Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,35 @@
+---
+tags: [hayula, rushd, arabic, qwen, reasoning]
+---
+# Part of **Hayula Agent OS**
+> **15 Specialists. 1 M2 Ultra. Zero Cloud.**
+---
+license: cc-by-4.0
+language: [ar, en]
+library_name: transformers
+tags: [arabic, hayula, rushd]
+---
+# hayulalab/Rushd-Router-8B
+**Multi-Specialist Query Router**
+Routes queries to the appropriate Rushd specialist. 99% routing accuracy on held-out set.
+| Property | Value |
+|----------|-------|
+| Base Model | Qwen3-8B + LoRA (r=16) |
+| Training | 22 min on M2 Ultra, val loss 0.014 |
+| Performance | 99% routing accuracy |
+| Framework | MLX (Apple Silicon) |
+| License | CC-BY-4.0 |
+## Usage
+```python
+from mlx_lm import load, generate
+model, tokenizer = load("Qwen/Qwen3-8B", adapter_path="hayulalab/Rushd-Router-8B")
+response = generate(model, tokenizer, prompt="Your query here")
+```
+## Citation
+```bibtex
+@misc{hayula-rushd-router-8b,
+  author = {Hayula Lab},
+  title = {hayulalab/Rushd-Router-8B},
+  year = {2026},
+}
+```