On Render, use the Standard plan or higher for the agent worker service. The included render.yaml is already configured with auto-scaling and health checks for production workloads.

✨ Features

• 🎙️ Real-time Voice Conversations - Natural voice interactions powered by LiveKit
• 🧠 Smart AI Agent - Uses GPT-5.2 for intelligent responses
• 🗣️ High-Quality TTS - Rime TTS with multiple voice options
• 🎧 Advanced STT - AssemblyAI for accurate speech recognition
• 🔇 Noise Cancellation - Built-in background noise reduction
• 🚀 Production-Ready Deployment - Pre-configured for Render with auto-scaling and Docker

📁 Repository Structure

render-voice-agent/
├── agent/                  # LiveKit voice agent
│   ├── agent.ts            # Agent implementation (GPT-5.2 + Rime + AssemblyAI via LiveKit Inference)
│   ├── AGENT_README.md    # Agent configuration and customization
│   ├── Dockerfile         # Production agent container
│   └── Dockerfile.dev     # Development agent container
├── app/                   # Next.js application
│   ├── api/              # API routes (LiveKit token generation)
│   ├── components/       # React components
│   │   └── VoiceAgent.tsx # Voice UI component
│   ├── page.tsx          # Main application page
│   └── globals.css       # Global styles
├── docker-compose.yml     # Full-stack local development (app + agent)
├── render.yaml           # Production deployment configuration (Render.com)
├── package.json          # Dependencies and scripts
└── README.md

💻 Local Development with Docker

Requirements for local development:

• Docker & Docker Compose
• At least 8 GB of RAM available for the agent (for AI model loading)
• Node.js 20 or higher (optional, for running without Docker)
• LiveKit Cloud account (sign up)
• Rime API key for TTS (LLM and STT use LiveKit Inference; no OpenAI or AssemblyAI keys needed)

Local setup:

1. Clone the repository:

git clone <your-repo-url>
cd render-voice-agent

2. Set up LiveKit Cloud credentials:

# Authenticate with LiveKit Cloud
lk cloud auth

# Generate credentials and save to .env
lk app env -w

This creates a .env file with:

LIVEKIT_API_KEY=...
LIVEKIT_API_SECRET=...
LIVEKIT_URL=wss://...

3. Add additional API key to .env:

RIME_API_KEY=your-rime-api-key

4. Start all services locally with Docker:

docker-compose up

This will:

• ✅ Automatically download AI model files during first build
• ✅ Start Next.js app on localhost:3000
• ✅ Start LiveKit voice agent

☕ First time? The initial build takes 2-3 minutes to download AI models.

5. Access your local application:

Open http://localhost:3000 in your browser and start a voice conversation!

Ready for production? See Deploy to Production (Render) above.

📦 Local Docker Commands

Start all services locally:

docker-compose up -d

Stop all local services:

docker-compose down

View local logs:

# All services
docker-compose logs -f

# Specific service
docker-compose logs -f app
docker-compose logs -f agent

Rebuild after code changes:

docker-compose up -d --build

Alternative: Run without Docker (local development):

# Step 1: Install dependencies and download AI models
npm install
npm run agent:download

# Step 2 (Terminal 1): Run the Next.js app
npm run dev

# Step 3 (Terminal 2): Run the LiveKit agent
npm run agent:dev

🎨 Customization

Change the Voice

Edit agent/agent.ts:

tts: new TTS({
  model: 'mistv2',
  voice: 'lagoon', // Try: rainforest, lagoon, astra
}),

Available voices: rainforest, lagoon, astra, and more. See Rime TTS docs.

Modify Agent Personality

Update the instructions in the Assistant class in agent/agent.ts:

instructions: `You are a friendly and helpful voice assistant...`

Tech Stack Details

Frontend:

• Next.js 16 - React framework
• LiveKit Components React - Pre-built voice UI components
• Tailwind CSS - Styling

Agent:

• LiveKit Agents - Voice agent framework
• Rime TTS - Text-to-speech
• OpenAI GPT-5.2 (via LiveKit Inference) - Language model
• AssemblyAI (via LiveKit Inference) - Speech-to-text
• Silero VAD - Voice activity detection

Infrastructure:

• Docker - Containerization
• Render - Cloud hosting
• LiveKit Cloud - Real-time communication platform

🔧 Environment Variables

Required environment variables (.env):

# LiveKit Cloud (get with: lk cloud auth && lk app env -w)
LIVEKIT_API_KEY=...
LIVEKIT_API_SECRET=...
LIVEKIT_URL=wss://...

# API Keys
# Required for TTS (Rime)
RIME_API_KEY=...          # Get from https://rime.ai/

# Optional: only if you switch from LiveKit Inference to provider plugins
# OPENAI_API_KEY=...     # For OpenAI plugin (LLM)
# ASSEMBLYAI_API_KEY=... # For AssemblyAI plugin (STT)

🐛 Troubleshooting

1. Agent won't start:

# Ensure LiveKit and Rime keys are set (OpenAI/AssemblyAI optional when using LiveKit Inference)
cat .env

# Download model files
npm run agent:download

# Check LiveKit Cloud authentication
lk cloud auth

2. No audio in browser:

• Ensure microphone permissions are granted
• Check browser console for errors (F12)
• Verify LiveKit Cloud connection in network tab
• Try a different browser (Chrome recommended)

3. Local Docker build fails:

# Ensure Docker is running
docker ps

# Try rebuilding from scratch locally
docker-compose build --no-cache

# Check local Docker logs
docker-compose logs agent

4. Local Docker port conflicts:

# Check what's using the port
lsof -i :3000

# Change port in docker-compose.yml if needed

5. Agent connection errors:

• Verify LIVEKIT_URL is correct (should start with wss://)
• Check LiveKit Cloud project is active: https://cloud.livekit.io/
• Ensure API keys match the LiveKit project
• Check agent logs: docker-compose logs agent

6. API key errors:

• Verify all API keys are valid and active
• Check API key permissions and quotas
• Ensure keys are properly set in .env
• For Render deployment, verify environment variables are set in the Render dashboard

Getting more help:

1. Local development: Check logs with docker-compose logs <service-name>
2. Production issues: Check the Render dashboard for logs and metrics
3. Verify all environment variables are set correctly (.env locally, Render dashboard for production)
4. See detailed troubleshooting in QUICK_START.md
5. Join the LiveKit Discord community

📚 Additional Documentation

• QUICK_START.md - Comprehensive local development guide with Docker
• agent/AGENT_README.md - Agent configuration and customization
• Render Documentation - Production deployment guides

External Resources:

• LiveKit Documentation
• LiveKit Agents Guide
• Rime TTS Documentation
• LiveKit Cloud Dashboard

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

MIT