AI Agent Council

A collaborative decision-making platform where multiple AI agents discuss and reach consensus on user queries.

Project Overview

The AI Agent Council is a system where:

1. Users can submit queries through a web interface
2. Multiple AI agents independently analyze the query
3. Agents engage in a discussion to share perspectives
4. Agents collaboratively reach a consensus
5. The final decision is presented back to the user

The system consists of:

• A Next.js web application with a modern UI
• Python-based AI agents that can be easily configured
• Ollama integration for local LLM inference using Llama3:8b

Installation

Prerequisites

• Node.js 18+ and npm/yarn
• Python 3.10+
• Ollama installed locally
  • Install on MacOS (Using Homebrew)

  $ brew install ollama

Setup

1. Clone the repository:

   git clone https://github.com/ninjonas/ai-council.git
   cd ai-council
   

2. Set up the Next.js application:

   cd web
   npm install
   npm run dev
   

3. Set up the Python agents:

   cd ../agents
   pip install -r requirements.txt
   python server.py
   

4. The web interface will be available at http://localhost:3000

Adding New Agents

To add a new agent:

1. Create a new folder in the agents/agent_instances directory:

   mkdir -p agents/agent_instances/your_agent_name/references
   

2. Configure the agent by creating a config.yml in the new folder:

   name: "Your Agent Name"
   description: "What this agent specializes in"
   personality: "How this agent should communicate"
   expertise: ["area1", "area2"]
   temperature: 0.7
   max_tokens: 1024
   system_prompt: |
     As Your Agent Name, your approach should be...

3. Add reference materials (optional):

   • Place PDF, Markdown or text files in the agents/agent_instances/your_agent_name/references folder
   • Agents will automatically incorporate these materials in their reasoning

4. Restart the agent server:

   cd agents
   python server.py
   

System Architecture

• Web Application: Next.js with TypeScript and Tailwind CSS
• Agent System: Python-based with WebSockets for real-time communication
• LLM Integration: Ollama with Llama3:8b for efficient local inference
• Data Exchange: Standardized JSON protocols for agent communication

Configuration Options

System Instructions

Users can provide system-wide instructions that all agents will follow during their discussion. This is optional and can be left blank.

Agent Configuration

Each agent's config.yml supports:

• name: Agent identifier
• description: Purpose description
• personality: Communication style
• expertise: List of specialization areas
• temperature: Creativity level (0.0-1.0)
• max_tokens: Maximum response length
• system_prompt: Default instructions for this agent

License

MIT