jupyter-opensysml-kernel

A Jupyter kernel for SysML v2, backed by the OpenSysML REPL. A notebook cell is a REPL submission: declarations accumulate into one model across cells, a bare expression is evaluated, and every % command of the sysml REPL — %eval, %instantiate, %action, %render, %render-document, %load, … — works as it does at the prompt. Views render as Mermaid, Graphviz (SVG), Markdown, CSV and JSON; documents as Markdown and HTML.

Install

Python 3.10 or later and a Jupyter front end (JupyterLab, Notebook, VS Code, nbconvert, …) are required.

pip install jupyter-opensysml-kernel

The wheel for your platform (Linux x64 and arm64, macOS Intel and Apple Silicon, Windows x64) bundles the sysml-jupyter-kernel binary of the OpenSysML release this package was built for and registers the sysml kernelspec under the Python environment it is installed into, so that is the whole install:

jupyter kernelspec list            # shows: sysml  …/share/jupyter/kernels/sysml
jupyter lab                        # pick "SysML v2 (OpenSysML)"

Install with the Python that runs your notebook server (or in its environment); pip uninstall removes the kernel and its kernelspec together.

Where no wheel applies, pip installs the sdist, which bundles no binary. Then one more command downloads the release's binary, verifies it against the SHA-256 digest the package ships for that release, and registers the kernelspec with the binary inside it:

python -m jupyter_opensysml_kernel install

Nothing unverified is installed: a download that does not hash to the pinned digest is refused, and so is a release the package pins no digest for. The same command registers the kernel for the user (--user) or under another prefix (--prefix DIR, --system) when the notebook server does not run in the environment the package is in, copying the bundled binary when there is one. To use a kernel built from source (go build ./cmd/sysml-jupyter-kernel) or installed another way, pass it explicitly:

python -m jupyter_opensysml_kernel install --binary ./sysml-jupyter-kernel

python -m jupyter_opensysml_kernel uninstall removes a kernelspec install wrote and the binary in it; jupyter kernelspec remove sysml does the same.

The conda-forge recipe for this package is kept in the OpenSysML repository under packaging/conda; once it is accepted there, conda install -c conda-forge jupyter-opensysml-kernel installs the kernel and its kernelspec into the environment directly. Until then, pip install inside the conda environment registers the kernel under that environment's prefix the same way.

A development snapshot of this package is published every night as jupyter-opensysml-kernel==<next release>.dev<yyyymmdd>, which pip install never picks up on its own; it carries the kernel of the same night's snapshot.

Documentation

• Using SysML in Jupyter — the guide chapter: cells, % commands, rich output, interrupts.
• REPL commands — every % command the kernel serves.

Development

pip install -e '.[dev]'
pytest
mypy jupyter_opensysml_kernel

The tests need no network: downloads are served from fixtures, and the wheel tests build from a staged stand-in binary (python -m build is required). The package's copy of release-digests.json is synced from client/release-digests.json by scripts/sync-release-digests.py; the release pipeline stamps the digests of the kernel binaries it built into the copy, then scripts/build-jupyter-kernel-dist.sh builds the sdist and, with each binary staged under jupyter_opensysml_kernel/bin/ and JUPYTER_OPENSYSML_KERNEL_PLATFORM=<goos>-<goarch> set, one wheel per platform (setup.py tags the wheel and adds the kernelspec as shared data).

License

Apache-2.0.