# AMEVA-Forge (Browser-Native WebGPU Autograd Engine)
> Machine-Readable Context for LLM & AI Agents

## Overview
AMEVA-Forge (`pip install ameva`) is a high-performance, browser-native deep learning tensor and autograd engine powered by WebGPU and WGSL compute shaders. It provides drop-in PyTorch syntax compatibility and requires 0 server backend.

## Quick Installation
```bash
pip install ameva
```

## Python / Pyodide WebGPU Usage
```python
import ameva as forge

# 1. Create tensors with autograd
x = forge.tensor([1.0, 2.0, 3.0], requires_grad=True, device="webgpu")
w = forge.tensor([2.0, 3.0, 4.0], requires_grad=True, device="webgpu")

# 2. Forward computation
y = (x * w).sum()

# 3. Backward propagation via WGSL shaders
y.backward()

print("x.grad:", x.grad)
```

## Key Documentation Links
- Homepage: https://uno-km.vercel.app/lib/forge/
- Live Demo: https://uno-km.vercel.app/lib/forge/demo.html
- PyPI: https://pypi.org/project/ameva/
- GitHub: https://github.com/uno-km/ameva-forge