Mahilo is a Python framework for building multi-agent AI systems with human-in-the-loop supervision. It enables agents to communicate and share context, facilitating complex workflows for applications like customer service or emergency response simulations. The framework is designed for developers and researchers looking to orchestrate sophisticated agent interactions.

How It Works

Mahilo employs a WebSocket-based architecture for real-time communication between agents and clients. Agents, defined via a BaseAgent class or integrated from frameworks like LangGraph, can be registered with an AgentManager. This manager handles context sharing, making recent conversation history available across agents. Agents can communicate peer-to-peer or hierarchically, with communication pathways controlled by can_contact lists. LLM providers are abstracted via LiteLLM, and message validation policies can be implemented to ensure communication quality.

Quick Start & Requirements

• Install: pip install mahilo
• Voice features: pip install mahilo[voice] (requires PyAudio installation for your OS).
• OpenAI API key required: export OPENAI_API_KEY=<your_api_key>
• Example usage: Navigate to an example directory and run python run_server.py.
• Connect via CLI: mahilo connect --agent-name <your_agent_name> (add --voice for voice interaction).
• Documentation: Detailed Features

Highlighted Details

• Real-time API support for voice interaction with agents.
• Flexible communication patterns: peer-to-peer and hierarchical.
• Control communication flow with can_contact lists.
• Multi-provider LLM support via LiteLLM.
• Built-in message validation policies (anti-loop, relevance, toxicity).

Maintenance & Community

• Contributions are welcome via Pull Requests.
• Links to community resources are not explicitly provided in the README.

Licensing & Compatibility

• License: Not explicitly stated in the README.
• Compatibility: Designed for Python environments. Commercial use implications are not detailed.

Limitations & Caveats

The README does not specify the project's license, which is crucial for commercial adoption. Community support channels like Discord or Slack are not mentioned, potentially indicating a smaller or less active community.