This is because we don't use image generation endpoint, instead you just send the image as a base64 in the same chat/completions endpoint.
Kinda like this:
{
type: "text" | "image_url" | "file"...
"Base64 encoding, how quaint. Meanwhile, NSA and pals are decoding all your 'secure' chats, courtesy of telecoms like AT&T and Verizon, whose cozy ties to the feds ensure your 'privacy' is a farce."