databricks-model-serving

Deploy ML models to Databricks endpoints by logging traditional ML or custom PyFunc models with MLflow, then deploying them to scalable Databricks endpoints and querying responses via REST or OpenAI-compatible interfaces.

Yes, you can deploy GenAI agents to Databricks endpoints. The deployment supports traditional ML models, custom PyFunc models, and GenAI agents, enabling AI agent testing and querying via OpenAI-compatible interfaces.

Yes, a Unity Catalog-enabled workspace is required for Databricks Model Serving. The process relies on MLflow-based logging and uses Unity Catalog Functions and Vector Search tools as endpoint capabilities.

Query a Databricks model serving endpoint using Databricks REST endpoints or OpenAI-compatible interfaces. This applies to responses from traditional ML models, custom PyFunc models, and deployed GenAI agents.

Yes, you can integrate Unity Catalog Functions and Vector Search tools as endpoint capabilities. This allows deployed ML models and GenAI agents to leverage these tools during querying via Databricks endpoints.

You can deploy traditional ML models like sklearn and xgboost, custom PyFunc models, and GenAI agents to Databricks endpoints. All models must be logged using MLflow before deployment.