Question

A Generative AI Engineer has deployed a registered chat model to a Databricks Model Serving endpoint. A downstream microservice team (not using Databricks) needs to send inference requests to the endpoint over HTTP from their own application. What must the engineer provide so the external team can successfully invoke the endpoint?

Accepted Answer

The endpoint's invocation URL, a Databricks personal access token (or service principal OAuth token) for the Authorization header, and the expected JSON request schema

Answer

A copy of the MLflow model artifacts and the requirements.txt so the team can load the model locally with mlflow.pyfunc.load_model

Answer

The Unity Catalog three-level model name and a SQL warehouse connection string to query the model as a table

Answer

The cluster ID of the serving compute and a JDBC driver so the team can connect through the Databricks Runtime