ChatOpenAI with ConversationalRetrievalChain is unable to retain the prior interaction

Short Course Q&A LangChain: Chat with Your Data
1

Could somebody please assist me in fixing the code and letting me know where I am missing out? Despite the fact that the data from earlier chats is preserved in buffermemory, it is not taken into account when responding.

Code -

PDF File Import

from langchain.document_loaders.pdf import PyPDFLoader

loader = PyPDFLoader(“titanic.pdf”)
pages = loader.load()

Splitting

from langchain.text_splitter import RecursiveCharacterTextSplitter, CharacterTextSplitter

text_splitter = RecursiveCharacterTextSplitter(
chunk_size = 1500,
chunk_overlap = 150
)
data = text_splitter.split_documents(pages)

Embeddings

from langchain.embeddings.openai import OpenAIEmbeddings

embedding = OpenAIEmbeddings(api_key=api_key)

Vector Storage

from langchain.vectorstores import Chroma

persist_directory = ‘database/’

vectordb = Chroma.from_documents(
documents=pages,
embedding=embedding,
persist_directory=persist_directory
)

Create Model

from langchain.chat_models import ChatOpenAI

llm_model = ChatOpenAI(api_key = api_key, temperature=0.2, model=“gpt-3.5-turbo”)

Memory

from langchain.memory import ConversationBufferMemory

memory = ConversationBufferMemory(
memory_key=“chat_history”,
return_messages=True
)

Conversational Retrieval Chain

from langchain.chains import ConversationalRetrievalChain
retriever=vectordb.as_retriever()

qa = ConversationalRetrievalChain.from_llm(
llm=llm_model,
retriever=retriever,
memory=memory
)

question = “can you prepare 5 questions, make sure the questions can be answered using one word”
result1 = qa({“question”: question})
print(result1[‘answer’])

print(‘Next Question’)

question = “provide 4 options to each of the above questions”
result2 = qa({“question”: question})
print(result2[‘answer’])

Output -

Response1 -

  1. Who were the victims of the implosion?
  2. Will OceanGate face criminal charges?
  3. What caused the Titan to fall apart?
  4. How much did the Titan submersible search operation cost?
  5. How did the pressure at depth lead to implosion?

Response2

I’m sorry, I don’t have the capability to generate multiple-choice questions. If you have any other questions or need information, feel free to ask.

memory.buffer

[HumanMessage(content=‘can you prepare 5 questions, make sure the questions can be answered using one word’),
AIMessage(content=‘1. Who were the victims of the implosion?\n2. Will OceanGate face criminal charges?\n3. What caused the Titan to fall apart?\n4. How much did the Titan submersible search operation cost?\n5. How did the pressure at depth lead to implosion?’),
HumanMessage(content=‘provide 4 options to each of the above questions’),
AIMessage(content=“I’m sorry, I don’t have the capability to generate multiple-choice questions. If you have any other questions or need information, feel free to ask.”)]