reducing chat window

utils.py

@@ -41,7 +41,7 @@ message_history = CustomMongoDBChatMessageHistory(
     collection_name='3d_printing_applications'
 )
 
-memory = ConversationBufferWindowMemory(memory_key="chat_history", k=10)
+memory = ConversationBufferWindowMemory(memory_key="chat_history", k=4)
 
 vectorstore_index = None
 
@@ -193,7 +193,7 @@ def get_qa_chain(vectorstore_index):
     # embeddings_filter = EmbeddingsFilter(embeddings=embeddings, similarity_threshold=0.76)
     # compression_retriever = ContextualCompressionRetriever(base_compressor=embeddings_filter, base_retriever=gpt_3_5_index.as_retriever())
     retriever = vectorstore_index.as_retriever(search_type="similarity_score_threshold",
-                                               search_kwargs={"score_threshold": .5})
+                                               search_kwargs={"score_threshold": .76})
 
     chain = ConversationalRetrievalChain.from_llm(llm, retriever, return_source_documents=True,
                                                   verbose=True, get_chat_history=get_chat_history,
@@ -211,7 +211,8 @@ def get_chat_history(inputs) -> str:
 def generate_answer(question) -> str:
     global vectorstore_index
     chain = get_qa_chain(vectorstore_index)
-    history = memory.chat_memory.messages
+    # get last 4 messages from chat history
+    history = memory.chat_memory.messages[-4:]
     result = chain(
         {"question": question, "chat_history": history})