GPT building: Display images delivered via actions


Is it possible to return an image (e.g., PNG) using an action in a Custom GPT, and then the GPT somehow displays this image? This feature would be extremely helpful in many cases where data visualization is desired, but without overburdening GPT-4 to accomplish this through the code interpreter. Additionally, certain data can be presented much more effectively and appealingly to the user if it could be returned as an image by the action.

Yes, actions can return an image hosted somewhere by URL

when I try this I only get a link to an image, did anyone figure out if there’s a way to show images?

That can happen… In your instructions you can add that the GPT should display images inline with the text.
Other technique I have seen is to add a field to the JSON structure returned by the API endpoint that is a “NoteForAssistant”, and put it there.

oh nice, that worked! thank you!

(I’ve updated the system prompt)

1 Like