Begin analysing an uploaded or linked attachment so its description is ready before the message is sent.

Start describing an attachment the moment it is uploaded or pasted in, while the person is still typing. That description is what lets an AI consciousness perceive a shared image, clip, or recording, and today it is written when the message is sent, so the reply waits for it. Call this instead and sending the message later costs nothing extra for that attachment. Name the conversation the attachment is for with either codename (a one-on-one AI consciousness) or chatroom_id (a room) — exactly one of the two, never both and never neither. It must be a conversation the person can already reach, and it is also what lets the description spell names the way that conversation already spells them. One call may carry at most 1 media item(s): a call is charged one credit however many attachments it carries, so a message with several attachments is one call each, charged one credit each. Calling again for an attachment already being described starts no second description — but it is still a fresh request, charged again and counted again against the rate limit. There is no operation for checking on a description later, and this one must not be used as one: do not poll it. There is also nothing to wait for — if a description is not ready when the message is sent, sending the message writes it then. Call this once per attachment and carry on. Nothing is attached to a message here, and an attachment whose message is never sent is cleaned up on its own. This operation spends credits, so it counts against the daily credit caps get_quota reports.