marimo-dev 0.1.8

Build and publish python packages from marimo notebooks

marimo_dev

A literate programming build system that converts Marimo notebooks into distributable Python packages. Inspired by nbdev.

What it does

Looks in /notebooks dir for self contained functions and classes and extracts them into a PyPI ready directory. Each notebook will be striped of the leading ##_, and XX_ and test notebooks will be ignored. Run md build to generate a proper Python package with __init__.py, module files, and llms.txt API documentation.

What it does not

An intentional effort to minimise magic project set up. You are responsible for setting up your pyproject.toml, and even mkdir notebooks. Note: I highly reccommend that you use uv my main motivation for making this was the dependency harmony betwwen marimo and uv

Project structure

my-project/
├── README.md
├── pyproject.toml
├── notebooks/
│   ├── 01_core.py
│   ├── 02_read.py
│   ├── ...
├── src/                 # auto created
│   └── my_package/
│       ├── __init__.py 
│       ├── core.py    
│       ├── read.py
│       └── ...
├── docs/                # auto created
│   ├── index.html       # Fut ver. 0.2
│   └── llms.txt
└── dist/                # auto created
    └──...

How it works

The build system parses notebooks via AST, extracts decorated exports (@app.function, @app.class_definition), and writes clean module files. It reads metadata from pyproject.toml and generates __init__.py with proper imports and __all__ exports.

The llms.txt file contains function signatures with inline documentation extracted from comments, formatted for LLM consumption. This provides a compact API reference. Note: this will work much better with fastcore.docments style function definitions

CLI usage

md build              # build package from notebooks/
md publish            # publish to PyPI
md publish --test     # publish to Test PyPI

Requirements

• Python 3.12+, Marimo, uv, pyproject.toml
  • for relative imports to work in multi module libraries add the following to your pyproject.tml

  • </code></pre>
    </li>
    </ul>
    pythonpath = ["src"]  ```</li>
    </ul>
    <p><strong>Tip</strong> <em>let marimo manages your <code>pyproject.toml</code> through its package tab, making dependencies visible and easy to update. When you add packages and remove them from the marimo package tab marimo will automaticly update your pyproject.toml</em></p>
    <h2>Install</h2>
    <pre lang="bash"><code>uv add marimo-dev
    

    Helpful

    uv sync --upgrade` # to update uv.lock and pyroject.toml in one go...
    uv cache clean     # General Trouble Shooting Tip
    

    • you need to manually update the version in pyproject.toml

    Module structure

    • core.py - Data model: Kind, Param, Node
    • read.py - Parse notebooks, extract exports, scan project
    • pkg.py - Write module files and __init__.py
    • docs.py - Generate signatures and llms.txt
    • build.py - Orchestrate the build
    • cli.py - Command-line interface