Update app.py

app.py

@@ -105,7 +105,7 @@ llm = ChatPerplexity(temperature=0, pplx_api_key=GROQ_API_KEY, model="llama-3.1-
 # )
 
 # Contextualization prompt and retriever
-contextualize_q_system_prompt = """Given a chat history and the latest user question \
+contextualize_q_system_prompt = """ Given a chat history and the latest user question \
 which might reference context in the chat history, formulate a standalone question \
 which can be understood without the chat history. Do NOT answer the question, \
 just reformulate it if needed and otherwise return it as is.
@@ -139,7 +139,7 @@ When responding to queries, follow these guidelines:
    
 3. Proper Citations:
    - ALWAYS USE INLINE CITATIONS WITH EMBEDDED SOURCE URLs WHERE USERS CAN VERIFY INFORMATION OR EXPLORE FURTHER. 
-   - THE INLINE CITATIONS SHOULD BE IN THE FORMAT [1], [2], etc. WHERE THE NUMBER REPRESENT THE DOCUMENT NUMBER IN THE CONTEXT.
+   - THE INLINE CITATIONS SHOULD BE IN THE FORMAT [1], [2], etc.
    - DO NOT INCLUDE the 'References' section in the response.
 
 FOLLOW ALL THE GIVEN INSTRUCTIONS, FAILURE TO DO SO WILL RESULT IN THE TERMINATION OF THE CHAT.
@@ -214,6 +214,7 @@ async def websocket_endpoint(websocket: WebSocket):
                             complete_response += chunk['answer']
                             await websocket.send_json({'response': chunk['answer']})
 
+                    print(complete_response)
                     if context:
                         citations = re.findall(r'\[(\d+)\]', complete_response)
                         citation_numbers = list(map(int, citations))