marimo

Create new notebook

$ marimo edit [notebook.py]

$ marimo edit [notebook.py]

$ marimo run [notebook.py]

$ marimo run --port [8080] [notebook.py]

$ marimo convert [notebook.ipynb] > [notebook.py]

$ marimo export html [notebook.py] > [output.html]

$ marimo new

$ marimo --version

marimo [edit] [run] [convert] [export] [options] [notebook.py]

marimo is a reactive Python notebook that stores notebooks as pure Python files. Unlike Jupyter, cells automatically re-execute when their dependencies change.
Notebooks are Python scripts, enabling version control with Git without JSON merge conflicts. Cells define functions and variables that form a dependency graph.
The reactive execution model ensures consistency. Changing a variable automatically updates all dependent cells. Hidden state and execution order issues from traditional notebooks are eliminated.
Run mode deploys notebooks as interactive web applications. Users interact with widgets while the Python backend handles computation.
Conversion from Jupyter preserves code and markdown cells. The reactive structure may need adjustment for notebooks that relied on execution order.

edit FILE

Edit notebook in browser.

Run notebook as interactive app.

Convert from Jupyter to marimo.

Export notebook (html, md, script).

Create new notebook from template.

Server port.

Server host.

Don't open browser.

Watch for file changes.

Run in sandbox environment.

Show version.

Show help.

Different mental model from Jupyter - linear execution not guaranteed. Some Jupyter magic commands not supported. Large notebooks may have performance implications from reactive updates.

marimo was created by Akshay Agrawal and Myles Scolnick in 2023 at Stanford. It addresses reproducibility issues in traditional notebooks by making execution deterministic through reactive programming.

jupyter(1), ipython(1), streamlit(1), python(1)