migrate-from-model-serving

Migrate an MLflow ResponsesAgent from Databricks Model Serving to Databricks Apps by extracting artifacts, converting predict() and predict_stream() methods to decorated functions, fixing import paths, and deploying with Databricks Asset Bundles.

Converting MLflow Model Serving endpoints to Databricks Apps involves scaffolding files, transforming ResponsesAgent predict methods into decorated functions, mapping resources from MLmodel to databricks.yml, and executing bundle deploy commands.

Yes, you can select either async or sync migration paths when moving MLflow agents to Databricks Apps, allowing flexibility in converting predict and predict_stream functions for the MLflow GenAI Server.

Mapping MLmodel artifacts to databricks.yml during Databricks Apps migration requires extracting artifacts, updating import paths, and configuring resource mappings in the databricks.yml file for Databricks Asset Bundles deployment.

Yes, migrating from Databricks Model Serving to Databricks Apps supports local testing, providing step-by-step workflows to validate decorated functions and artifact extraction before final bundle deployment.

Moving MLflow Model Serving to Databricks Apps requires strict transformation rules for ResponsesAgent methods and artifact extraction, meaning import paths and databricks.yml resource mappings must be manually fixed to ensure successful bundle deployment.