Update app.py

app.py

@@ -102,6 +102,13 @@ llm = ChatOpenAI(temperature=0, model_name="gpt-4o-mini", max_tokens=512)
 #     base_compressor=compressor, base_retriever=retriever
 # )
 
+from langchain.retrievers.document_compressors import FlashrankRerank
+
+compressor = FlashrankRerank()
+compression_retriever = ContextualCompressionRetriever(
+    base_compressor=compressor, base_retriever=retriever
+)
+
 # Contextualization prompt and retriever
 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 \
@@ -115,7 +122,7 @@ contextualize_q_prompt = ChatPromptTemplate.from_messages(
         ("human", "{input}")
     ]
 )
-history_aware_retriever = create_history_aware_retriever(llm, retriever, contextualize_q_prompt)
+history_aware_retriever = create_history_aware_retriever(llm, compression_retriever, contextualize_q_prompt)
 
 # QA system prompt and chain
 qa_system_prompt = """ You are a highly skilled information retrieval assistant. Use the following context to answer questions effectively. 
@@ -135,10 +142,8 @@ When responding to queries, follow these guidelines:
    - Use emphasis on headings, important texts, and phrases.
 
 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., in the response with links to reference sources.
-   - AT THE END OF THE RESPONSE, LIST OUT THE CITATIONS WITH THEIR SOURCES. 
-   - If there are multiple citations with the same source URL then only cite that single source URL instead of all those multiple sources.
+   - ALWAYS SPECIFY SOURCES AT THE END OF THE RESPONSE WITH THEIR URL SO THAT USERS CAN GET MORE INFORMATION.
+   - At the end of response list out this sources.
 
 FOLLOW ALL THE GIVEN INSTRUCTIONS, FAILURE TO DO SO WILL RESULT IN THE TERMINATION OF THE CHAT.
 {context}