Scaffolding

maldet scaffold generates a complete detector repository from a Jinja2 template. The output is a pip-installable Python package that satisfies the maldet protocol contracts out of the box.

Usage

maldet scaffold --template <rf|cnn> --name <detector_name> --out <directory>

Option	Default	Description
--template	rf	Template to use: rf or cnn
--name	(required)	Package and detector name
--out	.	Parent directory for the generated tree

Templates

rf — Random Forest baseline using SklearnTrainer. Generates: a pyproject.toml, maldet.toml, src/<name>/features.py, src/<name>/models.py, config.yaml, and Dockerfile.

cnn — PyTorch Lightning CNN using LightningTrainer. Generates the same structure with a LightningModule stub and Lightning-specific config.

After scaffolding

cd <out>
pip install -e .
maldet check      # validates maldet.toml
maldet describe   # prints manifest summary

Edit src/<name>/features.py to implement your feature extraction logic.