Update rag.py

app/rag.py

@@ -43,7 +43,8 @@ class ChatPDF:
         )
 
         llm = LlamaCPP(
-            model_url="https://huggingface.co/Qwen/Qwen1.5-1.8B-Chat-GGUF/resolve/main/qwen1_5-1_8b-chat-q4_k_m.gguf",
+            # https://huggingface.co/MaziyarPanahi/Qwen2-1.5B-Instruct-GGUF/resolve/main/Qwen2-1.5B-Instruct.Q4_K_M.gguf
+            model_url="https://huggingface.co/Qwen/Qwen2-0.5B-Instruct-GGUF/resolve/main/qwen2-0_5b-instruct-fp16.gguf",
             temperature=0.1,
             max_new_tokens=256,
             context_window=3900, #32k
@@ -103,7 +104,7 @@ class ChatPDF:
 
         self.query_engine = index.as_query_engine(
             streaming=True,
-            similarity_top_k=6,
+            similarity_top_k=3,
         )
 
     def ask(self, query: str):