multi-swarm 1.0.0

A framework for creating collaborative AI agent swarms

Multi-Swarm Framework

A powerful framework for creating collaborative AI agent swarms, leveraging multiple LLM providers including Claude and Gemini.

Features

• Create collaborative agent swarms with distinct roles and capabilities
• Support for multiple LLM providers (Claude and Gemini)
• Easy-to-use agent template creation
• Flexible agency configuration
• Built-in tools system
• Asynchronous communication between agents

Installation

pip install multi-swarm

For development installation with testing tools:

pip install multi-swarm[dev]

Quick Start

1. Set up your environment variables:

# .env
ANTHROPIC_API_KEY=your_claude_api_key
GOOGLE_API_KEY=your_gemini_api_key

2. Create a simple agency:

from multi_swarm import Agency, BaseAgent

# Create custom agents
class MyAgent(BaseAgent):
    def __init__(self):
        super().__init__(
            name="MyAgent",
            description="A custom agent for specific tasks",
            instructions="path/to/instructions.md",
            tools_folder="path/to/tools",
            model="claude-3-sonnet",
            temperature=0.7
        )

# Initialize agents
agent1 = MyAgent()
agent2 = MyAgent()

# Create agency with communication flows
agency = Agency(
    agents=[
        agent1,  # Entry point for user communication
        [agent1, agent2],  # agent1 can communicate with agent2
    ],
    shared_instructions="agency_manifesto.md"
)

# Run the agency
agency.run_demo()

Examples

Check out our example implementations in the examples directory:

1. Google Trends Analysis Agency: Shows how to create a specialized agency for analyzing Google Trends data. This example demonstrates:
   • Custom agent creation (CEO and TrendsAnalyst)
   • Communication flow setup
   • Agent instructions and tools structure
   • Basic agency configuration

Documentation

For detailed documentation, please visit our GitHub repository.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request. Here are some ways you can contribute:

• Add new agent types
• Implement useful tools
• Improve documentation
• Add tests
• Report bugs
• Suggest features

License

This project is licensed under the MIT License - see the LICENSE file for details.