fix: ensure AMD_INFERENCE_URL has http protocol prefix

agents.py

@@ -187,6 +187,8 @@ async def _call_amd_vllm(
 
     # Candidate endpoints
     base_url = AMD_INFERENCE_URL.rstrip("/")
+    if not base_url.startswith("http"):
+        base_url = f"http://{base_url}"
     candidates = [
         f"{base_url}/proxy/8000/v1/chat/completions",
         f"{base_url}/proxy/8001/v1/chat/completions",

app.py

@@ -217,7 +217,10 @@ async def api_get_telemetry():
     error_msg = None
     
     # Try current proxy endpoint
-    url = f"{AMD_INFERENCE_URL.rstrip('/')}/v1/models"
+    base_url = AMD_INFERENCE_URL.rstrip('/')
+    if not base_url.startswith("http"):
+        base_url = f"http://{base_url}"
+    url = f"{base_url}/v1/models"
     headers = {}
     if AMD_INFERENCE_TOKEN:
         headers["Authorization"] = f"Bearer {AMD_INFERENCE_TOKEN}"