Vscode jupyter notebooks do not find kernels installed by Pixi

[pakiessling]

Hi,
absolutely love Pixi!
However I have one small issue.

VScode does not seem to find my pixi enviroments in Jupyter notebooks.

This is a fresh pixi enviroment with ipykernel and pandas in it:

I can work around this by doing

pixi shell
ipython kernel install --user --name=my_pixi

Then I can see the Kernel in the category Jupyter Kernels.
However, from what I have seen this should not be necessary?
Am I doing something wrong?

Thank you!

OSX 14.4.1
Visual Studio Code - Insiders 1.96
Python pre-release 2024.21.2024112901
Pixi VSCode 1.0.4

[ruben-arts]

I'm personally not a big user of Jupyter, pinging @jjjermiah and @renan-r-santos as they are the maintainers of Pixi VSCode and pixi-kernel Which might be of help to you!

[jjjermiah]

@pakiessling could you share your pixi.toml manifest please?

[pakiessling]

Sure

[project]
authors = ["pakiessling <[email protected]>"]
channels = ["conda-forge"]
description = "Add a short description here"
name = "metadata"
platforms = ["osx-arm64"]
version = "0.1.0"

[tasks]

[dependencies]
pandas = ">=2.2.3,<3"
ipykernel = ">=6.29.5,<7"
pip = ">=24.3.1,<25"
openpyxl = ">=3.1.5,<4"

[jjjermiah]

couple considerations:

• after running pixi shell, can you try opening command palette and clearing the python env cache
  

then try again?

[pakiessling]

Hm after clearing the cache and once again entering pixi shell the enviroment suddenly appeared as "recommended". Interesting...

[herrgahr]

Hi there! Leaving this for everyone who still faces this problem and finds this thread via web search...

If your VS Code still can't find kernels, you may be using Code OSS instead of the official binary published by Microsoft.

In this case, you will probably see errors in the "Python" output panel and / or the "Jupyter" output panel, telling you you have to enable some "proposed" api via a command line switch.

The TL;DR of how to do this is to create the file ~/.config/code-flags.conf with the following contents:

--enable-proposed-api=ms-python.python
--enable-proposed-api=ms-toolsai.jupyter

Note: the filename code-flags.conf is corrent for Code OSS on Arch Linux, this may be different for your OS / distribution