🚩 Plotolo

A full-stack Python framework for creating interactive web applications from Jupyter Notebooks.

Experiments for serving Jupyter Notebook files through Tornado websockets as an interactive web application.

Guide

todo

How it works 🤷🏽

(Streamlit + Jupyter Notebook) =:

• The client connects to the server through websocket.

• The server reads and compiles the Notebook's source code and executes it cell by cell.

• The Notebook's source code calls widget functions.

• The widgets are sent to the client through the server.

• The client displays the widgets as UI components.

• The client sends widget input to the server on user interactions.

• The server partially re-executes the Notebook with the new input.

• +1 Works embedded inside Jupyter without running a server using in-browser Javascript.

So it is basically Streamlit, but it

• does not rerun the whole script on inputs, only the given cells (imitating the normal data science flow: when I modify a cell, I rerun it with the following cells).
• can work without a server and display widgets inside the Notebook during development.

So it is better. 💩

Example flow:

Architecture:

Project structure 🏗

This project is a monorepo containing the following:

• backend (lib): Tornado Web Server + IPython widgets
• frontend: Next.js Web Application
• widget(s): React Component Library

Detailed architecture:

How to run 🏃🏽

Currently only the dev mode is supported.

Prerequisites:

• Python >= 3.9 + PiP
• Node.js >= 18.0

Install dependencies:

pip install -r requirements.txt
cd ./frontend
npm install

Embedded mode from a Notebook:

Use an existing example or create a new Notebook:

cd ./examples
jupyter notebook

Standalone server mode:

Start backend:

python ./main.py

Start frontend:

cd ./frontend
npm run dev

Then go to http://localhost:3000

Start widget storybook:

cd ./widget
npm run storybook

Then go to http://localhost:6006