CODEWITHSUNDEEP
javascriptnode.jsocropenai-apichatgpt-api
Pipo
Pipo

Reputation: 5611

With node javascript : How use OpenAI api as OCR for a local image?

I try to use openApi's api as OCR in node with gpt-4o model from a local image .

const api_key = "mykey"
import OpenAI from 'openai';
import fs from "fs"

const openai = new OpenAI({
    apiKey: api_key, 
});

function convertToBase64(filePath) {
    const bitmap = fs.readFileSync(filePath);
    return Buffer.from(bitmap).toString('base64');
}

(async () => {
    const chatCompletion = await openai.chat.completions.create({
        model: "gpt-4o",
        messages: [
            {
                content: [
                    {
                        role: 'user',
                        content: 'Return me the text of this image.'
                    },
                    {
                        role: 'system',
                        content: 'image_url',
                        image_url: `data:image/png;base64${convertToBase64('./image12.png')}`
                    }
                ]
            }

        ]
    });

    console.log(chatCompletion.choices[0].message.content);

})();

I have the error message

(node:23388) [DEP0040] DeprecationWarning: The punycode module is deprecated. Please use a userland alternative instead. (Use node --trace-deprecation ... to show where the warning was created) url-state-machine.js:2 Uncaught BadRequestError Error: 400 Invalid chat format. Expected 'role' fields in all messages. at generate (path\node_modules\openai\error.mjs:41:20) at makeStatusError (path\node_modules\openai\core.mjs:268:25) at makeRequest (path\node_modules\openai\core.mjs:311:30) at processTicksAndRejections (internal/process/task_queues:95:5) error.mjs:41 Process exited with code 1

Do you know how to do it?

Upvotes: 1

Views: 627

Answers (1)

maximitt
maximitt

Reputation: 3

The role property should be located on the same level as the content:

    messages: [
      {
        role: 'user',
        content: [
          {
            content: 'Return me the text of this image.'
          },
          {
            content: 'image_url',
            image_url: `data:image/png;base64${convertToBase64('./image12.png')}`
          }
        ]
      }
    ]

Please note that your code also has other issues, as @ITgoldman pointed out.

Upvotes: 0

Related Questions