how do i decide what kind of pdf loader

It really depends on your usecase. Since your pdfs contain tables so amybe nougat PDF loader or unstructured one can be good for you

The data I have are 10 .json files that I generate on a daily basis. So I don't know if it's worth passing this data directly to OpenAI Assitant or using my system with LlamaIndex and Pinecone.

I upload this data on a daily basis and draw conclusions from it.
Thanks