In containers, tools like podman and docker are container engines. The idea of a container engine involves bespoke management of container images and the ability to seamlessly deploy and monitor containers, pods, deployments etc.

If we are aiming for a tool like ilab and the models it creates to be the "golden standard" for open source AI, our tool needs to be all encompassing.

A first step in this is making a few more "secondary" level commands under ilab, these should be something like:

config
model
data
taxonomy

This means we will now have commands that have 2 nouns followed by a verb: ilab model list for example.

Here are a few examples of commands that would exist in this new structure:

ilab config edit: edit ilab's configuration initialized when ilab first runs.
ilab model list: lists all models on the system, defaulting to models/
ilab model push: push model to a qualified registry.
ilab model top: gives memory usage information about a chat session, this could be applicable to server as well.

this organization of sub-commands under ilab will naturally uncover new ideas, commands, and use cases for ilab and hopefully incite a lot of community pull requests!

Here is the full structure of what I am proposing:

ilab
|
|_______model
|       |
|       |____quantize *
|       |____convert
|       |____download
|       |____train (--convert)
|       |____top *
|       |____serve   (-i)
|       |____chat
|       |____inference
|       |____push *
|       |____rm *
|       |____ls *
|       |____tag *
|       |
|_______data
|       |____generate
|       |____mix *
|       |
|_______config
|       |
|       |____create *
|       |____edit *
|       |____rm *
|       |
|_______tax
|       |
|       |____diff
|       |____check
|       |____download

where the starred commands are new. This issue is to track the initial conversion of the current structure to the new structure with the existing commands.