ScreamingOpossum/openai_interaction/image_recognition_test.py

from openai import OpenAI

client = OpenAI()
SUPPORTED_IMAGE_EXTENSIONS = ('.png', '.jpg', '.jpeg', '.webp', '.gif')


async def image_recognition(content: str, files: list[dict[str: str]]) -> str:
    input_data = [{'type': 'input_text', 'text': content}]
    input_data.extend(
        [
            {
                'type': 'input_image',
                'image_url': i_url.get('url'),
            }
            for i_url in files if i_url.get('filename').lower().endswith(SUPPORTED_IMAGE_EXTENSIONS)
        ]
    )
    response = client.responses.create(
        model='gpt-4.1-mini',
        input=[{
            'role': 'user',
            'content': input_data,
        }],
    )
    return response.output_text


if __name__ == '__main__':
    data = {
        'content': 'What\'s on these images?',
        'files': [
            {
                'filename': 'Вы_продоёте_рыбов.jpg',
                'url': 'https://upload.wikimedia.org/wikipedia/ru/8/8a/%D0%92%D1%8B_%D0%BF%D1%80%D0%BE%D0%B4%D0%BE%D1%91%D1%82%D0%B5_%D1%80%D1%8B%D0%B1%D0%BE%D0%B2.jpg',
            },
            {
                'filename': 'Putin_and_Macron_meeting_with_a_large_table.jpg',
                'url': 'https://upload.wikimedia.org/wikipedia/commons/a/a9/Putin_and_Macron_meeting_with_a_large_table.jpg',
            },
            {
                'filename': 'shitty_pdf_file.pdf',
                'url': 'https://example.com/shitty_pdf_file.pdf'
            }
        ],
    }
    print(image_recognition(**data))