Express 应用程序和 OpenAI API - createCompletion 方法不起作用

Question

Just experimenting with OpenAI's api and have a very basic express app up and running. What I'm trying to do is just get it to send me back an appropriate response with a basic input but it currently keeps failing.

I'm using Postman to iterate on the code on localhost. All packages are definitely installed and the API key is correct and specfied in the.env file.

My current working file is below. I'm sure I'll kick myself but can anyone spot what dumb thing I've probably done?

const express = require('express');
const app = express();
require('dotenv').config();
const bodyParser = require('body-parser');
app.use(bodyParser.json());
const axios = require('axios'); // Come back to this

const { Configuration, OpenAIApi } = require("openai");
const configuration = new Configuration({
    apiKey: process.env.OPENAI_API_KEY,
});

const openai = new OpenAIApi(configuration);

app.get('/api/v1', async (req, res) => {
    
  let body = {
      model: "text-davinci-003",        
      prompt: "How are you?",
      temperature: 1,
      max_tokens: 2086,
      top_p: 1,
      frequency_penalty: 0,
      presence_penalty: 0,
  };

  
  const response = await openai.createCompletion(body);

  res.send({ response });
});

// Listen for requests
app.listen(3000, function() {
    console.log('Server is listening on port 3000');
});

Error generated in terminal

/home/mint-pc/Desktop/projects/ebooks/api/node_modules/express/lib/response.js:1150
    : JSON.stringify(value);
           ^

TypeError: Converting circular structure to JSON
    --> starting at object with constructor 'ClientRequest'
    |     property 'socket' -> object with constructor 'TLSSocket'
    --- property '_httpMessage' closes the circle
    at JSON.stringify (<anonymous>)
    at stringify (/home/mint-pc/Desktop/projects/ebooks/api/node_modules/express/lib/response.js:1150:12)
    at ServerResponse.json (/home/mint-pc/Desktop/projects/ebooks/api/node_modules/express/lib/response.js:271:14)
    at ServerResponse.send (/home/mint-pc/Desktop/projects/ebooks/api/node_modules/express/lib/response.js:162:21)
    at /home/mint-pc/Desktop/projects/ebooks/api/ghost_writer.js:48:7

Answer 1

STEP 1: Run Express server

Run nodemon test.js in the terminal.

test.js

const express = require('express');
const app = express();

const bodyParser = require('body-parser');

app.use(bodyParser.json());

const { OpenAIApi } = require("openai");

const openai = new OpenAIApi();

app.post('https://api.openai.com/v1/completions', async (req, res) => {
    try {
        const request_options = {
            method: 'POST',
            headers: {
                'Content-Type': 'application/json',
                'Authorization': `Bearer xxxxx`
            },
            body: {
                model: req.body.model,
                prompt: req.body.prompt,
                max_tokens: 7,
                temperature: 0
            }
        }

        openai.createCompletion(request_options)
            .then((response) => {
                res.send({ response });
            })
            .catch(err => {
                console.log(err.message);
            })
    } catch (error) {
        console.log(error);
    }
});

app.listen(3000, function() {
    console.log('Server is listening on port 3000');
});

STEP 2: Provide model and prompt parameters in request body

{
  "model": "text-davinci-003",
  "prompt": "Say this is a test"
}

Screenshot:

Also, don't forget to provide your API key!

Screenshot:

STEP 3: Click "Send" to make an HTTP POST request

You should get 200 OK and the following response:

{
  "id": "cmpl-6ZNYkREx9KEThpq2keFD9E1Xic3q6",
  "object": "text_completion",
  "created": 1673890062,
  "model": "text-davinci-003",
  "choices": [
    {
      "text": "\n\nThis is indeed a test.",
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 5,
    "completion_tokens": 8,
    "total_tokens": 13
  }
}

Answer 2

Answer provided in this comment by @Boaz: "Inspect the response object. It's likely a complete HTTP response object and not just the response data..."

It was that.