Anybody running uv version on arch linux willing to give some insight?

[UnconnectedBedna]

It says "Install the CUDA 12.4 Toolkit" in the readme.
extra/cuda-tools 12.8.1-1 is available, is it back compatible?
The link provided in the readme does not have anything available for arch based distros so that seems out of the question even if I wanted to install from outside the repos. (Unless I were to hack and use a deb package or smthn)
Nothing on AUR for 12.4 either, just 12.0.

Is there really no way to use the pyenv and pip instead of installing cuda from repo?

Or am I "stuck" on the legacy non uv version?
And if using legacy, do I still need cuda installed locally even though I activate the .venv and the setup script uses pip?

I noticed it installed things in $HOME/.local/bin and $HOME/.keras when I ran the gui the first time, witch I really dislike, I set up a pyenv, .venv and activate it for a reason (outside of specific python), but I guess that is what it is...

But any insight into this before I start installing gigabytes of stuff that might or might not work...

Thank you in advance.

[UnconnectedBedna]

This repo adds to your $PATH inside you shell configuration files (bash and zsh) AND adds a .profile in your home directory meaning it does this at login, NOT WHEN RUNNING THE APPLICATION!

There is absolutely no defense for overriding stuff by adding to path this way!

This meant stuff that was supposed to be ran via OTHER pyenvs I use on my system got overridden!
sd-webui-forge memory stuff for example. Suddenly an image that usually takes 2s to generate took +30s!
I spent HOURS trying to figure it out, restoring backups etc.
I saw the files as you can see in my previous post, but I could not for the life of me think a repo this big would edit my COMPLETE SYSTEM AND SHELL this way!
BE AWARE OF THIS!