Max 2 key attempts

app.py

@@ -50,7 +50,7 @@ def get_healthy_keys(max_failures=2):
             v["failures"] = 0
             v["healthy"] = True
         return OLLAMA_KEYS[:3]
-    return healthy[:3]  # Only try top 3
+    return healthy[:2]  # Only try top 3
 
 @app.post("/v1/chat/completions")
 async def chat(req: Request):