Pipeline apps and web servers

Wrap a Pipeline in a Gradio demo or a lightweight web server for internal tools and prototypes.

Adapted for this playbook from the 🤗 Transformers documentation by Hugging Face. Official page: https://huggingface.co/docs/transformers/pipeline_gradio. Images © Hugging Face (documentation-images) unless noted. This is not a substitute for the upstream docs — verify against the current version.

Covers Hugging Face pages: pipeline_gradio, pipeline_webserver

Executive summary

Gradio gives stakeholders a UI in minutes. For service-style inference, follow Hugging Face’s web-server patterns — then graduate to Serve CLI or vLLM when you need concurrency and OpenAI-compatible APIs.

Gradio sketch

import gradio as gr
from transformers import pipeline

pipe = pipeline("sentiment-analysis")

def predict(text):
    return pipe(text)[0]

gr.Interface(fn=predict, inputs="text", outputs="json").launch()

Official guide: Machine learning apps.

Web server inference

Hugging Face documents serving Pipelines behind HTTP for simple deployments — see Web server inference. For multi-user LLM serving, prefer Serving or Inference engines.

Production caution

• Gradio demos are not a hardened multi-tenant gateway.
• Add auth, rate limits, request size caps and PII controls before exposing client data.

What to do next

Serving.