📦 Model shelf

Put a trained model on the shelf and it becomes a prediction API in seconds, with its own documentation. Call it from Python, from a Streamlit app, or from an n8n workflow.

1. Train a scikit-learn pipeline (linear or logistic regression, random forest, gradient boosting, XGBoost, …).
2. Package it with pack(...) from the packaging notebook. You get a .zip.
3. Upload it below. You get an endpoint and a documentation page to test it.

Upload a model

On the shelf

Calling a model

import requests
r = requests.post("https://<this site>/m/<name>/predict",
                  json={"rows": [{"column": "value", "other_column": 3}]})
r.json()["predictions"]

Every model's documentation page shows its columns and a ready-made example: click Try it out. The API of the shelf itself is at /docs. Only packaged models are accepted (ONNX, XGBoost JSON): no pickles, no code.