feat: stream_true and new stop_word

rag_pipeline/llm_interface.py

@@ -9,10 +9,11 @@ class LocalLLM:
 
     def __init__(self, model_path: str,
                  n_ctx: int = 2048,
-                 stop=["</s>", "User Question:"],
+                 stop=["</s>", "User Question:", " User Response:"],
                  max_tokens=500,
                  temperature: float = 0.5,
-                 repetition_penalty=1.25):
+                 repetition_penalty=1.25,
+                 stream=True):
         """
         Initialize the local LLM model via llama-cpp and Langchain.
 
@@ -27,6 +28,7 @@ class LocalLLM:
             max_tokens=max_tokens,
             temperature=temperature,
             repetition_penalty=repetition_penalty,
+            stream=stream,
             verbose=False
         )