This appears to implement nontrivial features such as constraint solving and the ability to easily fillet edges.

I like the idea of programming geometry, but never got the hang of the OpenSCAD language. I eventually settled on JSCAD.

I've found CadQuery to be a nice replacement for OpenSCAD. The viewer isn't as good, but using CadQuery + Jupyter Notebook is real nice. And it does fillets/chamfers, plus you can reference geometry. Bit of a learning curve though

There is also CQ-editor which is a graphical frontend, but I did not manage to get the official release to run on Ubuntu 22 as it threw some import errors.

Cadquery is great, I use it to parametrically create parts for electronics libraries (so a lot of similar looking geometries with slightly different dimensions).

My experience has been similar. By viewer, I meant the CQ-editor. I've also had issues getting it to run. IMO, it tries to do too much.

Since Jupyter-CadQuery [0] uses Juypter as an "editor", it's much more mature and stable. And if the viewer blows up, the notebook still works and can simply be reloaded. Debugging via replay is great.

It's also nice that you can install Jupyter-CadQuery on a different machine, either via Conda or Docker. And the standalone viewer looks awesome. It's a really exciting ecosystem.

[0] https://github.com/bernhard-42/jupyter-cadquery/

There's also implicitcad, that can use haskell, and does automatic fillets.

I have never used openscad, but from what I understand there would be 2 major differences. The first is the geometry engine. Openscad uses a mesh modeler, where cadseer uses boundary representation(opencascade). The second is the user interface. Openscad uses code/script to drive the model generation where cadseer uses user interaction to build a model generating dependency graph.

Cadseer is not better for stamping dies. That is just my background. I think cad is getting more into specialized industries now that the foundational stuff is more concrete. I don't think most people appreciate what it takes to develop a solid modeling kernel.