A strategy chooses boundaries, size, overlap, and metadata based on document structure and downstream needs. Chunks that are too small lose context, while chunks that are too large reduce retrieval precision and consume more tokens.
Chunking divides long source material into smaller units that can be indexed, retrieved, or placed into a model context.
A strategy chooses boundaries, size, overlap, and metadata based on document structure and downstream needs. Chunks that are too small lose context, while chunks that are too large reduce retrieval precision and consume more tokens.