databricks-model-serving

To deploy MLflow models as scalable REST APIs, log the model with MLflow, configure the endpoint, and deploy it. This workflow exposes traditional ML, custom PyFunc models, and GenAI agents as accessible REST APIs.

Yes, you can deploy GenAI agents on Databricks using ResponsesAgent or LangGraph patterns. This workflow supports integrating Unity Catalog function tools and vector search for retrieval-augmented workflows exposed as accessible REST APIs.

Setting up Databricks model serving endpoints requires Unity Catalog, Model Serving enabled, and compatible MLflow or foundation model APIs. You also need specific endpoint names from the SKILL.md file to execute the deployment workflow.

Yes, this workflow handles both classical ML and GenAI agent deployment. It provides a cohesive process to configure endpoints and expose models as scalable REST APIs while coordinating MLflow logging and Unity Catalog permissions.

The best way to coordinate MLflow logging and Unity Catalog permissions is using a cohesive deployment workflow. This approach manages endpoint configuration and monitoring while exposing ML models and GenAI agents as scalable REST APIs.