[Spike: 8hr] Investigate using JupyterLite for Dubois content

[jmunroe]

Context

We support https://github.com/HigherEdData/Du-Bois-STEM by hosting a JupyterHub https://dubois.2i2c.cloud/ on GCP. The cloud cost are running $400+/month. There is no persistent, per-user data.

The libraries (there are both R and Python notebooks) used are standard and may already be sufficiently supported to be served using a WASM / JupyterLite kind of solution.

What we need to do

Convert the notebooks in https://github.com/HigherEdData/Du-Bois-STEM (at least one R and one Python notebook) to work as browser only deployments using JupyterLite.

Identify functions and libraries where a WASM backed version of these notebooks works and does not currently work.

Use the notebooks found in https://github.com/HigherEdData/Du-Bois-STEM/tree/main/notebook_keys.

Definition of Done

• Explorer how to setup a serverless JupyterLite deployment
• Set up and test with one (or more) Python notebook
• Set up and test with one (or more) R notebook
• Write up findings (this works today, this does not yet work without further development)
• Meet with community representative from the Du Bois project to report back

[aprilmj]

We want to accomplish 2 things as we approach this work:

• Solve a problem for Dubois, so they have a lighter solution for right now
• Make sure that solution is supportable in the longer term
• Learn what we could/should do in the future - feed into our roadmap (could just leave ourselves an outline of a JupyterLite MVP that we fill in more completely in a future sprint, to be learned as we go)