My application has this vision-based feature. Currently, client side uploads the image to my server, and then my server send the image to OpenAI API as part of prompt.
I hope user can somehow directly send the image to OpenAI API, so the request sending would only take half of the time. I can also save my bandwidth money.
Any suggestions?
Welcome @Mengdi
In order to make calls to the OpenAI API, the request has to go through your server where the actual request is forwarded to the OpenAI API with the API key.
The image modality supports url uploads, which is what’s recommended in docs as well:
For long running conversations, we suggest passing images via URL’s instead of base64.
So you can give user the option to use url for images and then you won’t have to have them uploaded to your server, saving bandwidth costs. It should be noted that using url also has the downside where some websites/platforms deny having the image fetched which leads to an error.
2 Likes