To use this program, data and file lists need to be prepared. If configured correctly, the directory tree should look like this: . ├── data │ └── datasets │ ├── cv │ │ └── cv.ex │ ├── tr │ │ ├── ...

The schemas are written as plain, readable JSON — not inferred from Python types — so you edit exactly what the model sees. The model never sees your Python. It reads a JSON description of your tools ...