There's a more stable downstream known as Manim Community. One can use it, it has somewhat good documentation.
But even this leads to poor code, the API design and architecture decisions are very poor.
I don't like the hodgepodge produced for even the simplest animation.
And there is for Julia: https://github.com/JuliaAnimators/Javis.jl
My biggest problem with Manim is the long feedback loop as you (last time i used it) have to render the animation to view it. But it is a really valuable tool.
With p5.js you can get an instant feedback loop and have the animation frame number available with a little effort.
I coded this up and someone shared it here:
The Deep Learning papers one sees in NeurIPS, ICML, CVPR, etc. almost all have diagrams made in PP in them.
Using PP has made my life easier.
I even use Google Docs when I do not need animation.
I am a primary Linux user, and I keep at least one machine running Windows because of PP.
Also, as I said, the experience of writing code with Manim Community is not good.
If you have experience using prewritten libraries (i.e. you are not an academic or traditional Data Scienctist, and have proper programming experience), you will see that Manim is a bad tool. Not mincing words.
Yes, it takes time. But you get better and spee increases.