Chunking
Chunking is the way long documents are cut into pieces before indexing: the cuts decide what the search can find, so a good chunk is a self-contained passage that keeps its meaning when read alone.
No jargon, no math
In plain words, what is it?
Why cut at all? Models read finite windows, and retrieval works with focused passages: an embedding computed over 80 pages becomes an average that resembles nothing. The main strategies: fixed-size splits, or structural cuts on headings, paragraphs or clauses (for contracts) and rows (for tables), with an overlap of a few lines so a sentence split at a seam stays whole on one side or the other. Too small and meaning dies (“unless otherwise stated” without the sentence that follows); too big and the embedding dilutes, and search loses the exact answer. Every chunk keeps metadata: document title, section, clause number. Where it breaks: a table cut from its caption, a liability clause separated from its exceptions.
Terms that live nearby.
Embedding
An embedding is a list of numbers that captures the meaning of a piece of text, image or audio, so that content with similar meaning gets similar numbers and the computer can measure closeness with plain arithmetic.
ReadVector Database
A vector database is a database built to store numeric fingerprints of content (embeddings) and find the closest ones to any query in milliseconds, powering semantic search, recommendations and AI retrieval.
ReadRetrieval-Augmented Generation
RAG is an architecture where the AI retrieves relevant documents from your knowledge base before writing an answer, so it quotes your data instead of guessing from memory.
Read§ 03
Want this running in your stack?
We design and ship AI systems with permissions, logs and human approval. It starts with a 30-minute scoping call.