Voice DevTools

This UI provides a debug console for real-time AI voice interactions. It works with multiple realtime models (View supported models). Features include:

1. Cost Tracking: Know how much you’ve spent per voice interaction
2. Model Support: Supports open-source (MiniCPM-o) and closed-source S2S models like OpenAI’s GPT-4o Realtime (adding more soon!)
3. Voice and Chat UI
4. Session history and recording

Inspired by openai-realtime-console and openai-realtime-agents.

Quick Start

1. Get your API keys:

   • Outspeed API key
   • OpenAI API key

2. Set up environment:

   cp .env.example .env
   # Add your API keys to .env:
   # OPENAI_API_KEY="<your-openai-key>"
   # OUTSPEED_API_KEY="<your-outspeed-key>"
   

3. Install and run:

   npm install
   npm run dev
   

Visit http://localhost:3000 to access the console.

Usage

To modify agent prompt and tools, modify agent-config.js.

To modify the model parameters like (voice, version, etc.), edit model-config.js

Agents

There are three voice agent examples already present. You can choose them in the Session Config UI on the right before starting a session.

1. Dental Agent: Answers callers’ questions about working hours of a dental clinic
2. Message Agent: Takes callers’ messages for a person
3. Recruiter Agent: Talks to a candidate and asks questions about their background and availability

You can see the prompts in ./src/agent-config.ts file.

Supported Models

• MiniCPM-o (hosted by Outspeed)
• OpenAI Realtime models
• Moshi (Coming Soon)
• Gemini Multimodal Live (Coming Soon)

Deployment

You can see the deployment your agent to Cloudflare by following the steps at demo.outspeed.com/deploy, or you could run this locally and visit the deploy route.

License

MIT