Use when you cannot inspect media directly and the user provides images, screenshots, image-heavy PDFs, or video frames. Send each image blob to the configured multimodal alternate, read its response, then continue using that response as context.
Langue du texte source : anglais