Output values error with Agent and MongoDBChatMessageHistory

[JoseAlbDR]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

async getAssistantAnswer(document: string, question: string) {
    const { id } = await this.documentsService.findOne(document);

    const client = new MongoClient('mongodb://localhost:27017/chatbot');
    await client.connect();
    const collection = client.db('chatbot').collection('memory');

    const memory = new BufferMemory({
      returnMessages: true,
      memoryKey: 'chat_history',
      chatHistory: new MongoDBChatMessageHistory({
        collection,
        sessionId: id,
      }),
    });

    const prompt = ChatPromptTemplate.fromMessages([
      ['system', 'You are a helpful assistant'],
      new MessagesPlaceholder('chat_history'),
      ['human', '{input}'],
      new MessagesPlaceholder('agent_scratchpad'),
    ]);

    const vectorStore = this.vectorStoreService.createVectorStore(id);

    const retriever = vectorStore.asRetriever();

    const retrieverTool = createRetrieverTool(retriever, {
      name: id,
      description: `Tool to search information about ${document}`,
    });

    const tools = [retrieverTool];

    const agent = await createOpenAIFunctionsAgent({
      llm: this.model,
      tools,
      prompt,
    });

    const agentExecutor = new AgentExecutor({
      agent,
      tools,
      memory,
    });

    console.log(await memory.chatHistory.getMessages());

    const response = await agentExecutor.invoke({
      input: question,
    });

    return response;
  }

Description

I'm trying to add memory to an agent using BufferMemory and MongoDbChatMessageHistory, the messages are beeing stored correctly in the chatbot db under memory collection, but I`m getting the error:

Error: output values have 3 keys, you must specify an output key or pass only 1 key as output

I'm beeing using Lanchain js for a couple of days now but I'm new to the Agent concept, I implemented this with an ConversationChain with no problems, but I don't what to do with Agents, any help?

Sorry for my english, I tried my best ^^".

System Info

Platform: windows
Node version: 20.10.0
"@langchain/community": "^0.0.33",
"@langchain/core": "^0.1.36",
"@langchain/mongodb": "^0.0.1",
"@langchain/openai": "^0.0.14",

[JoseAlbDR]

I found a solution for my problem, here is the code for anyone that need it:

async getAssistantAnswer(document: string, question: string) {
    const { id } = await this.documentsService.findOne(document);

    const client = new MongoClient('mongodb://localhost:27017/chatbot');
    await client.connect();
    const collection = client.db('chatbot').collection('memory');

    const memory = new BufferMemory({
      returnMessages: true,
      memoryKey: 'chat_history',
      inputKey: 'input',
      outputKey: 'output',
      chatHistory: new MongoDBChatMessageHistory({
        collection,
        sessionId: id,
      }),
    });

    const prompt = ChatPromptTemplate.fromMessages([
      ['system', 'You are a helpful assistant'],
      new MessagesPlaceholder('chat_history'),
      ['human', '{input}'],
      new MessagesPlaceholder('agent_scratchpad'),
    ]);

    const vectorStore = this.vectorStoreService.createVectorStore(id);

    const retriever = vectorStore.asRetriever();

    const retrieverTool = createRetrieverTool(retriever, {
      name: id,
      description: `Tool to search information about ${document}`,
    });

    const tools = [retrieverTool];

    const agent = await createOpenAIFunctionsAgent({
      llm: this.model,
      tools,
      prompt,
    });

    const agentExecutor = new AgentExecutor({
      agent,
      tools,
      memory,
    });

    const chat_history = await memory.chatHistory.getMessages();

    const response = await agentExecutor.invoke({
      outputKey: 'output',
      input: question,
      chat_history,
    });

    return response.output;
  }

Now the agent is working with memory stored in a mongDB database:

{
response: {
  outputKey: 'output',
  input: 'Hello my name is Alberto',
  chat_history: [],
  output: 'Hello Alberto! How can I assist you today?'
}
}
{
response: {
  outputKey: 'output',
  input: 'Did I told you my name?',
  chat_history: [ [HumanMessage], [AIMessage] ],
  output: 'Yes, you mentioned your name when you first greeted me. How can I help you, Alberto?'
}
}
{
response: {
  outputKey: 'output',
  input: 'Whats my name?',
  chat_history: [ [HumanMessage], [AIMessage], [HumanMessage], [AIMessage] ],
  output: 'Your name is Alberto. How can I assist you today, Alberto?'
}
}