databricks-agent-bricks

You orchestrate multiple Databricks agents by configuring Supervisor Agents (MAS) to coordinate Knowledge Assistants, Genie Spaces, and endpoints into a single multi-agent workflow. This approach enables rapid composition of specialized agents within Databricks environments.

Yes, you can create and manage Genie Spaces for SQL exploration over Unity Catalog tables. This allows teams to build conversational analytics applications that query structured data directly through SQL-enabled interactions within Databricks.

To set up a Knowledge Assistant for document Q&A, you create and manage KA Bricks using the manage_ka API. You need documents stored in Volumes and a specific KA tile ID to enable the assistant to answer questions from your stored data.

Yes, building Databricks Agent Bricks requires access to Volumes, Unity Catalog, and model serving endpoints. These prerequisites provide the foundational data storage, governance, and compute resources needed to run conversational AI applications.

The best way to mix KA, Genie Spaces, and MAS in one workflow is through a combined Brick configuration. This involves using manage_ka, manage_genie, and manage_mas APIs to integrate document Q&A, SQL analytics, and multi-agent orchestration together.

Your Databricks Agent Brick requires a KA tile ID and Genie space ID because these identifiers serve as references for the manage_ka and manage_genie APIs to correctly route and manage the specific Knowledge Assistant and SQL exploration components.