databricks-model-serving

You can deploy MLflow models and GenAI agents to Databricks Model Serving endpoints by enabling Model Serving with MLflow support. This provides production-grade REST endpoints for traditional ML models, custom Python models, and AI agents.

Yes, you can deploy ChatAgent and ResponsesAgent workflows to serving endpoints. These GenAI agents can be enhanced by integrating Unity Catalog Functions and Vector Search tools for retrieval and tool usage, enabling advanced AI agent capabilities.

Before querying endpoints, you need to enable Model Serving with MLflow support and ensure appropriate workspace permissions. You must also have tooling configured for endpoint management and monitoring to successfully issue real-time inference requests.

You can test deployed models and monitor endpoint health by issuing real-time inference requests and checking endpoint status. This supports model testing, A/B testing workflows, and continuous monitoring of your Databricks serving endpoints.

Yes, Databricks Model Serving supports deploying custom Python models to scalable REST endpoints. This allows you to issue real-time inference requests and run A/B testing on your custom models alongside traditional ML workflows.