Update app.py

app.py

@@ -63,7 +63,8 @@ def initialize_qa_chain(temperature, max_tokens, top_k, vector_db):
         raise ValueError("Missing HF_TOKEN environment variable!")
     
     llm = HuggingFaceEndpoint(
-        repo_id="deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
+        # repo_id="deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
+        repo_id="Qwen/Qwen2.5-1.5B-Instruct",
         huggingfacehub_api_token=HF_TOKEN,  # Only needed if the model endpoint requires authentication
         temperature=temperature,
         max_new_tokens=max_tokens,
@@ -188,8 +189,10 @@ HF_TOKEN = os.getenv("AMAbot_r", "") # use for publishing
 if not HF_TOKEN:
     raise ValueError("Missing HF_TOKEN environment variable!")
 # Global InferenceClient for plain chat (fallback)
-client = InferenceClient("deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
-                        token=HF_TOKEN)
+client = InferenceClient(
+    # "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
+    "Qwen/Qwen2.5-1.5B-Instruct",
+    token=HF_TOKEN)
 
 # --- Auto-load vector database and initialize QA chain at startup ---
 try: