Model Serving Infrastructure and Inference Endpoints

What an inference endpoint exposes to consuming applications.

What does a model inference endpoint typically expose?