[petercorke]

The JupyterLite site has been installing the last release of the toolbox from PyPI, not the wheel built with it, since #848. This fixes that, and adds the checks that would have caught it. No change to the toolbox.

The bug

The bootstrap cell (#848, copied from MVTB's) preferred "the wheel bundled with this site" using Path("/pypi").glob(...). There is no /pypi folder in a JupyterLite kernel (checked in Chromium with the pinned kernel 0.6.1), so that branch never ran and it always did micropip.install("roboticstoolbox-python") from PyPI. "Try it Now" therefore did not track main, and fixes such as #860 were not in the site. The old piplite.install(...) cells did pick up the bundled wheel, because piplite looks in the site's own index first, so #848 regressed this. MVTB has the same code: petercorke/machinevision-toolbox-python#160.

Nothing noticed, because the Node check mounted the wheel in Pyodide's filesystem and redirected the pypi/ install to it.

The fix

In the browser, install through piplite (the site's index first), falling back to micropip when there is no piplite, for example under Node. The generated bootstrap cell is regenerated in all 13 notebooks; nothing else in them changes.

The checks

1. Node/Pyodide check (docs/lite/wasm_check/): a stand-in piplite installs the wheel (replacing the pypi/ redirect), and a new step plots a trajectory under the ipympl backend in real Pyodide and checks that the inline mode is chosen and a different image is pushed for every step. It fails with the code from before fix(plot): a trajectory is animated whatever %matplotlib says #860 (AssertionError: notebook-widget).
2. Real-browser check (docs/lite/browser_check/, new, Playwright and Chromium): builds a small site with the wheel and one notebook, runs it, and requires the bootstrap line; the hash of the installed PyPlot.py to equal the checkout's (so the wheel built here is the one installed); an ipywidgets slider; an ipympl figure of one pose; and at least 5 different images while an animation cell runs. It is a docs.yml step after the site is built, informational (continue-on-error) until it has passed once on main, with the screenshot kept as an artifact on failure. Run locally it passes in about 2 minutes, and fails with the old bootstrap (the toolbox in the browser is not the wheel built from this checkout ... installed the last release from PyPI).

This also confirms in a real JupyterLite site that ipywidgets and ipympl work there, and that a trajectory animates with the fix.

Please note

Merging this changes the generated bootstrap cell, and only that cell, in every notebook. It merges cleanly into a branch that has not edited that cell.

18 tests in tests/test_notebook_bootstrap.py (one new: the bootstrap installs through piplite and not /pypi). I have not seen the new CI step run in GitHub Actions yet. Written with AI assistance (Claude Code).

🤖 Generated with Claude Code

[codacy-production]

Not up to standards ⛔

🔴 Issues 4 critical · 6 high · 2 medium · 1 minor

Alerts:
⚠ 13 issues (≤ 0 issues of at least minor severity)

Results:
13 new issues

Category	Results
Security	1 minor 6 high 4 critical 2 medium

View in Codacy

🟢 Metrics 39 complexity · 0 duplication

Metric	Results
Complexity	39
Duplication	0

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 0.00%. Comparing base (6497cdc) to head (95a8d60).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@          Coverage Diff          @@
##            main    #863   +/-   ##
=====================================
  Coverage   0.00%   0.00%           
=====================================
  Files        146     146           
  Lines      17164   17164           
=====================================
  Misses     17164   17164           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.