Python language server never starts properly the first time the notebook is opened

[guidocioni]

Environment data

• VS Code version: 1.98.2
• Jupyter Extension version: v2025.2.0
• Python Extension version : 2025.2.0
• OS (Windows | Mac | Linux distro) and version: MacOS 13.7.4
• Python and/or Anaconda version: 3.12.8
• Type of virtual environment used: conda
• Jupyter server running: Local

Behaviour

I'm quite puzzled as to why never reported this issue because I've been seeing it for months on different machines.
Basically EVERY time I open a Notebook for the first time, the Python Language Server is not initialized correctly and this basically makes the whole IntelliSense part useless. The Notebook functionality is not affected though.

See as example this video capture.

bug.mov

What I'm doing is

• Opening a Notebook where a simple pandas import is done to show the package version.
• Even though the default environment is activated (and this environment has pandas installed), IntelliSense shows a message Import pandas could not be resolved from source
• Switching to base environment shows the same message, and this is expected because pandas is not installed there
• Switching back to default correctly loads the language server, and the message disappears (as expected because pandas is installed in this env).

Notes:

• Restarting the extensions host does solve the issue
• Leaving the VsCode window open and re-opening the Notebook does not cause the issue to appear (probably because the Extensions has been already restarted)
• Restarting the Python Language Server does not solve the issue

My guess is that there is something preventing the Extension to start properly the first time but I do not seem anything particular from the logs...
What could I look into to understand better where the issue is coming from?

Here are detailed logs of the python language server

Details

2025-04-02 10:10:31.106 [info] [Info  - 10:10:31 AM] (56418) Starting service instance "<default>"
2025-04-02 10:10:31.211 [info] [Info  - 10:10:31 AM] (56418) Setting environmentName for service "<default>": "3.12.8 (default conda)"
2025-04-02 10:10:31.211 [info] [Info  - 10:10:31 AM] (56418) Setting pythonPath for service "<default>": "/Users/guidocioni/miniforge3/envs/default/bin/python"
2025-04-02 10:10:31.212 [info] [Info  - 10:10:31 AM] (56418) No include entries specified; assuming /<default workspace root>
2025-04-02 10:10:31.212 [info] [Info  - 10:10:31 AM] (56418) Auto-excluding **/node_modules
2025-04-02 10:10:31.212 [info] [Info  - 10:10:31 AM] (56418) Auto-excluding **/__pycache__
2025-04-02 10:10:31.212 [info] [Info  - 10:10:31 AM] (56418) Auto-excluding **/.*
2025-04-02 10:10:31.260 [info] [Info  - 10:10:31 AM] (56418) Assuming Python version 3.12.8.final.0
2025-04-02 10:10:31.332 [info] [Error - 10:10:31 AM] (56418) File or directory "/<default workspace root>" does not exist.
2025-04-02 10:10:31.333 [info] [Info  - 10:10:31 AM] (56418) No source files found.
2025-04-02 10:10:31.346 [info] [Info  - 10:10:31 AM] (56418) Starting service instance "<default>"
2025-04-02 10:10:31.350 [info] [Info  - 10:10:31 AM] (56418) Setting environmentName for service "<default>": "3.12.8 (default conda)"
2025-04-02 10:10:31.350 [info] [Info  - 10:10:31 AM] (56418) Setting pythonPath for service "<default>": "/Users/guidocioni/miniforge3/envs/default/bin/python"
2025-04-02 10:10:31.350 [info] [Info  - 10:10:31 AM] (56418) No include entries specified; assuming /<default workspace root>
2025-04-02 10:10:31.350 [info] [Info  - 10:10:31 AM] (56418) Auto-excluding **/node_modules
2025-04-02 10:10:31.350 [info] [Info  - 10:10:31 AM] (56418) Auto-excluding **/__pycache__
2025-04-02 10:10:31.350 [info] [Info  - 10:10:31 AM] (56418) Auto-excluding **/.*
2025-04-02 10:10:31.375 [info] [Info  - 10:10:31 AM] (56418) Assuming Python version 3.12.8.final.0
2025-04-02 10:10:31.416 [info] [Error - 10:10:31 AM] (56418) File or directory "/<default workspace root>" does not exist.
2025-04-02 10:10:31.416 [info] [Info  - 10:10:31 AM] (56418) No source files found.
2025-04-02 10:10:31.431 [info] [Info  - 10:10:31 AM] (56418) Setting environmentName for service "<default>": "3.12.8 (default conda)"
2025-04-02 10:10:31.431 [info] [Info  - 10:10:31 AM] (56418) Setting pythonPath for service "<default>": "/Users/guidocioni/.vscode/extensions/ms-python.vscode-pylance-2025.3.2/dist"
2025-04-02 10:10:31.431 [info] [Info  - 10:10:31 AM] (56418) No include entries specified; assuming /<default workspace root>
2025-04-02 10:10:31.431 [info] [Info  - 10:10:31 AM] (56418) Auto-excluding **/node_modules
2025-04-02 10:10:31.432 [info] [Info  - 10:10:31 AM] (56418) Auto-excluding **/__pycache__
2025-04-02 10:10:31.432 [info] [Info  - 10:10:31 AM] (56418) Auto-excluding **/.*
2025-04-02 10:10:31.457 [info] [Error - 10:10:31 AM] (56418) File or directory "/<default workspace root>" does not exist.
2025-04-02 10:10:31.457 [info] [Info  - 10:10:31 AM] (56418) No source files found.
2025-04-02 10:10:31.755 [info] [Info  - 10:10:31 AM] (56418) Priority queue background worker(2) root directory: file:///Users/guidocioni/.vscode/extensions/ms-python.vscode-pylance-2025.3.2/dist
2025-04-02 10:10:31.755 [info] [Info  - 10:10:31 AM] (56418) Priority queue background worker(2) started
2025-04-02 10:10:31.887 [info] [Info  - 10:10:31 AM] (56418) Priority queue background worker(4) root directory: file:///Users/guidocioni/.vscode/extensions/ms-python.vscode-pylance-2025.3.2/dist
2025-04-02 10:10:31.887 [info] [Info  - 10:10:31 AM] (56418) Priority queue background worker(4) started

