I published our version to the snapcraft store at https://snapcraft.io/deepkit.
Deepkit is a GUI and CLI app to manage, execute, and track machine learning experiments. Machine learning experiments in a professional environment are usually done in Docker because of reproducibility and ease-of-use for GPUs (nvidia/cuda libraries for example).
Deepkit supports executing experiments on the host directly (using the shell) and in Docker. Docker is totally optional, so I’m fine with a non-auto-connected plug (if I understand that correctly). If the plug is non-auto-connected I can display an warning/dialog with introduction how the user can manually grant docker access to the deepkit snap.
I’m not sure if the process is correct. The latest deepkit version is currently in review-state “Manual review requested” since I added “docker” interface to the build now. Let me know if you need anything else or if I need to move that post to somewhere else.