Skip to main content

Use Genie in multi-agent systems

Preview

This feature is in Public Preview.

This page describes Genie agent systems and shows how to create a multi-agent system using Mosaic AI Agent Framework and Genie spaces.

Requirements

Benefits of Genie agents

Multi-agent systems consist of multiple AI agents working together, each with specialized capabilities. A Genie agent allows users to use natural language to query structured data.

Agent Framework allows agents to share answers as context for each other. So users can get answers that pull from multiple sources. You could have an already-deployed RAG agent that queries an unstructured document corpus and combine it with a Genie agent that queries a database of records.

Multi-agent systems typically consist of a supervisor agent that directs and manages context for specialized worker agents.

Multi-agent framework consisting of a supervisor and specialized agents.

Example notebook: Multi-agent system with Genie

The following notebook shows you how to create a multi-agent system that orchestrates these types of agents:

  • Genie Spaces
  • Agents deployed to Model Serving
  • UC function-calling agents built using LangGraph.

Multi-agent using Genie and LangGraph

Open notebook in new tab

Access Genie spaces with on-behalf-of-user authorization

On-behalf-of-user authorization allows an agent system to access a Genie space and its underlying resources with agent end user's permissions.

See On-behalf-of-user authentication for more information on how to set up on-behalf-of-user authorization with Genie spaces.