Dual-LLM Split Brain Chat System

A NodeJS/Express application implementing a dual-LLM "split-brain" architecture where two models collaborate through controlled communication patterns. Built entirely in CoffeeScript with real-time chat interface and Neo4j tool integration.

🧠 Architecture

• Alpha Brain (Left): Analytical, logical, sequential processing
• Beta Brain (Right): Creative, intuitive, pattern recognition
• Corpus Callosum: Inter-LLM communication layer
• Neo4j Integration: Shared knowledge graph

🚀 Quick Start

Prerequisites

1. Node.js 18+
2. Ollama with models:

   ollama pull llama3.1:8b-instruct-q4_K_M
   ollama pull qwen2.5-coder:7b-instruct-q4_K_M

3. Neo4j (Community Edition)

   docker run --name neo4j -p7474:7474 -p7687:7687 -d -e NEO4J_AUTH=neo4j/password neo4j:latest

Installation

# Clone and setup
git clone <your-repo>
cd dual-llm-chat

# Install dependencies
npm install

# Setup Neo4j schema
npm run neo4j:setup

# Start development server
npm run dev

Usage

1. Open http://localhost:3000
2. Select communication mode (Parallel, Sequential, Debate, Synthesis, Handoff)
3. Chat with both AI brains simultaneously
4. Watch the inter-brain communication in real-time

🎛️ Communication Modes

• Parallel: Both models process simultaneously
• Sequential: Alpha then Beta (or vice versa)
• Debate: Models challenge and refine each other
• Synthesis: Combine responses into unified output
• Handoff: One model takes over from the other

🛠️ Development

npm run dev          # Development with auto-reload
npm run compile      # Compile CoffeeScript
npm run test         # Run test suite
npm run lint         # CoffeeScript linting

📊 Features

• Real-time dual-brain chat interface
• WebSocket-based communication
• Neo4j knowledge graph integration
• Multiple orchestration modes
• Inter-LLM communication visualization
• CoffeeScript throughout

🔧 Configuration

Edit .env file for:

• Ollama host and models
• Neo4j connection
• Application settings

📝 License

MIT License - see LICENSE file