databricks-model-serving

Deploy MLflow models to Databricks Model Serving by logging the model with MLflow, then creating a scalable REST endpoint for production-grade serving. This workflow supports traditional ML models, custom PyFunc, and GenAI agents.

Serving GenAI agents on Databricks integrates with Unity Catalog functions and vector search tools to enable tool-enabled agents. You deploy the agent to a Model Serving endpoint for scalable query execution and status checks.

Yes, you can check endpoint status and manage deployments across MLflow and GenAI workflows. The process covers end-to-end deployment, logging, and query workflows for both traditional ML models and GenAI agents.

Vector search and Unity Catalog are required to deploy tool-enabled AI agents with full functionality. Integrating Unity Catalog functions and vector search tools provides the tool-calling capabilities for GenAI agents served via REST endpoints.

Querying a Databricks Model Serving endpoint for custom PyFunc models sends REST requests to the scalable deployed endpoint. The workflow handles end-to-end querying and logging for custom PyFunc and MLflow models.

Limitations of deploying GenAI agents to Databricks Model Serving endpoints include dependencies on Unity Catalog and vector search for tool-enabled features. The deployment workflow supports MLflow models, custom PyFunc, and GenAI agents.