I'm not sure you can succeed quickly on this.
Did you try on pytensor forum ?
Or better nowodays, make an AI try to solve that question.
If I had to do it for my local use on an isolated planet with no AI, I would look into adding g++ external tool to a VisualStudio https://stackoverflow.com/questions/59923813/how-to-use-g-compiler-in-visual-studio

... a gordian node-like method would be to use Anaconda: can't you ?

so I asked ChatGPT the following:
"I initially used WinPython to run pyscripts. I switched to PyCharm because I was able to run the pyscripts in the Anaconda environment. Is there a way to run pyscripts in WinPython but somehow link it to Anaconda environment?"

Response:
✅ Two Working Solutions
🔹 Option 1: Use WinPython IDE (Spyder or WinPython Control Panel) with Conda's Python
Find your Anaconda env's Python path
For pymc-env, it's usually:

makefile
Copy
Edit
C:\Users<you>\anaconda3\envs\pymc-env\python.exe
Configure WinPython to use it
If you're using WinPython Control Panel:

Open the WinPython Control Panel

Click “Register/Unregister” → point it to the conda env’s Python.exe

Alternatively, open WinPython Command Prompt, then launch:

bash
Copy
Edit
your_script.py

🔹 Option 2: Add Your Anaconda Env to WinPython Jupyter Kernel (if using notebooks)
If you want to use Jupyter from WinPython, but execute in your conda env:

bash
Copy
Edit
conda activate pymc-env
python -m ipykernel install --user --name pymc-env --display-name "Python (pymc-env)"
Then open Jupyter in WinPython, and select the "Python (pymc-env)" kernel.

I tried Option 1 by opening WPPM and tried to target the pymc-env in Anaconda but that failed, The Register functionality, correct me if I'm wrong, just registers the WinPython to the Windows environment but not necessarily set to Anaconda's environment.

ok, LLM are still not always perfect, or too perfect for us.

Option 1 doesn't make sense to me
Option 2 is unlikely Windows territory: it starts per "bash", as Anaconda is using a sort of compatibility layer to Windows

What I don't understand is the constraint that force you to try that exotic inter-breeding of Python stacks.
With a clearer picture of the problematic, maybe a clearer path may appear.

You need a master degree in IT, or a paid OpenAI, imho, to explore that world

Most users/electrons/watts are stuck on linux/anaconda/nvidia for these things, it's a heavy leafting.