I have been following the man, the myth, the legend Gregory John’s “How to Build an AI app with Bubble”. However, the method OpenAi uses to pass images created from prompts has changed since the video was recorded from URL that would remain for a few days, to a Base64 method.
Could someone please tell me what to do to save the base64 to the database and I will link the answer in the comments for future people.
Timecode relevant to trying to workout the backend workflow - 1:33:19
My core questions:
-What should the image api initialize call b64_json be set to - image/file/text?
I was reading your post. Mmm, I have some courses with Gregory too, and with many others.
Your set up, looks ok. But I want to share with you how I approach 90% of my WF regarding AI connections or services, for me it has been super helpful.
Normally, I added a 3 parthy tool - you don’t need it, strictly speaking- that smooth my WFs and integrated very easily. These two tools are: makecom and n8n.
I send the request and info with a simple webhook request, and take it back to bubble.
I like it. 100% sure you will handle that image “issue”, you are facing.
Note I have realised the reason it doesn’t return a URL is because I am using the gpt-image-1 model not dall-e-3 used in the video. At some point I must have seen it is cheaper to use gpt-image-1 and changed over. Gpt-image-1 does ONLY delivers in base64.
I dug a lot into this and paid someone to help but couldn’t find a way to save the base64 server side + associate to Post server side.
Let me know if you find out a way then people would be able to use the latest image model from openai.
