fix(mamba)

[ddobrinskiy]

• resolve errors in mamba (is it relocatable?) #413 (certificate errors)
• rewrite tests to check the actual package functionality

[mxcl]

So why is it we’re using this pre packaged mamba forge thing rather than building our own? I'm sure there's good reason but the comment doesn't actually specify it and I think we should add it.

[mxcl]

Yeah so… 878 hard coded prefixes…

This thing is bundled with a lot of packages we provide ourselves. I'm not familiar enough with python AI stuff yet to understand if we can break it apart, but I'm investigating.

Though I will start with anaconda

[ddobrinskiy]

So why is it we’re using this pre packaged mamba forge thing rather than building our own? I'm sure there's good reason but the comment doesn't actually specify it and I think we should add it.

I'm not sure there's actually an easy way to do that, the official docs basically say "just use older mamba to build newer version from source"

So brew also uses the pre-packaged version

https://github.com/Homebrew/homebrew-cask/blob/c6e8fd99bdfd5ec217ff3b53505f42214f5cba58/Casks/mambaforge.rb

[ddobrinskiy]

This thing is bundled with a lot of packages we provide ourselves. I'm not familiar enough with python AI stuff yet to understand if we can break it apart

My intuition is that we should not break it apart, because the idea of mamba is to provide multiple isolated enviroments, like venv.

The same way you would not want your venv to contain a copy of all packages that is uses, not refer to already-installed instances elsewhere

[mxcl]

it's a brew cask, which means it's just a slight convenience wrapper around a pre-packaged thing. So it's not really packaged by brew in other words.

We can do better. I will tackle this sooooonish I hope. Got a lot of balls in the air RN

[mxcl]

I managed to pkg conda and micromamba.

I have begun trying to package this from first steps as well. Mostly I'm struggling with understanding their install instructions since they seem to assume you have some kind of conda env already setup.