| Status |
Open for comments 💬 |
| Author(s) |
@kenafoster |
| Date Created |
24-07-2026 |
| Date Last updated |
24-07-2026 |
| Decision deadline |
TBD |
Should Data Science Pack Require Nebi Pack Installed Separately?
Summary
The data-science-pack currently assumes nebi-pack is installed. However, the "in-JupyterLab" Nebi works on its own. Furthermore, the local Nebi still supports connecting to OCI registries, so teams can still share "approved/curated" environments amongst themselves. The Remote<>Local functionality is extra that can be confusing.
Note that it is technically possible to install data-science-pack without Nebi pack, and appears at first glance that the end user behavior is as expected. This RFD is more oriented around asking if the intent is that DS pack should require a separate Nebi pack install.
User benefit
Users of the data-science-pack will not have to manage the nebi-pack as well. Though it provides additional features, it may not be necessary in all use cases so the implied dependency adds unnecessary administrative burden.
Design Proposal
Since most of the design is already in place, the only modifications would be to update the data science pack helm chart a bit to remove some resources that are unnecessarily created when remote Nebi is not there. Here is a draft PR:
#192
Alternatives or approaches considered (if any)
Alternative is the status quo.. but related... Here are a couple of related currently open issues around the remote<>local as well:
#50
nebari-dev/nebi#423
A more intuitive remote<>local feel and a way to describe how software packs interact and depend on each other (a bigger discussion than this issue) would make the case for maintaining the dependency.
Best practices
User impact
Unresolved questions
Should Data Science Pack Require Nebi Pack Installed Separately?
Summary
The
data-science-packcurrently assumesnebi-packis installed. However, the "in-JupyterLab" Nebi works on its own. Furthermore, the local Nebi still supports connecting to OCI registries, so teams can still share "approved/curated" environments amongst themselves. The Remote<>Local functionality is extra that can be confusing.Note that it is technically possible to install
data-science-packwithout Nebi pack, and appears at first glance that the end user behavior is as expected. This RFD is more oriented around asking if the intent is that DS pack should require a separate Nebi pack install.User benefit
Users of the data-science-pack will not have to manage the
nebi-packas well. Though it provides additional features, it may not be necessary in all use cases so the implied dependency adds unnecessary administrative burden.Design Proposal
Since most of the design is already in place, the only modifications would be to update the data science pack helm chart a bit to remove some resources that are unnecessarily created when remote Nebi is not there. Here is a draft PR:
#192
Alternatives or approaches considered (if any)
Alternative is the status quo.. but related... Here are a couple of related currently open issues around the remote<>local as well:
#50
nebari-dev/nebi#423
A more intuitive remote<>local feel and a way to describe how software packs interact and depend on each other (a bigger discussion than this issue) would make the case for maintaining the dependency.
Best practices
User impact
Unresolved questions