Description
cc @pavithraes @tcmetzger @philippjfr
Some thoughts:
I was thinking that we might simply have two versions of the intro and installation notebooks? Keep the ones we have now for actually running locally, but then have "pyodide" versions that do the necessary micropip installs, etc. Then a static hopefully we can point a tutorial.bokeh.org domain at the pyodide versions? I am less clear on how that might work, but I assume whatever happens on the Jupyter-lite site is something we can do ourselves? (These are ultimately just static pages after all).
What I would also wonder about is how to handle the helper functions and data that the tutorial use... should be just add a bokeh-tutorial package that can be pip / micropip installed that has all this stuff? Though the Jupyter-lite site has a data folder so maybe we can just point our our jupyter-lite setup at a repo tag or something?
Edit: here are the jupyter-lite docs for building your own on GH pages: https://jupyterlite.readthedocs.io/en/stable/quickstart/deploy.html
Activity
bryevdv commentedon Feb 19, 2025
I also think we might want to switch the install step from conda to a purely pip-based process to keep things consistent? We don't have many requirements at all so hopefully that would be simple.
ianthomas23 commentedon Feb 19, 2025
I've created a demo repo of the first page of the tutorial at https://github.com/ianthomas23/bokeh-tutorial-pyodide. It auto deploys to github pages at https://ianthomas23.github.io/bokeh-tutorial-pyodide/lab/index.html. The full dashboard works on the page, so I assume there shouldn't be too many problems getting the whole repo worklng.
I made just a couple of changes, moving the
datadirectory to within thenotebooksdirectory and using%pip install -q geopandasin the notebook cell before we use anygeopandascode. This would need some text explaining it. It is possible to configure it to auto open the first tutorial page.If everyone is in agreement I can switch this full repo to work the same. It works just as well locally of course.
Disclaimer: you may not all know that I work on JupyterLite as part of my day job 🎉
bryevdv commentedon Feb 19, 2025
I did not make that connection @ianthomas23 very happy to have any and all help and expertise on this :D
👍 from me!
pavithraes commentedon Feb 21, 2025
Thanks, @ianthomas23! I'm onboard to switch as well!