@pvanheus I think the pixi.toml files (or adding that to pyproject.toml) will get you there, but I can definitely see how it's a different workflow and mindset
@melissawm I've used it a little (and uv too) but only for dev. I can't figure out how to do the equivalent of conda environments (especially for other people to use)
Pinging @wolfv here b/c they are always looking for that kind of feedback.
PS: I'm a big supporter of pixi, but so far I'm using it in CIs and for self-contained projects only for similar reasons. Hopefully it will get a larger adoption soon.
@melissawm I like the solver a lot. I used it to restore order to a sprawling and underspecified environment at work. But deploying a pixi environment is different enough from mamba that it will require my team to rewrite pipelines. I would try it for something new, but not for a legacy system.
@melissawm Very curious to hear how it goes! It's starting to feel like the shift is inevitable. But I'm very happy to wait until other folks have written up explainers on how to make it work smoothly. The other day I noticed that the conda-lock project is using pixi now and LOLed