createImageVariation gives unclear error when input image is not square

[KennyLindahl]

Describe the bug

Using the function createImageVariation with a non square image results in the following error:

(node:58341) UnhandledPromiseRejectionWarning: Error: Request failed with status code 400
    at createError (/Users/kenny.lindahl/Dev/test/open-ai-gpt/node_modules/axios/lib/core/createError.js:16:15)
    at settle (/Users/kenny.lindahl/Dev/test/open-ai-gpt/node_modules/axios/lib/core/settle.js:17:12)
    at IncomingMessage.handleStreamEnd (/Users/kenny.lindahl/Dev/test/open-ai-gpt/node_modules/axios/lib/adapters/http.js:322:11)
    at IncomingMessage.emit (events.js:387:35)
    at endReadableNT (internal/streams/readable.js:1317:12)
    at processTicksAndRejections (internal/process/task_queues.js:82:21)

Solution:

Option 1: The client should know that the image is not square and throw an error and not call the API.

Option 2: Alternatively it should change the image size (add margin, not stretch the image content) so it can be sent to the API with a successful response.

To Reproduce

Complete node program that reproduces the issue:

const { Configuration, OpenAIApi } = require("openai");
const fs = require("fs");

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

(async () => {
  const response = await openai.createImageVariation(
    fs.createReadStream(__dirname + "/images/non-square-image.png"),
    2,
    "1024x1024"
  );

  console.log("-------------------");
  console.log(response);
})();

Code snippets

No response

OS

macOS Monterey: 12.5.1 (21G83)

Node version

v14.17.3

Library version

3.1.0

[Ignema]

I had the same issue with the response always being that it must be a valid PNG file and less than 4MB, but it didn't mention that it has to be a square so I was confused for hours not knowing what the issue even was.

I managed to bypass the issue by using the sharp library:

 const buffer = await sharp(Buffer.from(image.data, "base64"))
  .png()
  .resize(400, 400, {
    fit: sharp.fit.inside,
    withoutEnlargement: true,
  })
  .toBuffer();
  buffer.name = "input.png"
  
  const response = await openai.createImageVariation(
    buffer,
  )
  
  // Consume the response the way you want
  // [...]

Both options mentioned by @KennyLindahl are good in my opinion.

[rattrayalex]

Thank you for the report! Our upcoming v4 of this library does a better job of highlighting 4xx error messages; give it a try: https://github.com/openai/openai-node/discussions/182