AI Agent Interaction System

A multi-agent system that enables collaborative problem-solving through specialized AI agents. The system features a directory service for agent discovery and communication, with persistent memory storage using ChromaDB.

Features

• Multi-Agent Architecture: Multiple specialized agents working together
• Directory Service: Central service for agent discovery and message routing
• Persistent Memory: ChromaDB-based memory system for agents
• Interactive Interface: Command-line interface for direct agent interaction
• Tool Integration: Extensible tool system for agent capabilities
• Health Monitoring: Built-in health checks and status monitoring
• Logging System: Comprehensive logging with configurable levels

Prerequisites

• Python 3.x
• Poetry (Python dependency management)
• Required API keys:
  • OpenAI API key
  • Tavily API key (for web search capabilities)

Setup

1. Clone this repository

2. Install dependencies using Poetry:

   poetry install

3. Create a .env file based on .env.sample:

   OPENAI_API_KEY=your_openai_key_here
   TAVILY_API_KEY=your_tavily_key_here
   

Running the System

The system requires two terminals to run:

1. Start the main system (directory service and agents):

   poetry run python src/main.py

2. Launch the interaction interface:

   poetry run python src/live_interact.py

Available Agents

• MathAgent: Mathematical expert for calculations and problem-solving
• TimeAgent: Time and timezone specialist
• EditorAgent: Expert editor for proofreading and editing blog posts

Interactive Commands

• /lookup - List all available agents
• @agent_name message - Send direct message to agent
• @agent_name/status/new_status - Update agent status
• /show <collection_name> - Show documents in collection
• /list_col - List all collections
• /list_status - List all agent statuses
• /clear - Clear the screen
• /help - Show help message

API Documentation

• API specifications and OpenAPI documentation available at:
  • Directory Service: http://127.0.0.1:8000/docs
  • Individual Agent APIs: Available at their respective ports

Memory Management

The system uses ChromaDB for persistent storage with two types of memory:

• Short-term memory: Recent interactions and temporary data
• Long-term memory: Consolidated and processed information

Development

Environment Variables

• AGENT_DEBUG: Enable debug mode (default: false)
• AGENT_LOG_LEVEL: Set logging level (default: INFO)
• SERVER_LOG_LEVEL: Set server logging level (default: INFO)
• CONSOLE_LOGGING: Enable console logging (default: True)

Adding New Agents

Create new agent files in src/agents/ following the existing patterns:

'''python def create_new_agent(chroma_client: PersistentClient) -> BaseAgent: config = AgentConfig( agent_name="NewAgent", description="Description of the agent's capabilities", enabled_tools=["list_of", "enabled", "tools"], api_port=unique_port_number ) return BaseAgent(api_key=os.getenv("OPENAI_API_KEY"), chroma_client=chroma_client, config=config)

## License

This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.