To generate predictions, make POST requests at the /predict endpoint.