The huggingface_hub Python library now supports programmatic management of Hugging Face Inference Endpoints. Using the library, developers can create, update, pause, delete, and send requests to endpoints without touching the UI. The tutorial walks through a full example using the Zephyr 7B model with Text Generation Inference (TGI), covering endpoint creation with custom image configuration, waiting for readiness, running inference, pausing, and deleting the endpoint. This enables Infrastructure as Code workflows for GenAI deployments, useful for batch processing and automated model evaluation.
