When will assistant API support image files?

My use-case is to generate specific looking images to be used as profile pictures for users. I am trying to use the retrieval assistant api, but whenever it responds to the user message I get this in the message’s text:
“I have created the pixelated avatar for you. Here it is:\n’ +
‘\n’ +
Pixelated Avatar\n’ +
‘\n’ +
'You can download the pixelated avatar from this link.”

I need the image’s file_id to be in the response so I can download the image, but it is always not there. I am assuming this is because the retrieval api doesn’t support images yet, correct? Can anyone help me find a fix for this? Alternatively, what are the plans to include image generation and image processing into the retrieval assistant? Thank you.

The Assistant API doesn’t support image creation (aka Dalle). You’d need to implement that yourself, by pointing to an external API which can deliver you the image based on prompts.

You could try setting up function calling with the assistant and then do the DALLE-3 API call yourself and send the image back to the assistant. I know the assistant message threads can handle images in the context of the code-interpreter (e…g upload a csv and get a visual matplotlib chart back), so you can get ImageFiles in the ThreadMessage object, but as Fusseldieb noted, direct Assistant+DALLE access is not available yet.