Metadata-Version: 2.4
Name: triton-sleuth
Version: 0.0.2
Summary: A diagnostic toolkit for Triton GPU kernel development
Project-URL: Repository, https://github.com/biobenkj/triton-sleuth
Author-email: "Benjamin K. Johnson" <biobenkj@gmail.com>
License-Expression: MIT
License-File: LICENSE
Classifier: Development Status :: 1 - Planning
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3
Classifier: Topic :: Software Development :: Debuggers
Classifier: Topic :: Software Development :: Testing
Requires-Python: >=3.10
Description-Content-Type: text/markdown

# Triton-Sleuth

A diagnostic toolkit for [Triton](https://github.com/triton-lang/triton) GPU kernel development.

Sleuth investigates your kernels so you don't have to.

> **Status:** Pre-release. Under active development.

## What it will do

- **`sleuth check`** — Correctness testing against a PyTorch reference with auto-generated edge cases
- **`sleuth bench`** — Benchmarking vs PyTorch and torch.compile with zero boilerplate
- **`sleuth diagnose`** — Pattern-match known Triton failure modes and emit structured diagnostics

## Design principles

- **Minimal API surface** — Opinionated defaults, not a wall of config options
- **Structured output** — JSON diagnostics built for AI coding agent integration
- **Diagnostic intelligence** — Encodes hard-won debugging patterns so you don't repeat them

## Installation

```bash
pip install triton-sleuth
```

## License

MIT
