Preserve Hugging Face visitor headers through the Gradio bridge

llama_ui.py

@@ -118,7 +118,7 @@ def install_ui(app):
         # Keep Gradio's OAuth, queue, visitor identity and ZeroGPU scheduling context.
         # Do not call the GPU-decorated function directly from a FastAPI route.
         headers = {k: v for k, v in request.headers.items()
-                   if k.lower() in ("cookie", "authorization") or k.lower().startswith("x-hf-")}
+                   if k.lower() in ("cookie", "authorization") or k.lower().startswith("x-")}
         client = httpx.AsyncClient(timeout=httpx.Timeout(90, connect=5), trust_env=False)
         try:
             submitted = await client.post("http://127.0.0.1:7860/gradio_api/call/chat",