Hi folks,

I recently revisited my old post about combining Django with Jupyter for leveraging model methods for data analysis. If you missed it, check it out here:

Is there a guide for using Django in a Jupyter notebook in 2022? Getting Started

For anyone who might be following along, I’ve just come across django notebook.
It looks really cool, but I think they also have bumped up against the same files in the same root directory issue as I mentioned above.

However, Jupyter can sometimes be cumbersome due to its UI complexity and the git-unfriendly ipynb format that also stores query results. This is where Marimo comes in.

Enter Marimo

I recently discovered the open source project Marimo. Similar to Observable’s “reactive notebook” concept, Marimo offers a Pythonic approach that avoids hidden states, making notebooks reproducible and more maintainable. Its highlights include:

• reactive: automatically updates all affected cells and outputs when a cell is run
• interactive: seamlessly bind sliders, tables, plots, and more to Python without needing callbacks
• reproducible: ensures deterministic execution with no hidden state
• deployable: functions as both a script and an app
• developer-friendly: uses a git-friendly .py format with support for modern tools like GitHub Copilot, autocomplete, and code formatting

For more details, see the Marimo docs.

The tool seems like a promising middle ground between a django shell (or shell_plus) and writing a full-fledged Django management command. It offers rich widgets similar to Jupyter while saving your work in a standard Python script.

For example, to start a Marimo notebook session (after adding it to your dependencies), run:

BASHCopied
marimo edit 

This launches the notebook UI, where you can initialize your Django project with:

PYTHONCopied
import marimo as mo
import django
django.setup()

Then, interact with the ORM as you would in a Django shell:

PYTHONCopied
from apps.my_app.models import SomeModel
SomeModel.objects.all()

Working with Subfolders

One appealing aspect of Marimo is its ability to organize notebooks in a separate folder (e.g., analysis), keeping exploratory work version-controlled and isolated from the main project repo. From the project root, you can run:

BASHCopied
marimo edit ./analysis/my-analysis-notebook.py

This setup only requires adding the marimo library to your development dependencies, ensuring your open source Django project remains clean while you work privately on data analysis.

I'm very impressed with Marimo so far and would love to hear about others' experiences—both good and bad—with using it alongside Django.