📦 Customize Python Modules

This guide will give you a brief overview of package management in a browser-based Python environment and how you can specify custom dependencies for your Gradio apps via the quarto-gradio extension.

Available Python Modules

Let’s say we want to display a Plotly chart in a Gradio app. However, by default, we are only allowed to import Python modules that come with the Pyodide distribution used by Gradio Lite and plotly is not one of them.

Installation via gradio.requirements

You can specify any number of Python packages in the gradio.requirements metadata.

---
filters:
  - gradio

gradio:
  requirements:
    - plotly==5.24.0
---

Tip

Click on “Submit” to visualize the selected dataset using Plotly.

import gradio as gr
import plotly.express as px

datasets = {
    "iris": px.data.iris(),
    "gapminder": px.data.gapminder(), 
    "tips": px.data.tips()
}
default_dataset = list(datasets.keys())[0]


def plot_figure(dataset):
    if not dataset:
        return None
    df = datasets[dataset]
    return px.scatter(
        df,
        x=df.columns[0],
        y=df.columns[1], 
        color=df.columns[-1],
        title=f"{dataset.title()} Dataset"
    )


gr.Interface(
    fn=plot_figure,
    inputs=gr.Dropdown(
        choices=list(datasets.keys()),
        value=default_dataset,
    ),
    outputs="plot",
    flagging_mode="never"
).launch()

plotly==5.24.0 import gradio as gr import plotly.express as px datasets = { "iris": px.data.iris(), "gapminder": px.data.gapminder(), "tips": px.data.tips() } default_dataset = list(datasets.keys())[0] def plot_figure(dataset): if not dataset: return None df = datasets[dataset] return px.scatter( df, x=df.columns[0], y=df.columns[1], color=df.columns[-1], title=f"{dataset.title()} Dataset" ) gr.Interface( fn=plot_figure, inputs=gr.Dropdown( choices=list(datasets.keys()), value=default_dataset, ), outputs="plot", flagging_mode="never" ).launch()

You can read about supported Python requirement formats in the Reference section.

Installation via micropip

It is also possible to install Python packages via the micropip library at runtime. In the interactive example below, we install the faker package to generate mock data.

# Read more about micropip at https://micropip.pyodide.org/en/v0.2.2/project/api.html#micropip.install
import micropip
await micropip.install("faker")

import gradio as gr
from faker import Faker
import pandas as pd

fake = Faker()
data_types = {"Name": fake.name, "Address": fake.address, "Email": fake.email, 
              "Phone": fake.phone_number, "Job": fake.job}

def generate_data(data_type, count):
    data = [{"#": i+1, "Data": data_types[data_type]()} for i in range(count)]
    return pd.DataFrame(data)

demo = gr.Interface(
    fn=generate_data,
    inputs=[gr.Dropdown(list(data_types.keys())), gr.Slider(1, 10, value=3, step=1)],
    outputs=gr.Dataframe(wrap=True),
)
demo.launch()

plotly==5.24.0 # Read more about micropip at https://micropip.pyodide.org/en/v0.2.2/project/api.html#micropip.install import micropip await micropip.install("faker") import gradio as gr from faker import Faker import pandas as pd fake = Faker() data_types = {"Name": fake.name, "Address": fake.address, "Email": fake.email, "Phone": fake.phone_number, "Job": fake.job} def generate_data(data_type, count): data = [{"#": i+1, "Data": data_types[data_type]()} for i in range(count)] return pd.DataFrame(data) demo = gr.Interface( fn=generate_data, inputs=[gr.Dropdown(list(data_types.keys())), gr.Slider(1, 10, value=3, step=1)], outputs=gr.Dataframe(wrap=True), ) demo.launch()

Limitations

Warning

Not every Python package is supported by Pyodide. While gradio and many other popular packages (including numpy, scikit-learn, and transformers-js) can be installed in Pyodide, if your app has many dependencies, its worth checking whether whether the dependencies are included in Pyodide, or can be installed with micropip.

Source: Gradio Lite Documentation

Next Steps

Having covered the core features of the quarto-gradio extension, in the subsequent guides we’ll explore more advanced use cases.