databricks-model-serving

Deploy machine learning models to Databricks model serving endpoints by logging your model with MLflow, configuring a serving endpoint, and creating a scalable REST API to operationalize inference. It supports traditional ML, custom PyFunc models, and GenAI agents.

Yes, you can integrate Unity Catalog functions and Vector Search with AI agents. This combination enables tool-calling and retrieval capabilities within GenAI agents deployed to Databricks model serving endpoints for runtime querying.

Deploying GenAI agents and custom models requires Databricks DBR 16.1 or higher with Unity Catalog enabled. This environment provides the necessary foundation for logging, testing, and querying operationalized model serving endpoints.

MLflow supports logging custom PyFunc models for Databricks serving. You can package Python logic into PyFunc models, log them, and deploy them to scalable REST endpoints to operationalize custom machine learning workflows.

Query a Databricks model serving endpoint after deployment by issuing API requests to the scalable REST interface. You can monitor endpoint status and runtime query operationalized ML models or GenAI agents.