Hey guys, my first post!
Can a custom GPT access and display a pix from a file (word/rich text format) my knowledge base? If so, whats the best tips for instructing to do so?
Thanks!
Nope, it can’t, as per the explicit answer my GPTs gave me after I’ve uploaded docs with images in the knowledge base.
This would be nice to have as a feature added.
You need to host the images somewhere and then have an action that returns the image URL and instructs the GPT to display it inline.
I think if you list the names of images in a text document and then add a diplayImage[name] action that just returns the URL in some JSON structure like
{
"image": "https://myimage",
"notesForAssistant": "display the image inline"
}
…that would work.
Gosh, it looks like too much hassle…Thank you everyone
Hi,
I realise I’m a bit late to the party, but there is an another way without needing actions to add pictures to responses from custom GPTs. I’ve done this in an experimental GPT - ChatGPT - Ulbinghof Assistant
Here’s the text that I’ve added to instructions of the custom GPT:
When specifically asked for a picture of Ulbinghof, the Assistant will show the image named from the url https://www.ikangai.com/wp-content/uploads/2024/01/Lumiere.jpg
By asking “show me a picture of Ulbinghof” you’ll get a picture in your response. (It’s actually not the picture of Ulbinghof and just for illustrative purposes).
It’s a bit of an hack and you have to host the image somewhere publicly, but it works.