Jeff Coggshall

## Summary This PR adds support for **HunyuanOCR** (`hunyuan_vl`), a specialized multimodal model designed for high-resolution Optical Character Recognition. The architecture combines a 27-layer ViT vision encoder with a 24-layer...

Add design patterns / architecture overview to docs

3

This is a more architecturally focused addition to the docs. Did it mainly for my own understanding but think it might be useful to others!