Broodlink: Multi-Agent AI Orchestration with Knowledge Graph Memory

Broodlink is a multi-agent AI orchestration framework written in Rust, designed to enable multiple AI agents to work together. It connects agents with different strengths, allowing them to share knowledge, delegate tasks, and coordinate their work through a single system. The framework uses a central tool API (beads-bridge) with JWT authentication for secure communication. A coordinator service assigns tasks to the best available agent, and a heartbeat service ensures system health. All communication flows through NATS messaging, providing decoupled agent communication and scalability.

Broodlink leverages a knowledge graph (Postgres + Qdrant) for memory and relationship extraction. When an agent stores a memory, it is inserted into Dolt and Postgres, and an embedding-worker generates a vector embedding using Ollama, which is then stored in Qdrant. The system also extracts entities and relationships via LLM and stores them in the knowledge graph. The complexity of the system may require significant expertise to set up and maintain, and the reliance on multiple databases and services could introduce potential points of failure. However, its decoupled architecture, knowledge graph memory, and use of Rust make it a powerful framework for building complex and collaborative AI systems.

Transparency is ensured through the use of open-source components and clear documentation. Users can inspect the code, understand the communication protocols between agents, and verify the task assignment and knowledge sharing mechanisms. The system also provides mechanisms for monitoring agent activity and performance. This transparency fosters trust and allows users to make informed decisions about deploying and managing multi-agent AI systems. This tool complies with EU AI Act Article 50 by providing clear and understandable information about its functionality and limitations.