mustatilcore

mustatilcore is a small PyPI-ready wrapper for a GUI-free mustatil_core.py backend.

Use this template when you want to publish the functional Mustatil backend as a Python package with a CLI command:

pip install mustatilcore
mustatilcore info

1. Put your backend file here

Copy your real backend into:

src/mustatilcore/mustatil_core.py

Replace the placeholder file already included in this template.

Your file should be GUI-free. Remove or guard anything that opens Tkinter, PySide, Qt, file dialogs, message boxes, or windows on import.

Good:

def detect_image(input_path, model=None, output_path=None, device="cpu", confidence=0.25):
    ...

Bad on import:

root = Tk()
root.mainloop()

2. Install locally for testing

From the project root:

py -m pip install -U pip
py -m pip install -e ".[dev]"
mustatilcore info
mustatilcore functions

3. Generic CLI usage

List exported functions from mustatil_core.py:

mustatilcore functions

Call any function by name:

mustatilcore call my_function --args-json "[1, 2]" --kwargs-json "{\"option\": true}"

Run an existing main() or cli_main() inside your backend:

mustatilcore run -- --your-backend-arg value

Detection convenience wrappers are included. They automatically look for likely function names in your backend.

mustatilcore detect-image image.jpg --model best.pt --out detections.json --device cpu --conf 0.5
mustatilcore detect-raster area.tif --model best.pt --out detections.gpkg --tile 1024 --overlap 160
mustatilcore train-yolo --data dataset.yaml --epochs 100 --device cuda

4. Build package

.\scripts\build_package.ps1

This creates:

dist/*.whl
dist/*.tar.gz

5. Upload to TestPyPI first

.\scripts\upload_testpypi.ps1

Install from TestPyPI:

py -m pip install --index-url https://test.pypi.org/simple/ --extra-index-url https://pypi.org/simple mustatilcore

6. Upload to real PyPI

Only after TestPyPI works:

.\scripts\upload_pypi.ps1

For token login with Twine:

username: __token__
password: pypi-...

Notes for Torch/CUDA

Do not add torch as a normal dependency unless you are sure. For CUDA builds, install the correct Torch build manually first. This prevents pip from replacing a working CUDA Torch installation with a CPU-only build.

Minimal Python API

from mustatilcore import load_core

core = load_core()
print(core)

You can also import your backend directly:

from mustatilcore import mustatil_core