Options
Generated Chunks (0)
Frequently Asked Questions
What is RAG chunking?
Retrieval-Augmented Generation (RAG) requires splitting documents into smaller "chunks" before embedding them into a vector database. The chunk size directly affects retrieval quality.
Why does overlap matter?
Overlap ensures that a sentence split between two chunks is still retrievable. Without overlap, queries matching content at chunk boundaries may fail to find relevant results.
What is a good chunk size?
It depends on the use case. For Q&A, 200-500 characters is common. For summarization, 1000+ works better. Experiment with your specific content to find the optimal balance.
Characters vs Tokens?
This tool uses character-based chunking for simplicity. In production, consider token-based chunking to match your embedding model's limits (e.g., 512 tokens for some models).
What is semantic chunking?
Advanced RAG pipelines use "semantic chunking" which splits text by meaning (paragraphs, sections) instead of fixed sizes. Libraries like LangChain offer this out of the box.