The three steps
Split your documents into passages, find the passages relevant to a question, and give those passages to the model along with the question. The model answers from the text in front of it rather than from memory.
Why chunking decides quality
Passages that are too large bury the answer in noise; too small and they lose the context that makes them meaningful. Splitting on headings and paragraphs beats splitting on a fixed character count.
What still goes wrong
If retrieval returns the wrong passages, the answer is wrong no matter how capable the model. Most RAG debugging is search debugging.
