updated infrance to remove thinking token

inference.py

@@ -29,6 +29,8 @@ from models import (
     StoreObservation,
     dump_model,
 )
+from dotenv import load_dotenv
+load_dotenv()
 
 API_BASE_URL = os.getenv("API_BASE_URL", "https://router.huggingface.co/v1")
 MODEL_NAME = os.getenv("MODEL_NAME", "Qwen/Qwen3.5-397B-A17B")
@@ -269,11 +271,18 @@ async def get_model_action(
             {"role": "user", "content": user_msg},
         ],
         temperature=0.3,
-        max_tokens=1024,
+        max_tokens=4096,
     )
 
     content = response.choices[0].message.content or "{}"
     content = content.strip()
+
+    print(f"[DEBUG] Raw LLM response (first 500 chars): {content[:500]}", flush=True)
+
+    think_end = content.find("</think>")
+    if think_end != -1:
+        content = content[think_end + len("</think>"):].strip()
+
     if content.startswith("```"):
         content = content.split("\n", 1)[-1]
         if content.endswith("```"):