Thanks for making this! It looks like exactly what I want!
Pixi
Pixi is very easy to install.
Then the/a pixi.toml file could be tracked in this repo.
You could probably then do away with the separate requirements.txt and Docker builds.
A Pixi setup "task" could copy the example config to one for the user (and more).
Pixi also can store it's stuff inside pyproject.toml files and use those for installation.
I've used Python a lot over the years (myself, and transitioning user's code to "production") and realized it's fundamental design apparently makes it nearly impossible to enable a clean, all dependencies, installation. Sooo many third party systems (like Poetry) have tried to mitigate this.
The closest one can come (before full containerization) seems to be Conda. I've created top-level dep environment.yaml files to make things very easy to install (E.g. using micromamba). But conda doesn't let me get to the "last mile" of finishing an installation setup. Pixi seems to have done what I've been waiting a long time for.
Other languages, like Julia, have already been doing this sort of thing: Pkg.add(some_package) will automatically add the package to a top-level project file, which anyone can then take and use to install exactly the same things.
Pixi seems to have done a good job of helping Python catch up a bit.
Thanks for making this! It looks like exactly what I want!
Pixi
Pixi is very easy to install.
Then the/a pixi.toml file could be tracked in this repo.
You could probably then do away with the separate requirements.txt and Docker builds.
A Pixi setup "task" could copy the example config to one for the user (and more).
Pixi also can store it's stuff inside pyproject.toml files and use those for installation.
I've used Python a lot over the years (myself, and transitioning user's code to "production") and realized it's fundamental design apparently makes it nearly impossible to enable a clean, all dependencies, installation. Sooo many third party systems (like Poetry) have tried to mitigate this.
The closest one can come (before full containerization) seems to be Conda. I've created top-level dep environment.yaml files to make things very easy to install (E.g. using micromamba). But conda doesn't let me get to the "last mile" of finishing an installation setup. Pixi seems to have done what I've been waiting a long time for.
Other languages, like Julia, have already been doing this sort of thing: Pkg.add(some_package) will automatically add the package to a top-level project file, which anyone can then take and use to install exactly the same things.
Pixi seems to have done a good job of helping Python catch up a bit.