Update app.py

app.py

@@ -32,7 +32,22 @@ logging.getLogger("httpx").setLevel(logging.WARNING)
 logging.getLogger("gradio").setLevel(logging.WARNING)
 # قفل عام لحماية الوصول إلى LLM/Agent داخل الخيوط
 llama_lock = threading.Lock()
-
+model_path = hf_hub_download(
+	repo_id="bartowski/Qwen2.5-14B-Instruct-GGUF",
+	filename="Qwen2.5-14B-Instruct-IQ4_XS.gguf",
+    )
+    
+    llm = LlamaCpp(
+    model_path=model_path,
+    n_ctx=10000,
+    n_threads=os.cpu_count(),
+    n_gpu_layers=0,
+    f16_kv=True,
+    temperature=0.7,
+    top_p=0.7,
+    max_tokens=500,
+    verbose=False
+    )
 # تحقق من وجود توكن في متغير البيئة
 HF_TOKEN = os.getenv("HF_TOKEN")
 
@@ -451,22 +466,7 @@ def gradio_interface():
    # class DummyLLM:
     #    def __call__(self, *args, **kwargs):
      #       return ""
-    model_path = hf_hub_download(
-	repo_id="bartowski/Qwen2.5-14B-Instruct-GGUF",
-	filename="Qwen2.5-14B-Instruct-IQ4_XS.gguf",
-    )
     
-    llm = LlamaCpp(
-    model_path=model_path,
-    n_ctx=10000,
-    n_threads=os.cpu_count(),
-    n_gpu_layers=0,
-    f16_kv=True,
-    temperature=0.7,
-    top_p=0.7,
-    max_tokens=500,
-    verbose=False
-    )
     tracer = None
     if LANGSMITH_AVAILABLE and os.getenv("LANGSMITH_API_KEY"):
         try: