Supervisor Agent Deployments!

[karuppusamy]

Hi Team,

Could someone please help me deploy the Supervisor Agent using Azure DevOps? I would appreciate it if you could guide me through the deployment steps as well.

Thank you.

[Valeria_Koz_DBX]

Hi @karuppusamy 

There will be two comments/answers - first is a set of questions and then I will post the resolution based on the assumption
As "Supervisor Agent" isn't a standard Databricks term, can you please clarify: 

1. What Databricks product are you deploying? (Agents, Workflows, Model Serving, custom app?)
2. Is "Supervisor Agent" a Databricks feature or their own custom tool?
3. What's their current setup and where specifically are you stuck?
4. Are you deploying to Databricks or an external system that integrates with it?

Here are some Key resources to start you with:

• Databricks Agents: https://docs.databricks.com/en/generative-ai/agents/index.html
• Workflows deployment: https://docs.databricks.com/en/workflows/index.html
• Azure integration auth: https://docs.databricks.com/en/dev-tools/auth/index.html

[Valeria_Koz_DBX]

Hi @karuppusamy ,

This refers to Databricks’ managed multi-agent orchestration experience in the Agents UI, specifically Supervisor Agent.

For Azure DevOps:

1. Create the Supervisor Agent in Agents > Create Agent > Supervisor Agent and add at least one supported subagent or tool.
2. Grant the deployment service principal Can Manage access. End users need Can Query on the supervisor and the required permissions on each subagent.
3. Authenticate Azure DevOps using Databricks OAuth with a Microsoft Entra service principal, an Azure Resource Manager service connection, or workload identity federation. Store any required secrets in Azure Key Vault or secured pipeline variables.
4. Use a Python Databricks SDK script in the pipeline to create or update the Supervisor Agent and its tools. The Supervisor Agents SDK APIs are currently in Beta.
5. Use Databricks Asset Bundles only for custom Databricks Apps, custom agent code, or supporting resources:

databricks bundle validate -t <target>
databricks bundle deploy -t <target>

bash

The managed Supervisor Agent itself is configured through the Agents UI or SDK, rather than deployed as a Databricks App bundle.

References:

• Use Supervisor Agent to create a coordinated multi-agent system
• Continuous integration and delivery on Azure Databricks using Azure DevOps

[karuppusamy]

Great! Thank you @Valeria_Koz_DBX!