I haven't followed up. When I last looked, I found the problem space
is too large for one person (or project) to do it all. So the job is to
glue together lots of good OSS tools -- which is a very pythonic task.
The absolute requirement for Knowledge-Based-Engineering is an API which
allows a script to do anything a human can do. E.g.:
1. For 3D mechanical CAD, there is OpenCascade, with the pythonic
OpenCascade requires registration and is big download.
2. For 2D mechanical CAD, there is PythonCAD
The explicit intent to provide full scriptability (anything a human can
do via the GUI, a script can do via the API).
3. For EE schematics and simulation, there is OpenCollector and
specifically gEDA suite.
Not pythonic, but people have written glueware scripts in python to tie
the pieces together.
4. For fancy 3D objects and animations, Blender has the power and is
scriptable in python. It comes from the world of animations, but the
math doesn't care if you do EE 3D models instead.
5. We should all be concerned over SGI selling the OpenGL patents to
Microsoft, so at least look to Mesa, and perhaps to alternative 3D
6. I don't do GUIs much, but I understand form others that PyQT's
slot-and-signal architecture is well-respected, that *many* OSS projects
use PyGTK, and that folks who use wxPython are looking at "wax" as a
more pythonic layer. I finesse the whole issue by claimintg "GUIs are
for humans to do the work. I write code so computers can do the work."