Building an application with OpenAI api that analyses multiple PDFs with bank account statements. Wh

[ancient grove]

I have multiple bank accounts in a few different countries. I want to be able to ask questions about it.

HOW I CURRENTLY DO IT:

1. I download all of my bank account statements (PDFs, CSVs, images...) and my family's (~20 statements, some are as long as 70 pages, some are 2 pages).
2. I upload them to ChatGPT.
3. I ask questions about them.

THE APP I WANT TO BUILD:

1. I upload all of my bank account statements to the app .
2. The answers to a set of pre-defined question are retrieved automatically.

HOW DO I ACHIEVE THIS?
I'm new to using the OpenAI api. I don't know how to achieve this. Some questions:

1. Can I submit PDFs, CSVs and images all through the same api call?
2. Which model can do this?
3. For the specific case of PDFs: is it better to
   ....a) convert to image and have openai answer questions about images? or
   ....b) extract text from the PDF and have openai find answers to questions on text?
4. Are there going to be problems with very long PDFs? What are some techniques to avoid such problems?

[loud hollow]

you need several api calls, probably over a hundred for all file types, logic and multiple files.

text is always better than images

[ancient grove]

@loud hollow how do I make these calls? I asked chatgpt and itself says models like 4o and o1 don't take in pdfs