This library provides a Python framework for building hierarchical multi-agent systems, enabling a supervisor agent to orchestrate specialized agents. It's designed for developers and researchers creating complex agentic workflows that require task delegation and controlled communication, offering a structured approach to multi-agent coordination.

How It Works

The core of the library is the create_supervisor function, which builds a LangGraph StateGraph to manage agent interactions. It uses a tool-based handoff mechanism, allowing the supervisor LLM to delegate tasks to specific, specialized agents. This approach facilitates modularity and allows for clear separation of concerns within the multi-agent system.

Quick Start & Requirements

• Install via pip: pip install langgraph-supervisor
• Requires Python >= 3.10.
• OpenAI API key is needed for the example models.
• Official documentation: https://langgraph.ai/langgraph-supervisor/

Highlighted Details

• Supports creating multi-level hierarchies by nesting supervisors.
• Offers flexible message history management (full_history or last_message).
• Allows customization of handoff tools, including their names, descriptions, and arguments.
• Provides options for adding memory (short-term and long-term) via checkpointers and stores.
• Includes a create_forward_message_tool for direct message forwarding.

Maintenance & Community

The project is part of the LangGraph ecosystem, indicating active development and community support. Specific contributor details or community links are not explicitly mentioned in the README.

Licensing & Compatibility

The library is released under the MIT License, permitting commercial use and integration with closed-source projects.

Limitations & Caveats

The provided examples primarily use langchain-openai and gpt-4o, suggesting a strong dependency on OpenAI models. While the framework is flexible, adapting it to other LLM providers might require custom configurations.