Many user workflows assume that they have control over their own environment. For example, by using conda to install custom packages. For these users, they also assume that this environment will persist between their hub sessions.
While there are some risks to letting users control their own environments, this is a very common workflow. We should find some way to make it technically possible to persist user environments between sessions, and to document best-practices in how to do this.
This has already been discussed and requested a few places:
This is an important task because it's very common across many data science workflows, and is part of many training sessions (e.g. "managing your environment" is a standard part of most intro courses). I suspect that this is something many people will just "assume" exists already so we will get more requests about it if we don't have a story for this use-case.
Tasks to complete
[ ] Decide how we'd like to incorporate this into our hub infrastructure setup (e.g., nb_conda_kernels)
Summary
Many user workflows assume that they have control over their own environment. For example, by using
condato install custom packages. For these users, they also assume that this environment will persist between their hub sessions.While there are some risks to letting users control their own environments, this is a very common workflow. We should find some way to make it technically possible to persist user environments between sessions, and to document best-practices in how to do this.
This has already been discussed and requested a few places:
Value
This is an important task because it's very common across many data science workflows, and is part of many training sessions (e.g. "managing your environment" is a standard part of most intro courses). I suspect that this is something many people will just "assume" exists already so we will get more requests about it if we don't have a story for this use-case.
Tasks to complete
nb_conda_kernels)