Question

A GenAI engineer builds a RAG chain in a development Databricks workspace and logs it to MLflow, registering it in Unity Catalog. The organization uses a separate production workspace for serving. The team wants to promote the exact same logged model artifact from dev to prod without re-running the training/logging notebook in production, ensuring bit-for-bit parity of the deployed chain. What is the recommended approach on Databricks?

Accepted Answer

Register the model in a Unity Catalog catalog shared across both workspaces, then create the production serving endpoint referencing the same registered model version by its Unity Catalog three-level name and version number.

Answer

Export the model artifact to DBFS in the dev workspace, copy the files manually to the prod workspace DBFS, and re-log them with mlflow.log_artifacts before serving.

Answer

Re-run the model-logging notebook in the production workspace using the same code so an identical model version is created locally, then deploy it.

Answer

Use the MLflow Model Registry webhook to automatically retrain and re-register the model in the production workspace whenever a dev version is promoted to the Production alias.