Multi-Swarm Framework

A powerful framework for creating collaborative AI agent swarms, supporting multiple LLM providers and advanced agent management capabilities.

Latest Version: 1.0.5

What's New

• Plugin system for extensible agent capabilities
• Centralized YAML configuration for model management
• Enhanced experimental model support
• Improved version compatibility handling
• Enhanced knowledge coordination across agents
• Improved temporal awareness in document retrieval
• Better knowledge gap detection and analysis
• Advanced knowledge synthesis capabilities
• Optimized collaborative search functionality
• Improved communication flows between agents
• Enhanced state management and persistence
• Better error handling and recovery mechanisms

Features

Plugin System

• Extensible plugin architecture for custom tools
• Built-in plugin registry and management
• Example tools including Git integration
• Easy-to-implement plugin interface
• Automatic plugin discovery and loading
• Plugin health monitoring and lifecycle management

Multi-LLM Support

• Claude (Anthropic) integration with latest models:
  • claude-3-5-sonnet-latest (balanced)
  • claude-3-5-opus-latest (research)
• Gemini (Google) integration with latest models:
  • gemini-2.0-flash-exp (high performance)
• Automatic model selection based on agent roles and tasks

Advanced Agency Management

• Flexible agent creation and configuration
• Sophisticated communication flows between agents
• State persistence and thread management
• Automated task distribution and coordination
• Knowledge synthesis and gap analysis

Built-in Tools

• Plugin-based tool architecture
• Git integration for version control
• Code Interpreter for executing Python code
• RAG system with temporal awareness
• File Search with semantic capabilities
• Docker integration for isolated environments
• Customizable tool creation framework

Installation

pip install multi-swarm

For development:

pip install multi-swarm[dev]

For GPU support:

pip install multi-swarm[gpu]

Quick Start

from multi_swarm import Agency, Agent
from multi_swarm.plugins.tools.development import GitTool

# Create agents with specific roles
ceo = Agent(
    name="CEO",
    description="Manages overall strategy and coordination",
    llm_provider="claude",  # Will use claude-3-5-sonnet-latest
    provider_config={
        "model": "claude-3-5-sonnet-latest",
        "api_version": "2024-03"
    }
)

developer = Agent(
    name="Developer",
    description="Handles technical implementation",
    llm_provider="gemini",  # Will use gemini-2.0-flash-exp
    provider_config={
        "model": "gemini-2.0-flash-exp",
        "api_version": "2024-01"
    }
)

# Register tools with agents
git_tool = GitTool()
developer.register_tool(git_tool)

# Create agency with communication flows
agency = Agency([
    ceo,  # Entry point for user communication
    [ceo, developer],  # CEO can communicate with Developer
])

# Run the agency
agency.run()

Plugin Development

Create custom plugins by extending the base plugin class:

from multi_swarm.plugins.base import PluginBase, PluginMetadata

class CustomTool(PluginBase):
    """Custom tool implementation."""
    def __init__(self):
        metadata = PluginMetadata(
            name="custom_tool",
            version="1.0.0",
            description="Custom tool description",
            author="Your Name",
            requirements=["required-package"],
            tags=["custom", "tool"]
        )
        super().__init__(metadata=metadata)
    
    def get_capabilities(self):
        return ["custom_capability"]

Task-Specific Model Selection

The framework automatically selects the best model based on the agent's role:

Claude Models

• Code Generation & Development: claude-3-5-sonnet-latest
• Research & Analysis: claude-3-5-opus-latest
• Technical Writing: claude-3-5-sonnet-latest

Gemini Models

• Data Processing: gemini-2.0-flash-exp
• Real-time Operations: gemini-2.0-flash-exp
• Machine Learning Tasks: gemini-2.0-flash-exp

Documentation

For detailed documentation, visit:

• Getting Started Guide
• API Reference
• Examples
• LLM Providers

Requirements

• Python 3.8+
• Dependencies:
  • anthropic>=0.18.1
  • google-generativeai>=0.3.2
  • pydantic>=2.0.0
  • python-dotenv>=1.0.0
  • docker>=6.1.0
  • sentence-transformers>=2.2.0
  • faiss-cpu>=1.7.4 (or faiss-gpu for GPU support)
  • transformers>=4.38.0
  • torch>=2.0.0
  • numpy>=1.24.0
  • click>=8.0.0

Optional Dependencies

Development Tools

• pytest>=8.0.0
• pytest-asyncio>=0.25.0
• pytest-cov>=6.0.0
• black>=24.0.0
• isort>=5.0.0
• mypy>=1.8.0

GPU Support

• faiss-gpu>=1.7.4
• torch>=2.0.0 (with CUDA support)

License

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

Contributing

Contributions are welcome! Please read our Contributing Guidelines and Code of Conduct.

Support

• Report issues on GitHub Issues
• Join discussions in GitHub Discussions

Authors

• Bart Van Spitaels (bart.vanspitaels@gmail.com)