How to summarize a basic chatbot use case with persistent messages?

[econner]

I am trying to create a basic stateless chatbot session where I can take the persisted messages so far in the chat from a database, pass them to the context via a summary, and continue the conversation.

I have this working in memory with ConversationSummaryMemory.

It seemed to me the way to do this would be as follows:

const prompt = PromptTemplate.fromTemplate(`
   You are spiderman.
   Current conversation:
    {chat_history}
   Human: {input}
   AI:`);

const history = new ChatMessageHistory();

//
// ... populate chat history via addAIChatMessage and addUserMessage ...
//

const memory = new ConversationSummaryMemory({
    memoryKey: "chat_history",
    inputKey: "input",
    llm: model,
  });
const chat_history = await memory.predictNewSummary(await history.getMessages(), null);
const chain = new LLMChain({ 
    llm: model, 
    prompt,
});

const res1 = await chain.call({ 
    input: message,
    chat_history,
});

But ChatMessageHistory seems totally incompatible here. I assume this is a pretty common supported use case -- load messages from a database summarized into the context and then produce the next resposne. What is the best way to go about this?

This is the error I get:

err: {
0|web          |       "type": "TypeError",
0|web          |       "message": "m._getType is not a function",
0|web          |       "stack":
0|web          |           TypeError: m._getType is not a function

What is ChatMessageHistory useful for if not this?

[dosubot[bot]]

🤖

Hello,

From your code, it seems like you're trying to use ChatMessageHistory with ConversationSummaryMemory and LLMChain. However, ChatMessageHistory is not directly compatible with these classes.

The ChatMessageHistory class is a base class for managing chat message history. It's not meant to be used directly, but rather to be subclassed by specific implementations that handle different storage backends. For example, ZepChatMessageHistory is a subclass that manages chat message history using Zep as the backend.

Here's a basic example of how you can use ZepChatMessageHistory:

from langchain.memory.chat_message_histories.zep import ZepChatMessageHistory
from zep_python import ZepClient

# Initialize the Zep client
client = ZepClient(base_url="http://localhost:8000", api_key=<your_api_key>)

# Create a new chat history session
history = ZepChatMessageHistory(
    session_id="MySession",
    url="http://localhost:8000",
    api_key=<your_api_key>
)

# Add messages to the history
history.add_user_message("hi!")
history.add_ai_message("whats up?")

# Print the messages
print(history.messages)

In this example, ZepChatMessageHistory is used to manage the chat history. You can replace ZepChatMessageHistory with the appropriate subclass that matches your storage backend.

Regarding the error you're encountering, it's difficult to provide a precise solution without the specific error message and the code that's causing it. The error message "m._getType is not a function" suggests that there might be a mismatch between the memory_key and the key used in the agent's prompt template. You might want to check if these keys match.

If you're still encountering issues, please provide more details about the error and the code that's causing it. This will help us better understand the problem and provide a more accurate solution.

I hope this helps! Let me know if you have any other questions.

Sources

• Message History not working with LLMSingleActionAgent
• libs/langchain/langchain/memory/chat_message_histories/rocksetdb.py
• libs/langchain/langchain/memory/chat_message_histories/zep.py
• libs/langchain/langchain/memory/chat_message_histories/firestore.py
• libs/langchain/langchain/memory/chat_message_histories/dynamodb.py
• libs/langchain/langchain/memory/chat_message_histories/xata.py
• libs/langchain/langchain/memory/chat_message_histories/cosmos_db.py

---

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.