Corpus Preparation & Chunking
Experiment with chunking strategies on sample documents. See exactly how text gets split, measure overlap, and test which strategy retrieves the best results.
Chunking Strategy Lab
Choose a sample document, pick a chunking strategy, and adjust the parameters. Watch the document split in real time and see how each strategy handles boundaries differently.
Retrieval Quality Tester
Enter a query and see which chunks each strategy would retrieve. The tester scores retrieval quality based on keyword overlap and context completeness.
Enter a query or click an example to see retrieval results across all three chunking strategies.
Key insight: Chunking is the most underrated step in RAG. Poor chunking — splitting mid-sentence, losing headers, or creating chunks too small for context — is the number one cause of bad retrieval. Semantic chunking preserves meaning but is slower. Fixed-size is fast but crude. Recursive splitting balances both by trying natural boundaries first.