Metadata-Version: 2.5
Name: decision-tune
Version: 0.0.4
Summary: Train your own local decision model on your own data, with an honest report card. Early development.
Project-URL: Homepage, https://github.com/abe238/decision-tune
Author: Abe Diaz
License-Expression: Apache-2.0
License-File: LICENSE
Classifier: Development Status :: 1 - Planning
Requires-Python: >=3.10
Description-Content-Type: text/markdown

# decision-tune

Train your own local decision model on your own data, with an honest report card.

**Status: early development (0.0.x). Not usable yet.** This release reserves the name and ships a CLI that only prints its status.

The plan: a local open-weight model drafts labels, you label a small answer key blind, decision-tune trains [Laya](https://github.com/NandhaKishorM/laya) on your Mac, and the model only leaves your machine through a report card that blocks known silent evaluation traps. Serving is through [Ollaya](https://github.com/ollaya-dev/ollaya).

```bash
pip install decision-tune   # or: npx decision-tune
decision-tune
```

**Use responsibly.** Models you train with decision-tune are only as good as your data and answer key. Don't use them for high-stakes decisions about people (hiring, credit, housing, health, legal) without independent validation and any review the law requires. Provided "as is" under Apache-2.0, without warranty.

Apache-2.0.
