How a RAG chatbot answers from your own content (and where it still fails)
A RAG chatbot answers in two steps: it retrieves the passages of your own content that match the question, then writes an answer from those passages only, declining when nothing matches. This guide walks one question through both steps, explains chunks, embeddings and keyword search in plain words, lists the ways retrieval still fails on a real store and the content fix for each, and shows how to test it.