Chunking Your Documents

Definition: Chunking is splitting long documents into smaller passages before embedding them. You retrieve and feed the model chunks, not whole documents.

Why chunk at all?

• Models have a limited context window — you cannot paste a 300-page manual
• Smaller chunks give more precise retrieval — you fetch the exact relevant paragraph, not a whole chapter
• It keeps prompts shorter and cheaper

Run it yourself

The demo splits text into chunks of a fixed number of words. Try changing the chunk size and watch how the passages change.

Getting chunk size right

Too small and a chunk loses context; too large and retrieval gets fuzzy and expensive. Real systems often use a few hundred words per chunk, with a little overlap between chunks so ideas are not cut in half.

💡 Tip: chunking strategy is one of the biggest levers on RAG quality — more on that later.