Gradio — Build ML Demos & AI Apps in Python

What it is

Gradio is the standard Python library for building interactive ML demos and AI web applications. You create chat interfaces, image generators, audio processors, and data tools with a few lines of Python -- no frontend code required. It automatically generates a web UI and an optional public share link.

ML engineers and researchers who need a quick web interface for their models are the primary users. Gradio is used extensively on Hugging Face Spaces and by teams that want to demonstrate model capabilities without building a full frontend.

How it saves time or tokens

Without Gradio, turning a Python model into a web demo requires building HTML/CSS/JS frontend code, setting up a web server, and handling file upload/download logic. Gradio replaces all of this with a single function call. The gr.Interface class maps Python function inputs/outputs to UI components automatically.

The share=True flag generates a public URL through Gradio's tunnel service, eliminating the need to deploy to a server for sharing demos with stakeholders.

How to use

1. Install Gradio:

pip install gradio

1. Create a simple interface:

import gradio as gr

def greet(name):
    return f'Hello, {name}!'

gr.Interface(fn=greet, inputs='text', outputs='text').launch()

1. Open http://localhost:7860 in your browser. Add share=True to get a public URL.

Example

A chat interface with streaming responses:

import gradio as gr

def chat(message, history):
    response = ''
    for char in f'You said: {message}':
        response += char
        yield response

gr.ChatInterface(fn=chat, type='messages').launch()

This creates a full chat UI with message history, streaming token display, and a text input box -- all from seven lines of Python.

Related on TokRepo

• Coding Tools -- Developer tools for building and testing AI applications
• AI Tools for Research -- Research tools that pair with demo interfaces

Common pitfalls

• Gradio's default server binds to 0.0.0.0, making it accessible on your network. Use server_name='127.0.0.1' for local-only access when working with sensitive models.
• Share links expire after 72 hours. For permanent deployments, host on Hugging Face Spaces or deploy behind a reverse proxy.
• Large file uploads can time out with the default settings. Increase max_size in the component configuration for applications that process large videos or datasets.