Introduction

Pyodide and PyScript enable running a native CPython interpreter in browser. There are several projects that leverage this including jupyterlite. The idea of this project is to enable easily running code via an annotation in a remote ipykernel or locally in the web browser seamlessly. This is similar to marking a function via the decorator dask.delayed and having the option to scale out a given computation.

The only dependency for this project is cloudpickle installed in the ipykernel environment.

Usage

cd src
python -m http.server &
jupyter-server --ServerApp.allow_origin='*' --ServerApp.token='asdfqwerzxcvqwer'

Load in your web browser localhost:8000 and run the following code.

import asyncio
import jupyter

kernel_id, kernel = await jupyter.jupyter_client.ensure_kernel()
print(await kernel.send_code('costrouc', 'print(1)'))


def add(a, b):
    return a + b


@kernel.delayed
def mult(a, b):
    return a * b


@kernel.delayed
def get_host():
    import os
    import socket
    return os.getcwd(), socket.gethostname(), os.getuid()


print(add(1, await mult(3, 4)))
print(await get_host())

costrouc/pyodide-jupyter-kernel

Introduction

Usage