Enable opening notebooks without a kernel

[jtpio]

References

Fixes #3170

Build on top of the upstream JupyterLab changes.

Code changes

• Allow opening notebooks without a kernel
• Test

User-facing changes

notebook-no-kernel.mp4

Backwards-incompatible changes

None

[github-actions]

👈 Launch a Binder on branch jtpio/notebook/issue-3170-open-without-kernel

[Copilot]

Pull request overview

Adds standalone Notebook support for opening notebooks without starting or reusing a kernel.

Changes:

• Enables the upstream no-kernel context-menu action.
• Propagates no-kernel preferences through notebook URLs.
• Adds an end-to-end test confirming no session starts.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.

File	Description
app/package.json	Enables the no-kernel notebook plugin.
packages/application-extension/src/index.ts	Applies no-kernel preferences when opening routed notebooks.
packages/docmanager-extension/src/index.ts	Propagates no-kernel state to new tabs.
ui-tests/test/filebrowser.spec.ts	Tests opening without a running session.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Daniele-Cangi]

Hi @jtpio, I tested this branch locally and found two edge cases around carrying the no-kernel state between the tree page and the notebook tab.

The first is that shouldStart === false is currently enough to serialize kernel=none, while the receiving page expands that back to { shouldStart: false, shouldReuse: false }. This changes the semantics of notebookStartsKernel=False: it should prevent auto-starting a kernel, but still allow reconnecting to an existing session.

The second is after explicitly opening with Notebook (no kernel). If a kernel is then started, kernel=none remains in the URL. Reloading the page therefore applies the no-reuse preference again instead of reconnecting to the running session.

I added regression coverage and a small fix on top of the current PR head (e2ef70022df07f2b15c2b1420dc124e4896148ef):

Daniele-Cangi@40217c3

The two regression tests fail on the original PR head and pass with the patch. The full Chromium filebrowser.spec.ts suite also passes (9/9). The reload case additionally verifies that it reconnects to the same session ID and that only one server session remains for the notebook.

Happy to open this as a small PR against your feature branch if that would be useful, or you can cherry-pick the commit directly.