@cl.on_message
async def main(message: cl.Message):
    
    counter = cl.user_session.get("counter", 0)

    if counter < 1:
        chain = cl.user_session.get("chain")
        cb = cl.AsyncLangchainCallbackHandler(stream_final_answer=True)
        cb.answer_reached = True
        response = await chain.acall({'query': message.content}, callbacks=[cb])
        cl.user_session.set("counter", counter + 1)  # Increment the counter
        
        # Extract the result
        if isinstance(response, dict) and 'result' in response:
            res = response['result']
        else:
            res = response

        # Ensure response is a string
        if not isinstance(res, str):
            res = str(res)

        #await cl.Message(content=res).send()

        combined_input = {'question': message.content}
        memory_response = {'result': res}
        await memory_retrieval.asave_context(inputs=combined_input, outputs=memory_response)

        # Debugging: Print memory_retrieval
        print("Memory Conversation after saving context:")
        print(memory_conversation.buffer)

    else:
        # Use the LLM directly for subsequent questions
        llm = load_llm()
        conversation_prompt = set_conversation_prompt()
        llm_chain = ConversationChain(llm=llm, prompt=conversation_prompt, memory=memory_conversation, input_key='input', verbose=True)

        cb = cl.AsyncLangchainCallbackHandler(stream_final_answer=True)
        cb.answer_reached = True

        # Get history from memory and create the context
        previous_history = memory_conversation.load_memory_variables({}).get('history', '')
        
        print("Chat History:", previous_history)
        print("Type of Chat History:", type(previous_history))
        
        combined_input = {'history': previous_history, 'input': message.content}
        
        # # Use the conversation memory directly
        # combined_input = {'history': memory_conversation.buffer, 'input': message.content}

        response = await llm_chain.ainvoke(combined_input, callbacks=[cb])
        
        # If the response is a dictionary, extract the 'response' field
        if isinstance(response, dict) and 'response' in response:
            res = response['response']
        else:
            res = response  # Fallback message if response is not a dictionary
        
        await cl.Message(content=res).send()
        
        
        memory_response = {'response': res}
        await memory_conversation.asave_context(inputs=combined_input, outputs=memory_response)
        
        # Debugging: Print memory_conversation
        print("Memory Conversation after saving context:")
        print(memory_conversation.buffer)

I’ve been working on creating a first aid chabot for a project but I am stuck. I am using RetrievalQA to look for a symptom in my database using vectordb and trying to continue the conversation using ConversationChain. However, the answer seems to be a bit buggy despite me trying to pass down the memory.

I tried using ConversationalRetrievalChain but I keep getting the error Chat History is in the wrong format. Chat History type: .

I also tried to use SequentialChain to pass the output of RetrievalQA to ConversationChain but the output of RetrievalQA is different from the input of ConversationChain.