LLM Roundtable

A Council of LLMs visualized as Knights of the Round Table. Multiple AI models deliberate on your questions through a 3-stage process inspired by Karpathy's llm-council.

How It Works

1. Initial Responses - Each knight (LLM) provides their perspective on your question
2. Peer Review - Knights anonymously rank each other's responses
3. Synthesis - King Arthur synthesizes all perspectives into a final verdict

The Knights

Knight	Model	Personality
King Arthur	llama-3.3-70b-versatile	Wise ruler who synthesizes viewpoints
Sir Lancelot	openai/gpt-oss-120b	Strategic, bold, action-oriented
Sir Gawain	qwen/qwen3-32b	Diplomatic, balanced perspectives
Sir Percival	moonshotai/kimi-k2-instruct	Idealistic, spiritually-minded
Sir Galahad	meta-llama/llama-4-scout-17b-16e-instruct	Morally rigorous, principled

Features

• 3D visualization of knights around the Round Table
• Visual feedback showing which knight is currently "speaking"
• Rankings showing which knight's response was most valued by peers
• Expandable responses to see each knight's full answer
• Final synthesized answer from King Arthur

Tech Stack

• React Native + Expo
• React Three Fiber (3D rendering)
• Groq API (LLM inference)
• TypeScript

Getting Started

1. Clone the repo

2. Install dependencies

   npm install

3. Create .env file with your Groq API key

   EXPO_PUBLIC_GROQ_API_KEY=your_api_key_here
   

4. Start the app

   npx expo start

Get a Groq API Key

Sign up at console.groq.com to get your free API key.

License

MIT