mayooear / gpt4-pdf-chatbot-langchain

GPT4 & LangChain Chatbot for large PDF docs
https://www.youtube.com/watch?v=ih9PBGVVOO4
14.94k stars 3.02k forks source link

source output #441

Closed Mullet-bsd closed 8 months ago

Mullet-bsd commented 12 months ago

how do i change the source out put?

dosubot[bot] commented 12 months ago

🤖

Hello @Mullet-bsd! It's good to see you back. I hope you've been doing well.

To change the source output, you would need to modify the API endpoint '/api/chat' to return the desired output. This could involve changing the logic in the server-side code that handles this endpoint, or changing the endpoint itself to point to a different server-side function or external API.

The '/api/chat' endpoint is currently pointing to the server-side function named 'handler'. This function processes the request, checks if it's a POST request, and if a question is included in the request. It then sanitizes the question, creates a vector store from an existing index, retrieves documents, creates a chain, and invokes the chain with the sanitized question and chat history. The response is then sent back to the client.

Here is the function:

export default async function handler(
  req: NextApiRequest,
  res: NextApiResponse,
) {
  const { question, history } = req.body;

  //only accept post requests
  if (req.method !== 'POST') {
    res.status(405).json({ error: 'Method not allowed' });
    return;
  }

  if (!question) {
    return res.status(400).json({ message: 'No question in the request' });
  }
  // OpenAI recommends replacing newlines with spaces for best results
  const sanitizedQuestion = question.trim().replaceAll('\n', ' ');

  try {
    const index = pinecone.Index(PINECONE_INDEX_NAME);

    /* create vectorstore*/
    const vectorStore = await PineconeStore.fromExistingIndex(
      new OpenAIEmbeddings({}),
      {
        pineconeIndex: index,
        textKey: 'text',
        namespace: PINECONE_NAME_SPACE, //namespace comes from your config folder
      },
    );

    // Use a callback to get intermediate sources from the middle of the chain
    let resolveWithDocuments: (value: Document[]) => void;
    const documentPromise = new Promise<Document[]>((resolve) => {
      resolveWithDocuments = resolve;
    });
    const retriever = vectorStore.asRetriever({
      callbacks: [
        {
          handleRetrieverEnd(documents) {
            resolveWithDocuments(documents);
          },
        },
      ],
    });

    //create chain
    const chain = makeChain(retriever);
    const pastMessages = history
      .map((message: [string, string]) => {
        return [`Human: ${message[0]}`, `Assistant: ${message[1]}`].join('\n');
      })
      .join('\n');

    //Ask a question using chat history
    const response = await chain.invoke({
      question: sanitizedQuestion,
      chat_history: pastMessages,
    });

    const sourceDocuments = await documentPromise;

    res.status(200).json({ text: response, sourceDocuments });
  } catch (error: any) {
    res.status(500).json({ error: error.message || 'Something went wrong' });
  }
}

This function does not directly point to an external API but uses the 'pinecone' and 'OpenAIEmbeddings' libraries to create a vector store and retrieve documents. You can find this function in the pages/api/chat.ts file.

If you want to change how the source output is displayed in the chat interface, you would need to modify the code in the return statement of the Home function. This is where the chat interface is rendered, and it uses the state variables messages and sourceDocs to display the chat messages and source documents. You can find this function in the pages/index.tsx file.

I hope this helps! If you have any other questions, feel free to ask.

Sources


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.