Hi,
I was wondering if it would be possible to make an option for Jupyter Lab launch that would NOT create a new virtual environment, but use the default system environment instead?
Thanks,
Bojan
We don’t recommend installing packages without a virtual env or container, as it makes it harder to cleanup and work on different projects with different requirements.
So that’s a no, right?
We will not be adding it as a feature
OK, fine, then I’ll just make my own version of the app that allows users far more flexibility.
I understand the general position, however, Jupyter can look through and even register multiple environments/kernels as options.
I find that I almost never use the Dashboard Jupyterlab because it is a “bare Jupyterlab” environment, where I most frequently need extra packages. One good example is the Nvidia Legate ecosystem. This works very nicely on Spark and is recommended to deploy via Conda. The Dashboard feature deploys via a uv venv particular to whatever directory you select… you can’t easily use Legate with the Dashboard.
It would be really nice to be able to select from the available Conda or top-level virtual environments, perhaps limited to those which have jupyterlab installed, when starting from the Dashboard. Then we could choose to work within a curated and reproducible environment with the tooling for a particular project, rather than a minimal-Jupyterlab venv that is probably fragile and incomplete.