Chunking
Splitting a document into parts that can be indexed, searched, and passed to a model separately while preserving context about the source.
On this page
Source content checked . English translation published 2026-10-27.
In business automation
Chunks need to be small enough to retrieve and complete enough to understand. Dividing by headings or units of meaning can preserve context better than arbitrary cuts in the middle of a rule.
Examples
These examples illustrate a possible workflow. Adapt the connections and rules to your own systems.
Business process example
A returns policy is divided by topic, and each chunk retains its heading, document ID, and date. A search for a damaged product returns the conditions and exceptions together instead of half a sentence from the policy.
Relevant tools
- LlamaIndex
- Document ingestion and indexing tools let you adapt chunking to search needs.
What to check
- Do not leave an exception to a rule in a disconnected chunk without context. Check what is retrieved for real questions, not just file sizes.