databricks-agent-bricks

To create an AI agent for document Q&A on Databricks, you configure a Knowledge Assistant that uses RAG to query documents stored in Unity Catalog volumes. This turns your cataloged documents into a conversational bot without extensive boilerplate code.

Yes, you can use natural language to query SQL data by configuring Genie Spaces over Unity Catalog tables. This enables natural-language-to-SQL exploration, allowing users to ask analytics questions conversationally.

A Supervisor Agent in Databricks multi-agent orchestration combines Knowledge Assistants, Genie Spaces, custom model endpoints, and external MCP servers into a single routing interface. It directs conversational queries to the appropriate specialized agent or function.

Yes, you need Unity Catalog volumes and tables to run these Databricks AI agents. The Knowledge Assistants require volumes for document storage, Genie Spaces require tables for SQL exploration, and the system relies on model serving endpoints to function.

You route user requests by building a Supervisor Agent that orchestrates multiple components within a Databricks workspace. It can direct policy questions to a Knowledge Assistant, analytics to a Genie Space, and actions to an MCP server.

The primary limitation is strict platform dependency; building conversational AI with Databricks agent bricks requires Databricks Unity Catalog volumes, tables, and model serving endpoints to function. Without these environment prerequisites, the agents cannot operate.