initial README

README.md

@@ -1,10 +1,22 @@
 ---
-title: Path D Humanizer
-emoji: 🐢
-colorFrom: yellow
-colorTo: gray
+title: Path D Adversarial Humanizer
+emoji: 🧠
+colorFrom: green
+colorTo: blue
 sdk: docker
-pinned: false
+app_port: 7860
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Path D Adversarial Humanizer
+
+Adversarial paraphrasing inference Space for the humanizer-v3 spec
+(Path D). Paraphraser = Qwen3-4B (4-bit) + optional LoRA. Guidance
+detector = roberta-large-openai-detector. Per-token autoregressive
+loop with detector-guided argmin selection.
+
+## Endpoint
+
+`POST /generate` with payload `{prompt_text, system_prompt,
+max_new_tokens, top_p, top_k, temperature}`.
+
+Same contract as the original training Space serve.py.