Notebooks

Diagnostics in cells

In VS Code each Jupyter notebook cell reaches the language server as its own document (vscode-notebook-cell: URIs). PyPilot checks Python cells like files: missing imports, imports that cannot run on the environment's Python, and removed attributes, with the same quick fixes. Markdown cells are skipped, and IPython magics and shell escapes (%pip install …, !ls) are masked out before analysis.

Quick fixes work across the notebook: select several imports to install them in one step, or use Install all N missing packages in this notebook.

Untitled notebooks are checked too; a saved notebook additionally lets PyPilot find local modules next to it.

A kernel that starts

A notebook can only run in an environment that has ipykernel. In projects with notebooks, whenever PyPilot sets up the environment or installs a package, it makes sure .venv has:

ipykernel>=6.31.0,<7

and warns when it is missing. In uv projects it is added as a dev dependency.

Why this range:

• <7: on ipykernel 7.x (reported on 7.1.0 and 7.2.0) VS Code notebooks can hang with cells stuck "pending", while 6.31.0 runs them. The maintainers' advice is to pin below 7 (vscode-jupyter#17228).
• >=6.31.0: the newest 6.x, released after Python 3.14 and tested on it. The 3.14 "kernel won't connect" failure was specific to 7.0.0 (ipykernel#1445). The floor also stops a cached old release from resolving.

If you install ipykernel yourself, your requirement is used instead.

Selecting the kernel

With pypilot.autoSelectVenv (on by default) the VS Code extension makes the project environment the active interpreter, which is also what the Jupyter kernel picker recommends. In pixi projects that is .pixi/envs/<name>. PyPilot: Use Project Environment (.venv / pixi) does it on demand.