Missing context if total number of chunks exceed 100

[Ari-stack123]

I'm trying to make app that will analyze each document and ask couple of questions to each file separately, then save answers and sources to csv. For programming I use ChatGPT because I have 0.01% skills. Right now everything works fine except when document is big enough, model doesn't get context. On printscreen is example with Orca_paper.pdf. I'm using the default embedding model, and groq API.