Need help: Use OpenAI Function Calling with LangChain

[its-anas]

1. I want to utilize the new OpenAI Functions to enable my chatbot to retrieve data from an external API and use it in the response (in case the user requested that data and provided a specific ID).

2. After retrieving the data, I want the user to verify their email address that is contained within the data, so I can provide them with the answer.

I don't know how implement in my code! Can someone help please?

import { ChatOpenAI } from 'langchain/chat_models/openai';
import { ConversationalRetrievalQAChain } from 'langchain/chains';

const model = new ChatOpenAI({
  temperature: 0,
  modelName: 'gpt-3.5-turbo',
  callbacks: [],
});

const chain = ConversationalRetrievalQAChain.fromLLM(
  model,
  vectorstore.asRetriever(),
  {
    qaTemplate: QA_PROMPT,
    questionGeneratorTemplate: CONDENSE_PROMPT,
    returnSourceDocuments: false,
  },
);

const response = await chain.call({
  question: sanitizedQuestion,
  chat_history: history || [],
});

[Ehzuiq]

Hello, has the above requirement been implemented? I am also facing the same issue.

[brerdem]

Been interested in this as well. I don't know the exact answer but I'm pretty sure there should be an agent which wraps the CRQA chain which wraps the LLM. For what I know so far, there is no way to call a function without initiating an agent.

[mulib]

Hello, function calling method makes a lot of sense and looks very promising.
Unfortunately, i could not find JS/TS documentation that shows how to use it.
found many python examples, but couldn't find any example in JS/TS on the internet..
any plan to support function calling for JS/TS? Do you familiar with existing working example/docs?

[cbothra]

Hi Guys, did anybody got a solution?