[DonJayamanne]

@rchiodo Please can you have a look.

[guidocioni]

Other things that I tried:

• completely excluding all other Python enviroments from the jupyter kernels (using jupyter.kernels.excludePythonEnvironments), this means default is automatically picked once I open a notebook
• try to unset python.defaultInterpreterPath
• restarting linter, formatter...

Unfortunately nothing fixes this issue, which is coming from pylance.

Until now the only way to I have to fix this is to restart the Extension Host KEEPING the notebook opened in VsCode. Any other method fails.

[rchiodo]

What's your version of Pylance? This could be this issue here:
microsoft/pylance-release#7045

There was a problem with responding to kernel changes which should be fixed in our latest prerelease bits. I can't reproduce this issue with our latest Pylance prerelease bits.

[guidocioni]

What's your version of Pylance? This could be this issue here: microsoft/pylance-release#7045

There was a problem with responding to kernel changes which should be fixed in our latest prerelease bits. I can't reproduce this issue with our latest Pylance prerelease bits.

I just installed Pylance 2025.3.105 and I still see the same behaviour. Opening for the first time doesn't properly load the language server. Right now I excluded all python environments but one so when I open the notebook it auto selects default, I don't know if this is another edge case...
After selecting again default in the kernel picker it starts to work properly.

bug.mov

[rchiodo]

That seems different. Not even sure how you get a 'default' kernel? When I create a notebook, I just get this:

Which would show those pandas errors until we receive a kernel selection event.

@DonJayamanne we still use the jupyter extension's API to get the kernel for a notebook on open. Would that return this default kernel?

[guidocioni]

That seems different. Not even sure how you get a 'default' kernel? When I create a notebook, I just get this:

Which would show those pandas errors until we receive a kernel selection event.

@DonJayamanne we still use the jupyter extension's API to get the kernel for a notebook on open. Would that return this default kernel?

Default is the name of a conda env, the one that i always use to run Jupyter notebooks 😉
It gets auto selected at start in my case because I disabled all other python environments (see the comment befor about jupyter.kernels.excludePythonEnvironments).
Could it be that, in case the kernel is not selected from the user but auto selected by vscode, then no update event is triggered?

[rchiodo]

Could it be that, in case the kernel is not selected from the user but auto selected by vscode, then no update event is triggered?

That's likely, but I wouldn't expect an update on open. There should just be a value returned when we ask for the 'current' kernel. This works fine on saved notebooks (they don't need an update event after open), so I suspect the API we call in the jupyter extension isn't handling this default case for some reason.

[guidocioni]

Could it be that, in case the kernel is not selected from the user but auto selected by vscode, then no update event is triggered?

That's likely, but I wouldn't expect an update on open. There should just be a value returned when we ask for the 'current' kernel. This works fine on saved notebooks (they don't need an update event after open), so I suspect the API we call in the jupyter extension isn't handling this default case for some reason.

Is the kernel picked supposed to be persisted? Because one of the reason why I excluded the other python interpreters is exactly because every time I open the notebook I have to select the kernel again and again...

[rchiodo]

Yeah it's supposed to be persisted on save.

[guidocioni]

Yeah it's supposed to be persisted on save.

Ok then it's yet another issue ... Do you know how to understand why that is not happening?
But anyway do you think what I'm seeing is expected behavior or not?

[rchiodo]

Kernel save works for me, so not sure what the issue would be. Should probably enter a separate issue.

I wouldn't say the 'default' not working would be by design though. It should work. I need to repro to see if it's a problem in the Jupyter extension or the Pylance extension.

[rchiodo]

What exactly do you put for the 'excludePythonEnvironments'? It's very vague about how it matches.

[guidocioni]

Kernel save works for me, so not sure what the issue would be. Should probably enter a separate issue.

I wouldn't say the 'default' not working would be by design though. It should work. I need to repro to see if it's a problem in the Jupyter extension or the Pylance extension.

I think you should be able te reproduce by just excluding all python interpreters besides the one you want. However, if you don't have any save issue in theory in your case it will automatically load that kernel anwyay so I'm not sure it is useful.

Just to understand better, once you save a notebook with a kernel and then you reopen it, it then automatically select that kernel and correctly use this as interpreter in pylance?

The only difference I could think about is that I have multiple conda environments, and the base one is NOT the one I'm using for jupyter. Maybe you only have 1 python installation? Which means it will always load the right one no matters what....