Goal
Chunk the documents, then run a keyword search for refund. Notice which page comes first, and which one is missing.
A RAG system — retrieval-augmented generation — answers questions about your own documents by finding the relevant passages first, then handing them to a language model. Everything depends on that finding step.
Documents are first cut into chunks, so the system can retrieve a paragraph rather than a whole manual. The oldest way to search them is by keyword: score each chunk by the words it shares with the question, weighting rare words more (that scoring is BM25, the classic algorithm behind most search boxes).
Try it
Search for "refund". The damaged-items page mentions "a full refund" and wins. The returns page — the one you actually want — says "Refunds", plural, and never the exact word. Keyword search doesn't know they're the same.