How to provide user context information using ConversationalRetrievalChain

[levalencia]

I have a chatbot which searches for documents in Azure Cognitive Search, then gets the results into ChromaDB and then gets a response back to the user using OpenAI. It works perfectly fine.

However I have failed to provide user context to the chatbot. Basically I want to tell: User is called John Smith, He is 70 years old, He likes baking and gardening.

Why?
Because in my documents I have information about ages, names and interests of my users. SO if the user asks, Who likes the same as me? Then the bot should know that I like baking and gardening and then use those keywords to search. However I have failed miserably in all my attempts.

I tried writing in the prompt:
If user asks for his name, answer John Smith
If user asks for his age, answer 70.

I also tried adding hardcoded chat history.

I also tried using the messages array, but that didnt work either.

Part of my code:

    memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
    embeddings = OpenAIEmbeddings(deployment=OPENAI_EMBEDDING_DEPLOYMENT_NAME, model=OPENAI_EMBEDDING_MODEL_NAME, chunk_size=1)
    user_input = get_text()   
    retriever = AzureCognitiveSearchRetriever(content_key="content")

    
    llm = AzureChatOpenAI(
        openai_api_base=OPENAI_DEPLOYMENT_ENDPOINT,
        openai_api_version=OPENAI_API_VERSION ,
        deployment_name=OPENAI_DEPLOYMENT_NAME,
        openai_api_key=OPENAI_API_KEY,
        openai_api_type = OPENAI_API_TYPE ,
        model_name=OPENAI_MODEL_NAME,
        temperature=0)
    
    docs = get_relevant_documents(retriever, user_input)
    vectordb = Chroma.from_documents(documents=docs, embedding=embeddings, persist_directory='Store')

    if 'chain5' not in st.session_state:  

        general_system_template = """. Use the following pieces of context to answer the users question. 
        If users asks for a phone number just say you are not allowed to reveal personal information.
        If users asks for an age just say you are not allowed to reveal personal information.       
        If you cannot find the answer from the pieces of context, just say that you don't know, don't try to make up an answer.
        ----------------
        Context: 
        {context}"""
        general_user_template = "Question:```{question}```"

        #whatsyourname = "Tell me something about yourself?"
        #something = "I am John Smith, I have 70 years and I like painting."
        #howcanihelpyou = "How can I help you?"
        messages = [
                    SystemMessagePromptTemplate.from_template(general_system_template),
                    HumanMessagePromptTemplate.from_template(general_user_template)
        #            SystemMessagePromptTemplate.from_template(whatsyourname),
        #            HumanMessagePromptTemplate.from_template(something),
        #            SystemMessagePromptTemplate.from_template(howcanihelpyou)                   
        ]
        qa_prompt = ChatPromptTemplate.from_messages( messages )
    
        st.session_state['chain5'] = chain =ConversationalRetrievalChain.from_llm(
                llm=llm,
                memory=memory,
                retriever=vectordb.as_retriever(),
                chain_type="stuff",
                verbose=True,
                combine_docs_chain_kwargs={'prompt': qa_prompt}
            ) 
        
    if user_input:         
        output = st.session_state.chain5({"question": user_input, "chat_history": st.session_state.chat_history5})    
        st.session_state.past5.append(user_input)
        st.session_state.generated5.append(output["answer"])
        st.session_state.chat_history5.append((user_input, output["answer"]))