databricks-model-serving

Deploy MLflow models on Databricks by creating model serving endpoints, which packages your trained model into a scalable, low-latency REST API. This workflow supports both autolog and manual logging for classical ML models.

Yes, Databricks model serving supports custom PyFunc models and GenAI agents. You can build and serve agents using frameworks like ResponsesAgent and LangGraph through scalable REST endpoints.

Databricks model serving integrates directly with Unity Catalog functions and Vector Search tools, allowing you to register models, manage endpoints, and connect external tools within your Databricks workspace.

Query and monitor Databricks serving endpoints using the SDK, REST APIs, or MCP tools. This allows you to programmatically update endpoints and retrieve predictions from your deployed MLflow or GenAI models.

Yes, Unity Catalog registration is a core requirement for serving models on Databricks. It handles the governance and registration of MLflow models, custom PyFunc models, and GenAI agents before endpoint deployment.