A team registers an AI agent to Unity Catalog and deploys it to a Model Serving endpoint. The agent queries a Vector Search index at inference time. To ensure the endpoint can access the Vector Search index without hardcoded credentials, what must be configured during the logging step?

Question

Accepted Answer

B. The `resources` parameter in `log_model()` must declare the Vector Search index, enabling automatic authentication passthrough using the endpoint's service principal.. Automatic authentication passthrough for Databricks resources is enabled by declaring those resources in the `resources` parameter of `mlflow.langchain.log_model()` or `mlflow.pyfunc.log_model()`. At deployment time, Databricks configures the endpoint's service principal to authenticate to the declared resources, eliminating the need for hardcoded credentials.

A team registers an AI agent to Unity Catalog and deploys it to a Model Serving endpoint. The agent queries a Vector Search index at inference time. To ensure the endpoint can access the Vector Search index without hardcoded credentials, what must be configured during the logging step?

Related Questions

Discussion