Help with curl for GPT 4 Vision Macro (v11.0.1)

Questions & Suggestions
1

Help with curl for GPT 4 Vision Macro (v11.0.1)

Hi there, would appreciate getting any advice on getting the OpenAI vision endpoint to work with curl in KM terminal. I'm guessing there's some sort of escaping I need to do since I'm encoding the image as base64 but I just haven't been able to figure out how or what the true issue is.

Here's a link to the OpenAI documentation for Uploading a base 64 encoded image.

Thanks ahead of time for the help!

GPT 4 Vision.kmmacros (4.1 KB)

Keyboard Maestro Export
Keyboard Maestro Export1676×2180 282 KB

Hi there, would appreciate getting any advice on getting the OpenAI vision endpoint to work with curl in KM terminal. I'm guessing there's some sort of escaping I need to do since I'm encoding the image as base64 but I just haven't been able to figure out how or what the true issue is.

Here's a link to the OpenAI documentation for Uploading a base 64 encoded image.

Thanks ahead of time for the help!

2

Not able to be much help because I can't test this but:

Rather than a KM variable for the base 64 data (which might exceed the KM environment space), I'd use the OpenAI function to write the data to a file and access the data in the file. You'd just have to pass it the KM variable path to the screenshot. And no escaping required.

3

Hey thanks for the response! Trying to make sure I understand your suggestion.

Are you saying for me to save the base 64 string as a file, then load the file in the shell script with the path?

4

Yes, that's what the OpenAI documentation does. And it makes sense for Keyboard Maestro because there are issues with large variables overruning its environmental space.

5

I'll give it a go again tomorrow and see what I can come up with. In the OpenAI docs it's still sending a base 64 string, so I guess what you're saying is that there would be a difference in converting the image to base 64 at the time of the bash script running, rather than storing it in a variable then inserting into the curl script.

Thanks for taking the time to answer.

6

The difference is that writing the base 64 data to a file does not have the memory restriction in environment space that trying to store it in a Keyboard Maestro variable.