How to send an image to GPT 4 vision?

hi All,

I just received word from Retool that the new GPT 4 models are now updated in retool. This includes GPT 4 vision.

I was wondering - how can I send an image, via retool, to ChaptGPT 4 vision?

I've uploaded my image to Retool storage, as a test. And in my chat prompt I am trying to somehow reference this imagem in a format that can be sent to GPT - but I'm not really getting anywhere.

Does anyone know how this might be possible?


Sorry - I think I was a bit premature, I've figured this out.... at least on Retools end (seems that my OpenAI account isn't setup for it - but I'll figure that out, I think with some pre-paid credits)

So, solution is - Retool has a separate AI feature added for GPT vision. See below:



Update: So far, I don't think a) the Retool API for vision is correct (needs to be the preview API) and b) I don't think Base64 data (I.e. Retool storage) can be passed (awaiting confirmation) as I think it might be too long.


Hey @maillme, we are able to use the gpt-4-visual model with base64 data coming from Retool Storage. Mind sharing your setup and error received?

Hi @joeBumbaca - thanks for the reply.
Actually - I sent my setup to @dworsky - you can see it below.

I tested the API manually (separate API resource) - and it seems to work fine, but when I attempt to use through the retool built-in model, I get an error:

And here is the API working "manually" using an img_url....

Got it, thanks. And to confirm these are both (RetoolAI action and REST) using your OpenAI key now. When you use your key via the API it works, but says you don't have access when using the built in model?

Does switching to the Retool key and using the built in model work?