Sphinx-Gallery JupyterLite helpers#

_sphinx_gallery_jupyterlite has two independent roles:

a Sphinx extension that injects the documentation release value into the JupyterLite warning/bootstrap content; and
public helper callables used from sphinx_gallery_conf to prepare notebooks and reset optional library state between gallery examples.

Enable the release-version hook#

extensions += [
    "scikitplot._externals._sphinx_ext._sphinx_gallery_jupyterlite",
]

At builder-inited the extension uses app.config.release as the authoritative documentation version when it is non-empty. Outside a Sphinx build it falls back to installed scikit-plots package metadata and finally to "unknown" with a warning.

Connect it to Sphinx-Gallery#

The helper callables are referenced by dotted path:

sphinx_gallery_conf = {
    "reset_modules": (
        "scikitplot._externals._sphinx_ext._sphinx_gallery_jupyterlite.reset_others",
        "matplotlib",
        "seaborn",
    ),
    "jupyterlite": {
        "notebook_modification_function": (
            "scikitplot._externals._sphinx_ext._sphinx_gallery_jupyterlite."
            "notebook_modification_function"
        ),
    },
}

`notebook_modification_function`#

The notebook hook prepends the generated JupyterLite notebook with the Scikit-Plots environment/bootstrap cells needed by the current source. It uses explicit detection tokens for optional packages and dataset-fetching patterns rather than installing every optional dependency unconditionally.

The function asserts the Sphinx-Gallery notebook-cell helper contract after prepending its cells so an incompatible upstream API change fails visibly instead of quietly producing a malformed notebook.

`reset_others`#

The reset hook runs garbage collection and, when the optional libraries are installed, restores the library state the gallery relies on. The current implementation has explicit handling for scikit-learn, Plotly and PyVista. Missing optional packages are not treated as errors.

Customization boundary#

The module exposes its behavior through documented module-level constants for package-detection tokens, HTTP setup and Pyodide base imports. Override those constants before Sphinx begins processing notebooks if a documentation project needs a different JupyterLite bootstrap policy.

This helper is for documentation builds. It should not become a runtime requirement of ordinary Scikit-Plots plotting code.