Question

A GenAI engineer has built a custom retrieval-augmented chain in Python that orchestrates a vector search retriever, a custom re-ranking function, and a foundation model call. The chain is not implemented using a supported LangChain construct — it is plain Python with several helper modules. The engineer needs to log this chain to MLflow so it can be deployed to a Model Serving endpoint that accepts JSON requests and returns text responses. Which approach correctly packages this chain for serving?

Accepted Answer

Log the chain using mlflow.pyfunc with a custom PythonModel subclass, specifying the code paths for helper modules and defining a predict() method along with a model signature and input example.

Answer

Serialize the chain object with pickle and upload the pickle file to a Unity Catalog volume, then point the serving endpoint directly at the volume path.

Answer

Use mlflow.langchain.log_model since any Python chain can be logged with the LangChain flavor regardless of implementation.

Answer

Convert each helper module into a Unity Catalog Python function and register only the foundation model to the endpoint, calling the functions at inference from the client.