marimo-functions

Flexible interface for defining functions with marimo UI inputs

What is a "Marimo Function"

When running analysis in marimo, it is common to have a set of inputs which are all mapped as inputs to a particular function. The marimo-functions library provides a simple framework for implementing this pattern of user interaction.

The MarimoFunction class has an inputs attribute which maps keyword arguments to marimo UI elements. The type attribute of each input specifies the UI element, and all of the remaining kwargs are passed to the constructor.

The run() method then takes the kwargs provided by the user and runs any method that consumes them appropriately.

For example:

from marimo_functions import MarimoFunction
from sklearn.manifold import TSNE as SKTSNE


class TSNE(MarimoFunction):
    """
    A class representing a t-SNE (t-distributed Stochastic Neighbor Embedding) function.
    """
    inputs = {
        "n_components": {
            "type": "number",
            "label": "Number of components",
            "value": 2,
            "start": 1,
            "step": 1
        },
        "perplexity": {
            "type": "number",
            "label": "Perplexity",
            "value": 30,
            "start": 5,
            "stop": 50,
            "step": 1
        },
        "early_exaggeration": {
            "type": "number",
            "label": "Early exaggeration",
            "value": 12.0,
            "start": 1.0,
            "stop": 100.0,
            "step": 0.1
        },
        "random_state": {
            "type": "number",
            "label": "Random State",
            "value": 0
        }
    }
    name = "t-SNE"
    description = "T-distributed Stochastic Neighbor Embedding."

    @classmethod
    def run(cls, **kwargs) -> SKTSNE:
        return SKTSNE(**kwargs)

Using a Function

To make use of a function in marimo, simply prompt the user for the inputs:

# func: MarimoFunction
inputs = func.prompt()
inputs

Note that by using inputs as the last line in the marimo cell, it is presented as a UI element to the user.

And then run the function using the values provided by the user:

output = func.run(**inputs.value)

Picking a Function

In addition to the single function, this library provides a mechanism for selecting a function of interest from a dropdown menu.

from marimo_functions import PickFunction

selected_function = (
    PickFunction(
        functions=[
            FunctionA,
            FunctionB,
            FunctionC,
            FunctionD,
            FunctionE
        ]
    )
    .prompt(
        label="Select a function",
        value="Function A"
    )
)
selected_function

The function selected by the user can be populated by the value attribute.

func = selected_function.value

# Display the description of the function
mo.md(func.description)

Interactive Example

To run a marimo notebook displaying this functionality:

1. Set up the local environment (uv sync)
2. Launch the notebook (uv run marimo edit example.py)