uv env installation

[pollytur]

• think from which python do we want to support things (if 3.9 too old?)
• clean up the dependencies
• remove requirements.txt and check (and fix if needed) the .toml installation
• Add uv.lock file and add it

[AbirajKangotra]

Hi! I’d like to work on this issue as part of exploring the Experanto project for GSoC 2026.

I can help clean up the dependencies, check the installation via 'pyproject.toml', remove 'requirements.txt' if appropriate, and add the 'uv.lock' file for reproducible environments.

Please let me know if this approach sounds good.

[reneburghardt]

Thanks for your interest @AbirajKangotra !
The switch from requirements.txt to pyproject.toml was already done in #105
Check and test python version, clean up dependencies (might be done already) and then finally add uv.lock is something we rather see with one of the main maintainers to make sure everything is tested properly and with (some maybe undocumented) objectives of this project already in mind.

[pollytur]

We probably need to wait a bit for uv since #81 is changing the base python version and a couple of other versions

[AbirajKangotra]

Hi @pollytur and @reneburghardt,

Thanks for the guidance!

I’ve started contributing to related areas and recently opened a PR improving documentation to better understand the workflow.

I’ll now focus on checking Python version compatibility and cleaning up dependencies as suggested.

Looking forward to contributing further!

[reneburghardt]

Hi @AbirajKangotra , seems like my message was not clear, but we would like to keep this issue for a maintainer with more experience and background knowledge regarding Experanto. So for your own sake of saving time and effort, please refrain from this issue as it otherwise might be a waste of your time.

[AbirajKangotra]

Hi @reneburghardt,

Thanks for the clarification, I appreciate it. I think I misunderstood your earlier comment.

I’ll explore other parts of the repo. If you have any suggestions on where I can contribute effectively, I’d like to know.

Thanks again!

[schewskone]

This is concerning PR #81 and just a note for the future. I think it would be nice to also add custom dependency groups for cpu and gpu supported versions (similar to odor-model repo).
Cuda is only used in GPU accelerated decoding in that PR so we could keep the environement lighter by making cuda kernels optional.