diff --git a/scripts/gpu-self-heal.py b/scripts/gpu-self-heal.py
index 3229ca7..9598463 100755
--- a/scripts/gpu-self-heal.py
+++ b/scripts/gpu-self-heal.py
@@ -141,6 +141,7 @@ def evaluate_rules(data, history):
                 host_key = k
                 break
         if not host_key:
+     print(f\"[warn] no GPU_HOSTS mapping for '{name}' — remote rules (incl. Rule 7) skipped for this GPU\")
             host_key = name.replace(" ", "-").lower()
         
         temp = gpu.get("temp_c", 0)
@@ -384,7 +385,7 @@ def run_inference_test(model="syslog-auto"):
         req = Request("http://192.168.68.116:4000/v1/chat/completions",
                       data=body, headers={
                           "Content-Type": "application/json",
-                          "Authorization": "Bearer sk-litellm-7f96080dd99b15c36bd4b333b58a6796"
+                          "Authorization": "Bearer " + os.environ.get("LITELLM_API_KEY", "")
                       })
         with urlopen(req, timeout=30) as resp:
             return resp.status == 200