nb_conda_kernels not detecting conda environments

[dagingehelgoy]

I am struggeling to get my nb_conda_kernels to detect my conda envs. While jupyter kernelspec only sees my base environment, conda has listed 3.

(base) jupyter kernelspec list
Available kernels:
  python3    C:\ProgramData\Anaconda3\share\jupyter\kernels\python3

# conda environments:
#
base                  *  C:\ProgramData\Anaconda3
vsb                      C:\ProgramData\Anaconda3\envs\vsb
test_gpu                 C:\Users\DayNoone\.conda\envs\test_gpu

However, it looks like nb_conda_kernels finds one of these environments (test_gpu)

[tdpetrou]

Looks like this is an issue with conda channel that you have downloaded nb_conda_kernels from. The latest commit was not pushed to conda defaults or conda-forge. It is available on jupycon.

conda install jupycon/label/dev::nb_conda_kernels

Whats happening is that the active environment is not showing up. This is an issue with the manager.py file. Before the last commit it has this line in the _all_envs method which was skipping the current environment.

if env_path == sys.prefix:
    continue

I would just install from the above channel. As a bad workaround you can start your notebooks from an environment other than the one you would like to actually run.

[Zsailer]

This difference between your kernelspec list and your notebook's available kernels is the consequence of nb_conda_kernels' design.

It does not add kernelspecs to your jupyter data files. Rather, it dynamically creates kernels from your conda environments when jupyter is launched using a custom CondaKernelSpecManager object.

[mcg1969]

The screenshot you posted is for the nb_conda extension, which is not the same as nb_conda_kernels.

[mcg1969]

Also, jupyter kernelspec cannot make use of nb_conda_kernels... yet. See here.