Update chatbot_rag.py

chatbot_rag.py

@@ -61,7 +61,7 @@ def build_qa():
         model=model,
         tokenizer=tokenizer,
         max_new_tokens=256,       # control length of response
-        temperature=0.5,          # more deterministic
+        temperature=0.2,          # more deterministic
         do_sample=False,          # no randomness (deterministic answers)
         top_p=0.9,                # nucleus sampling
         repetition_penalty=1.2,   # 🚀 reduce loops/repeats
@@ -69,8 +69,6 @@ def build_qa():
         return_full_text=False
     )
     
-    # Wrap into LangChain LLM
-    llm = HuggingFacePipeline(pipeline=pipe)
     
     # 🔹 Wrap in LangChain LLM
     llm = HuggingFacePipeline(pipeline=pipe)
@@ -82,15 +80,13 @@ def build_qa():
         input_variables=["context", "question"], 
         template=""" 
         Use the context to answer the question in short factual sentence.
-        Answer ONLY in one short, plain sentence.
+        Answer ONLY in vert particular ONE short, plain sentence.
         Do not explain, speculate, or roleplay. 
         Answer from the docs - Answer in plain natural language. - 
         Do not include code, imports, functions, or explanations of how to implement code. - 
         If you don't know, just say "I don't know." 
         
         
-    
-    
         Question: 
         {question}