Update README.md

README.md

@@ -1,23 +1,10 @@
-# Exoskeleton Reasoning — Hugging Face Space
-
-A zero-setup Gradio demo for **Exoskeleton Reasoning**. It asks the model to return a compact JSON with slots:
-`question, evidence, claims, sources, final_answer` and renders a structured panel + the raw JSON.
-
-## Try locally
-```bash
-pip install -r requirements.txt
-export EXOSKELETON_MODEL_ID=Inpris/humains-junior   # or another compatible instruct model
-export HF_TOKEN=hf_xxx  # if the model is gated
-python app.py
-```
-
-## Space secrets / variables
-- **HF_TOKEN**: (Secret) your access token if the model is gated.
-- **EXOSKELETON_MODEL_ID**: (Variable) defaults to `Inpris/humains-junior`.
-- **DEVICE_MAP**: (Variable) defaults to `auto`. Set to `cuda` on GPU Spaces.
-- **MAX_NEW_TOKENS**: (Variable) default 512.
-- **TEMPERATURE**, **TOP_P**: sampling controls.
-
-## Notes
-- On free CPU Spaces, first token latency can be high. Consider enabling GPU or using a quantized checkpoint.
-- If JSON parsing fails (some models may add prose), the app falls back to showing the raw text as the final answer.
+---
+title: Exoskeleton Reasoning Demo
+emoji: 🦴
+colorFrom: indigo
+colorTo: blue
+sdk: gradio
+sdk_version: "4.44.0"
+app_file: app.py
+pinned: false
+---