React front-end Jupyterhub backend

I’m looking to set up a server environment for students where they can log into an isolated environment, use a react front-end to select some dropdown options (dataset, ml algorithm, columns etc) and then some statistics/graphs are generated. The idea is that people with little code experience can try out machine learning.

I was wondering if this is something I could manage with Jupyterhub as the account login/setup/security/isolation is taken care of?

Ideally beginners would only see the React front-end and advanced users could switch between the Front-end and Jupyterhub.

Any pointers would really be appreciated!

Voila can do something along these lines without needing React.
See Deploy Your Machine Learning Model with Flask or Voila | by Stephanie Kirmer | Geek Culture | Medium and go to the Voila Gallery and click on some of the boxes to see examples running.

Voila renders notebooks built with ipywidgets to apps. Once you know what you are doing, it’s easy to use the URL to switch to the Jupyter Notebook view of the same notebook. You can go here press launch binder and then select from the last three notebooks among the list of available notebooks to get a sense of how it all works together. Especially focus on the last two notebooks for the Voila aspects.

This demo by Fastai using Voila to show a classifier of images trained on bears seems similar to what you describe, in broad strokes. Except right now it is broken; I’m not familiar with it enough to recall how to fix it. I thought I helped someone with getting a similar site to work in the last few months. UPDATE: this variation looks to work better!

2 Likes

This is perfect thank you so much!

There is also Mercury that works directly with Jupyter notebooks, too. I haven’t tried it yet.

1 Like