Feeding text data into a Custom GPT

I would like to develop a custom GPT analysing national sustainability strategies, and I have around 60 of them (PDF files). However, I can’t upload more than 10 of them - it simply throws an error (without any details) while saving the GPT. I saw in some another thread that there is such a limitation on the number of files. What would be the best strategy to bypass it? Merge 60 files in 10 batches, with each containing 6 texts? Deploy a vector database and pass an API to access it in the GPT Actions? Or something else?

Ideally, I need that the response from the GPT also points me to a citation from an uploaded document, e.g., a “German strategy aims at achieving SDG 7 by this and that, see p. XX” Is there a way to engineer a prompt to achieve this?

1 Like