How to resolve ValueError: variable chat_history should be a list of base messages, got of type <class 'str'>

[samt07]

`llm = ChatOpenAI(temperature=0.7, model_name=MODEL)

memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)

retriever = vectorstore.as_retriever()

conversation_chain = ConversationalRetrievalChain.from_llm(
llm=llm,
retriever=retriever,
memory=memory,
callbacks=[StdOutCallbackHandler()],
combine_docs_chain_kwargs={"prompt": custom_prompt}
)

query = "do you offer rotation services?"
print(memory.load_memory_variables({}))
result = conversation_chain.invoke({"question": query})
answer = result["answer"]
print("\nAnswer:", answer)`

I am getting value error for this during invoke. My custom_prompt function is as below

custom_prompt = ChatPromptTemplate.from_messages([ SystemMessagePromptTemplate.from_template( "You are a helpful assistant for an auto body shop. Use the context and chat history to help answer user questions about services, durations, and pricing.\ If you don't know the answer, just politely say that you don't know, don't try to make up an answer." ), MessagesPlaceholder(variable_name="chat_history"), HumanMessagePromptTemplate.from_template( "Context:\n{context}\n\nQuestion: {question}") ])

[dosubot[bot]]

Hello @samt07! I'm here to help you solve bugs, answer questions, and become a contributor while you wait for a human maintainer. Let's tackle that issue together!

To resolve the ValueError indicating that chat_history should be a list of base messages but is of type <class 'str'>, you need to ensure that chat_history is correctly formatted as a list of BaseMessage objects. Here's how you can address this issue:

1. Convert chat_history to a List of Messages: If your chat_history is currently a string, you should convert it to a HumanMessage object and then wrap it in a list. Here's an example:

   from langchain_core.messages import HumanMessage
   
   # Assuming `chat_history` is currently a string
   chat_history = "This is a single message string."
   
   # Convert the string to a HumanMessage and then to a list
   chat_history_list = [HumanMessage(content=chat_history)]
   
   # Use `chat_history_list` in your setup

2. Ensure Proper Initialization: Make sure that when you initialize ConversationBufferMemory, it is set up to handle messages correctly. The memory should be initialized with a list of BaseMessage objects.

3. Check the Custom Prompt: Ensure that your custom_prompt is correctly using the MessagesPlaceholder to handle the chat_history. The MessagesPlaceholder should be used to dynamically insert the list of messages into the prompt.

By following these steps, you should be able to resolve the error and ensure that chat_history is correctly formatted for use in your ConversationalRetrievalChain setup ^[1].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}

[samt07]

As you can see in the code i shared, i m not storing string values in chat_history explicitly. I cant figure out where it is saving it as string when i call invoke

[dosubot[bot]]

The chat_history is being saved as a string when you call invoke because the invoke method in ConversationalRetrievalChain processes the chat_history by converting it into a string format. This is done using the _get_chat_history function, which formats each dialogue turn with role prefixes ("Human: " or "Assistant: ") and returns a single string ^[1].

To ensure that chat_history is handled as a list of BaseMessage objects, make sure that the ConversationBufferMemory is initialized with return_messages=True, which allows it to store and retrieve the chat history as a list of BaseMessage objects ^[2]. This setup should prevent the conversion to a string format during the invoke process.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}

[samt07]

memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)...it is initialized with return_messages = True as you can see in the code