New version :

Alright, no sugarcoating — this thing takes in your neural network JSON, runs it through a Node.js backend, cooks it with a Python-LaTeX script, and spits out a clean PNG diagram. Minimal effort, max clarity.

• You send a JSON with your model layers.
• Backend runs a Python script using PlotNeuralNet (built on LaTeX/TikZ).
• It gives you back a crispy PNG diagram. Use it wherever you want.

• Windows 10/11 (Mac/Linux work too, but commands vary)
• Python 3.8+ (works well with 3.9)
• Node.js 14+
• LaTeX (MiKTeX or TeX Live — must include pdflatex)
• ImageMagick (with magick command and PDF support)
• Git (optional, if you're cloning)

Using git:

git clone https://github.com/guider23/PlotNeuron
cd PlotNeuron

Or unzip the folder and cd into it.

No pip installs. Pure standard library — simple and clean.

npm install

• Install MiKTeX or TeX Live.
• Ensure pdflatex is accessible from terminal:

pdflatex --version

• Install from here
• During install, check:
  • "Install legacy utilities (e.g., convert)"
  • "Install development headers for C/C++"
• Confirm it works:

magick --version

• Still no PNG? Install Ghostscript and add to PATH for PDF support.

node server.js

Send a JSON to http://localhost:3000/generate. Example:

{
  "layers": [
    { "type": "Conv", "filters": 64, "position": "(0,0,0)" },
    { "type": "Pool", "position": "(2,0,0)" },
    { "type": "Dense", "units": 128 }
  ]
}

Use Postman, curl, whatever you vibe with.

Using curl:

curl -X POST http://localhost:3000/generate -H "Content-Type: application/json" -d "{\"layers\":[{\"type\":\"Conv\",\"filters\":64,\"position\":\"(0,0,0)\"},{\"type\":\"Pool\",\"position\":\"(2,0,0)\"},{\"type\":\"Dense\",\"units\":128}]}"

Open:

http://localhost:3000/outputs/main.png

Boom. There’s your architecture.

• Dynamic JSON Handling — No static files. Backend takes your live JSON and builds LaTeX on the fly.
• Absolute Paths — No “where’s my file” errors. Full paths passed to Python.
• Clear Errors — Any LaTeX/ImageMagick fails? The server yells clearly, not cryptically.

• FileNotFoundError — Check your paths and installed stuff.
• LaTeX/ImageMagick not working — Confirm they're in your PATH.
• No PNG Output — Server logs are your map.

• Want new layer types? Hack scripts/generate.py, inside generate_arch_from_json().
• Wanna go big? Check out the OG repo: PlotNeuralNet

That’s it. Clean project, no clutter. You know the vibe. Just clone → install → run → post → PNG. If anything's off, read the console — it talks back.