migrate-from-model-serving

To migrate an MLflow ResponsesAgent from Databricks Model Serving to Databricks Apps, you extract the agent code and wrap its predict() and predict_stream() methods using @invoke and @stream decorators to create an Apps-ready entry point.

Converting ResponsesAgent predict methods for Databricks Apps involves wrapping the existing predict() and predict_stream() methods with @invoke and @stream decorators, which adapts the logic for the Databricks Apps environment.

Yes, you can convert your MLflow ResponsesAgent to async operation during Databricks Apps deployment, as the migration supports both async and sync paths to handle the transition effectively.

When moving a ResponsesAgent to Databricks Apps, the migration process handles state management by wiring the checkpointer or store for either stateful or stateless operation within the new environment.

Yes, migrating a ResponsesAgent to Databricks Apps produces a deployable bundle that includes production deployment scaffolding and a databricks.yml file for final configuration.