About the only way is if you hosted the images on a web server, and if not just a preset list of links, perhaps had an API action to give the URLs so the AI could produce them as markdown in its response.
Only user uploads can be given as vision inputs.