Update app.py

app.py

@@ -105,7 +105,8 @@ knowledge_base = KnowledgeBase()
 # repharser_llm = ChatNVIDIA(model="mistralai/mistral-7b-instruct-v0.3") | StrOutputParser()
 repharser_llm = ChatNVIDIA(model="microsoft/phi-3-mini-4k-instruct") | StrOutputParser()
 validation_llm = ChatNVIDIA(model="microsoft/phi-3-small-8k-instruct") | StrOutputParser()
-instruct_llm = ChatNVIDIA(model="mistralai/mixtral-8x22b-instruct-v0.1") | StrOutputParser()
+instruct_llm = ChatNVIDIA(model="mistralai/mistral-7b-instruct-v0.2") | StrOutputParser()
+#instruct_llm = ChatNVIDIA(model="mistralai/mixtral-8x22b-instruct-v0.1") | StrOutputParser()
 relevance_llm = ChatNVIDIA(model="nvidia/llama-3.1-nemotron-70b-instruct") | StrOutputParser()
 answer_llm = ChatOpenAI(
     model="gpt-4o",              
@@ -369,7 +370,7 @@ extract_validation_inputs = RunnableLambda(lambda x: {
 validation_chain = (
     extract_validation_inputs
     | relevance_prompt
-    | validation_llm
+    | instruct_llm
     | RunnableLambda(safe_json_parse)
 )
 
@@ -422,7 +423,7 @@ def RExtract(pydantic_class: Type[BaseModel], llm, prompt):
 
 knowledge_extractor = RExtract(
     pydantic_class=KnowledgeBase,
-    llm=validation_llm,            
+    llm=instruct_llm,            
     prompt=parser_prompt        
 )
 
@@ -469,14 +470,10 @@ def chat_interface(message, history):
         "vectorstore": vectorstore,
         "bm25_retriever": bm25_retriever,
     }
-
-    hybrid_result = hybrid_chain.invoke(inputs)
-    hybrid_result["validation"] = validation_chain.invoke(hybrid_result)
-    
     full_response = ""
 
     # Stream the response to user
-    for chunk in answer_chain.stream(hybrid_result):
+    for chunk in full_pipeline.stream(inputs):
         if isinstance(chunk, dict) and "answer" in chunk:
             full_response += chunk["answer"]
             yield full_response