Notes Are Too Big to Just Paste In

Split notes into paragraph-sized chunks so retrieval can point at the exact passage that answers a question.

Part of: Personal Notes Brain

You have hundreds of markdown files: daily journals, meeting notes, recipe scraps, half-finished project plans. You want to ask "what did I decide about the budget last March?" and get a real answer with a pointer back to the note. Pasting your whole vault into one prompt does not work. It is too big, too expensive, and mostly irrelevant to any single question. The answer is RAG , or Retrieval-Augmented Generation. The RAG loop 1. Index every note once by splitting it into small, retrievable pieces called chunks . 2. Retrieve only the chunks that relate to the question, at question time. 3. Synthesize an answer from those chunks and cite which chunk backed which claim. This lesson builds step 1. The rest of the project builds steps 2 and 3. Why chunk at all A whole note is usually too coarse. One journal entry might cover breakfast, a work decision, and a doctor's appointment. If your question is only about the work decision, retrieving the entire entry wastes tokens on two paragraphs you don't need. Chunking splits each note into paragraph-sized pieces, so retrieval hands back the one paragraph that matters instead of a whole day. Each chunk records where it came from in note and

Challenge: Chunk the Notes