Image uploads vs. file uploads

My understanding is there is no difference in the way data is stored between files and images.

I get that image uploader component allows for camera or camera roll access.

If that is correct, then when uploading a PDF of a document or a photo of the PDF it sounds like it wouldn’t matter what is uploaded.

If a file has multiple pages, how does that work (so not multiple files, just 1 file with 2 or more pages) in terms of what is displayed?

Upload a PDF or an image, can be multi-page PDF or multiple images. Then use OpenAI integration to read data and return result.

Files will be renewal notices and/or invoices. I want to speed up the process of entering data from the files/images and saving the correct data to the correct field, i.e. due date, supplier, amount.

Because all docs will be formatted differently from various suppliers, I’m using OpenAI to tell me what data is in each one.

Help needed
:point_right: Does it matter how these docs are uploaded?

For example, will a multi-page PDF, and it could just be two pages, be harder to work with in OpenAI vs. multiple images? I can’t think of many examples where you’d need more than one page, I guess a car owner might just upload a 2-3 page doc that has the info we need on one page, plus terms and conditions. The idea is to make it easy so we won’t ask them to upload page 2 of 3 instead of just the whole thing.

:point_right: What field type should I use? Multi-file? Multi-image? Image? Text? Again, the idea is an owner can just upload a file or a photo without do much fiddling around.

Have you tried using “Document to Text” column/action? I don’t know if it works for images. If not, you can just extract the file extension and determine whether to use Document to Text or Image to text?

I have no idea if Document to Text works well with multiple pages, but I expect it to.

I will give it a go, thanks.

Basically because all doc will be structured differently I thought of sending the doc to Open AI. Not 100% sure how that would work, but I get that the API needs some text or maybe a PDF? as input and then I prompt it from there.

Interested to know if anyone else has done this.