HarisIqbal88/PlotNeuralNet ? reverse-engineered prompt

Build me a small toolkit for making neural network diagrams in LaTeX, with a simple Python way to describe layers and connections and then generate the TikZ and tex files automatically.

I want it to be easy to create common blocks like convolution, pooling, softmax, and arrows between them, so I can quickly draw model diagrams for papers and presentations. Please include a few working example scripts and the matching LaTeX output so I can see how to use it right away.

Make sure the diagrams actually compile cleanly with a helper script, and keep the code organized so it is easy to add new layer shapes later. If anything depends on current LaTeX or Python details, look up the latest docs online if you need to. Also, please include clear defaults and sensible spacing so the generated figures look polished without much manual tweaking.