Agent Bricks

Create and manage Databricks Agent Bricks - pre-built AI components for building conversational applications.

Overview

Agent Bricks are three types of pre-built AI tiles in Databricks:

| Brick | Purpose | Data Source | |-------|---------|-------------| | Knowledge Assistant (KA) | Document-based Q&A using RAG | PDF/text files in Volumes | | Genie Space | Natural language to SQL | Unity Catalog tables | | Multi-Agent Supervisor (MAS) | Multi-agent orchestration | Model serving endpoints |

Prerequisites

Before creating Agent Bricks, ensure you have the required data:

For Knowledge Assistants

• Documents in a Volume: PDF, text, or other files stored in a Unity Catalog volume
• Generate synthetic documents using the unstructured-pdf-generation skill if needed

For Genie Spaces

• Tables in Unity Catalog: Bronze/silver/gold tables with the data to explore
• Generate raw data using the synthetic-data-generation skill
• Create tables using the spark-declarative-pipelines skill

For Multi-Agent Supervisors

• Model Serving Endpoints: Deployed agent endpoints to orchestrate
• These could be custom agents, fine-tuned models, or other deployed services

MCP Tools

Knowledge Assistant Tools

create_or_update_ka - Create or update a Knowledge Assistant

• name: Name for the KA
• volume_path: Path to documents (e.g., /Volumes/catalog/schema/volume/folder)
• description: (optional) What the KA does
• instructions: (optional) How the KA should answer
• tile_id: (optional) Existing tile_id to update
• add_examples_from_volume: (optional, default: true) Auto-add examples from JSON files

get_ka - Get Knowledge Assistant details

• tile_id: The KA tile ID

delete_ka - Delete a Knowledge Assistant

• tile_id: The KA tile ID to delete

Genie Space Tools

IMPORTANT: Before creating a Genie Space, you MUST first inspect the table schemas using get_table_details to understand the data. This allows you to:

• Select the most relevant tables for the use case
• Write sample questions that reference actual column names and data patterns
• Create a description that accurately explains the data model

Genie Space Creation Workflow:

1. Call get_table_details(catalog, schema) to fetch table schemas
2. Analyze the columns, data types, and relationships
3. Select tables appropriate for the user's use case (prefer silver/gold over bronze)
4. Generate 5-10 sample questions based on actual columns and business context
5. Write a description explaining what users can explore
6. Call create_or_update_genie with the prepared content

create_or_update_genie - Create or update a Genie Space for SQL exploration

• display_name: Display name for the space
• table_identifiers: List of tables (e.g., ["catalog.schema.table1", "catalog.schema.table2"])
• warehouse_id: (optional) SQL warehouse ID (auto-detects if not provided)
• description: (optional) What the space does - explain the data model and relationships
• sample_questions: (optional) List of sample questions that reference actual columns
• space_id: (optional) Existing space_id to update

get_genie - Get Genie Space details

• space_id: The Genie space ID

delete_genie - Delete a Genie Space

• space_id: The Genie space ID to delete

Multi-Agent Supervisor Tools

create_or_update_mas - Create or update a Multi-Agent Supervisor

• name: Name for the MAS
• agents: List of agent configurations:
• name: Agent name
• endpoint_name: Model serving endpoint name
• description: What this agent handles (used for routing)
• description: (optional) What the MAS does
• instructions: (optional) Routing instructions
• tile_id: (optional) Existing tile_id to update
• examples: (optional) List of example questions with question and guideline fields

get_mas - Get Multi-Agent Supervisor details

• tile_id: The MAS tile ID

delete_mas - Delete a Multi-Agent Supervisor

• tile_id: The MAS tile ID to delete

Typical Workflow

1. Generate Source Data

Before creating Agent Bricks, generate the required source data:

For KA (document Q&A):

1. Use `unstructured-pdf-generation` skill to generate PDFs
2. PDFs are saved to a Volume with companion JSON files (question/guideline pairs)

For Genie (SQL exploration):

1. Use `synthetic-data-generation` skill to create raw parquet data
2. Use `spark-declarative-pipelines` skill to create bronze/silver/gold tables

2. Create the Agent Brick

Use the appropriate create_or_update_* tool with your data sources.

3. Wait for Provisioning

Newly created KA and MAS tiles need time to provision. The endpoint status will progress:

• PROVISIONING - Being created (can take 2-5 minutes)
• ONLINE - Ready to use
• OFFLINE - Not running

4. Add Examples (Automatic)

For KA, if add_examples_from_volume=true, examples are automatically extracted from JSON files in the volume and added once the endpoint is ONLINE.

Best Practices

1. Use meaningful names: Names are sanitized automatically (spaces become underscores)
2. Provide descriptions: Helps users understand what the brick does
3. Add instructions: Guide the AI's behavior and tone
4. Include sample questions: Shows users how to interact with the brick
5. Use the workflow: Generate data first, then create the brick

See Also

• 1-knowledge-assistants.md - Detailed KA patterns and examples
• 2-genie-spaces.md - Detailed Genie patterns and examples
• 3-multi-agent-supervisors.md - Detailed MAS patterns and examples