Huggingface
CertifiedTasks that call Hugging Face Inference APIs.
Set your Hugging Face token, model endpoint, and payload (text, image, or other inputs) to run hosted models via the Inference API and return predictions for downstream tasks.
Huggingface
Tasks that call Hugging Face Inference APIs.
Set your Hugging Face token, model endpoint, and payload (text, image, or other inputs) to run hosted models via the Inference API and return predictions for downstream tasks.
How to use the Hugging Face plugin
Call any model hosted on the Hugging Face Inference API from Kestra flows.
Authentication
Set token to your Hugging Face API token. Store it in a secret.
Tasks
Inference sends a request to a Hugging Face hosted model endpoint and returns the prediction. Set modelId to the model's repository path (e.g., facebook/bart-large-cnn) and inputs to the payload appropriate for that model type — text for NLP tasks, base64-encoded data for vision models. The response shape varies by model and task type; check the model's API documentation on the Hugging Face Hub for the expected input and output format.
