You want to package a chain, log its versions, and deploy it behind an endpoint. Which pair fits?
Why D is the answer
MLflow and Model Serving
MLflow logs and versions the app, and Model Serving hosts it.
In detail
- MLflow logs and versions the application.
- Model Serving hosts it behind an endpoint.
- Log the dependencies with the model, so it runs the same way when served.
Use case
An illustrative scenario, not a real incident.
A team builds a chain, logs each version with MLflow, and deploys one behind a Model Serving endpoint. A bad release can be swapped for the earlier version.
See it in different ways
flowchart TD A["RAG model"] --> B["Model flavor"] A --> C["Embedding model"] A --> D["Retriever"] A --> E["Dependencies"] A --> F["Input example"] A --> G["Model signature"]
sequenceDiagram participant C as pyfunc chain participant M as MLflow participant U as Unity Catalog participant S as Model Serving C->>M: Log the model M->>U: Register a version U->>S: Deploy to an endpoint
flowchart LR S0["Build the chain"] --> S1["Log with MLflow"] --> S2["Register a version"] --> S3["Serve behind an endpoint"]
Key point: MLflow tracks versions, and Model Serving hosts them.
Why the other options are wrong
- A. Delta Sharing and a dashboard
- Delta Sharing shares data, and a dashboard does not serve a chain.
- B. A notebook and a shared drive
- A shared drive gives no versioned deployment or endpoint.
- C. Git and email
- Git tracks code, but it does not host an endpoint.