Richard & Mark,
Thank you both for your suggestions.
I actually made some progress, as there were 2 different issues :
1- the syntax of the multimodal json curl must be
{"type" : "image_url", "image_url" : {"url":
"data:image/jpeg;base64,{_BASE64_}"}}
instead of
{"type" : "image_url", "image_url" : {"url": "data:image/jpeg;base64
{_BASE64_}"}}
I found the tip in an obscure blog post from early 2023.
Even the latest ChatGPT-4o3 missed the point.
2- I still have a problem with the base64 encoding of a jpeg file in LC.
Long story short, I am trying to send a curl request to a vision AI
model running with
LM studio. I tried several models that didn't give much details in the
error messages,
except for "bad request" or "data must be base64 encoded jpeg".
But the last one I tried returned :
{"error":"Input buffer has corrupt header: VipsJpeg: Corrupt JPEG data:
13 extraneous bytes before marker 0x8e\nVipsJpeg: Unsupported marker
type 0x8e"}
Here is my code for base64 encoding :
put URL ("file:" & "myfile.jpg") into timagedata
put base64encode(timagedata) into tBase64Image
replace return with "" in tBase64Image
Thanks,
jbv
_______________________________________________
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode
