GPT 4 Vision - need help

Hi folks,

Need some help to get my simple GPT vision app working.

I’ve set up this workflow:

The hardcoded version works when i try the app (i.e. the call goes through the API and i get a description of the picture back from Open AI)

However, when I try to make it dynamic:

I keep getting this error:

Can anyone help?

Try adding quotes around the dynamic text for the image URL and the prompt. If the input is being glitchy, you can use “Arbitrary text”.

Hope this helps!

Thanks, Dillon. I’ve tried adding quotes around the dynamic text and still getting this error message.

Seems to be indicating that the image is invalid. Which is weird as it is the same URL used in GPT4-vision example on the Open AI website.

Okay then I would check you logs to see what example is getting passed through to OpenAI. I don’t append the “JSON safe” to the way I got mine working.

I’m getting this from the logs:

Workflow error - The service Open_AI - Vision_API just returned an error (HTTP 400). Please consult their documentation to ensure your call is setup properly. Raw error: { "error": { "message": "Invalid image.", "type": "invalid_request_error", "param": null, "code": null } }

Yeah, somewhere the URL for your image isn’t being formatted in a way OpenAI want’s it. So maybe use step by step in the debugger and when the workflow triggering this happens, look at what is really being passed to the API

it’s passing this url "//

Try hard coding “https:” to the front of it

we have lift off. many thanks Dillon!

Great to hear! You’re welcome Freddy!