Update app.py

app.py

@@ -47,15 +47,19 @@ def calculate_embedding_cost(text):
     return total_tokens, total_tokens / 1000 * 0.0004
 
 
+# Define the metadata for filtering
+metadata = {'source': 'https://mimtsstac.org/sites/default/files/session-documents/Intensifying%20Literacy%20Instruction%20-%20Essential%20Practices%20%28NATIONAL%29.pdf'}
+
+metadata_filter = {"source": {"$eq": metadata['source']}}
+
 def ask_with_memory(vector_store, query, chat_history=[]):
     from langchain.chains import ConversationalRetrievalChain
     from langchain.chat_models import ChatOpenAI
     
     llm = ChatOpenAI(model_name='gpt-3.5-turbo', temperature=1, openai_api_key=openai_api_key)
-    
     # The retriever is created with metadata filter directly in search_kwargs
-    retriever = vector_store.as_retriever(search_type='similarity', search_kwargs={'k': 3, 'filter': {'source':{'$eq': 'https://mimtsstac.org/sites/default/files/session-documents/Intensifying%20Literacy%20Instruction%20-%20Essential%20Practices%20%28NATIONAL%29.pdf'}}})
-    
+    retriever = vector_store.as_retriever(search_type='similarity', search_kwargs={'k': 3, 'metadata': metadata_filter})
+
     chain= ConversationalRetrievalChain.from_llm(llm, retriever)
     result = chain({'question': query, 'chat_history': st.session_state['history']})
     # Append to chat history as a dictionary