Ok RAG, but what about data extraction from documents?

Hi everyone,
I'm working on my RAG system — or rather, thinking about how to make it strong and general-purpose. I've started with the first step: extracting data from uploaded documents.

I'm facing a lot of issues — there are many models, many open-source ones, and many others that are quite costly — but none can guarantee that 100% (or even close) of the data will be extracted correctly from my PDFs, DOCX files, or other formats.

I also have another problem: I'm Italian and want to build this for an Italian audience, so the documents will be in Italian — and some extractors don’t handle that very well.

So my question is: what kinds of systems, tools, or approaches do you use to extract all the information from your documents before the chunking and embedding phase?

Let me know, thanks!