Jeremy Tuloup

Ah looks like it's not packaged for conda. We should open a PR on https://github.com/conda-forge/staged-recipes to add it.

> What is the reason conda-forge packages work, but pip packages don't? They will eventually be supported: https://github.com/jupyterlite/xeus-python-kernel/pull/102

> I suffer from conda-forge packaging fatigue :) Well often you can generate most of the recipe with `grayskull`: https://github.com/conda/grayskull

> fails downstream repos building with the error `...and cannot be used with 'export *'.`. You mean in a downstream repo reusing the extension, but without webpack?

Not sure it's worth looking into this for now. We know this kernel is not meant to be used with JupyterLab at least for now. Also folks usually create a...

> it parses the imports and automatically installs standard packages that are being imported: Right, iirc even code like the following would trigger the install? ```py if False: import matplotlib...

Actually this can already easily be tested on any JupyterLite deployment that has `anywidget` installed as a build dependency (to properly pick up the lab extension). For example here: https://jtp.io/lit/lab/index.html?path=pyodide/anywidget.ipynb...

> I'm guessing you need a preconfigured JupyterLite instance with this enabled. Yes, the `anywidget` Python package needs to be installed in the build environment so its JupyterLab extension can...

Indeed this would be a low-friction way for trying something out quickly! @manzt how do you see something like this integrated in the repo? I quickly looked into adding this...

> This could complicate the Netlify build and setup. Maybe a first step could be to deploy this JupyterLite demo to GitHub pages, and link to a full blown JupyterLite...