Glossary · Data & retrieval

Chunking

Splitting a document into overlapping segments small enough to embed and precise enough to retrieve.

What is chunking?

Chunking divides cleaned content into pieces of a set character length with a set overlap — 1200 and 120 by default, both tunable per call. Each chunk is embedded and stored separately, so retrieval returns the passage that answered the query rather than the whole document.

Why it matters

Chunk size is the main quality dial in a retrieval pipeline. Too large and a match returns mostly irrelevant text; too small and the passage loses the context that made it meaningful. Overlap is what stops an answer that straddles a boundary from being lost.

Related terms

See it in context

Back to the full glossary

Put your agent to work. Keep the decision.

Give your agent a real browser, authorize the accounts it may use, and keep control of every step. Repeated runs compile into skills that replay without a model.