Ingest long policies into vector search.
Retrieve only relevant sections for the model context and reduce unrelated tokens.
What is the main purpose of chunking?
Chunking defines useful retrieval units so relevant passages can be selected and passed within the model's context limit.
Detailed explanation
Chunking and access control are separate concerns.
Appropriate chunks help retrieval return useful sections while controlling context size.
Chunking changes input preparation, not model parameters.
Chunking does not verify the content or its freshness.
Check official documentation
Compare the following points with the official documentation.
Read the definition of chunking in Task 2.1 and the Bedrock Knowledge Bases chunking guide. Compare preserving a policy-section boundary with splitting through it, and identify how retrieval units affect the supplied context.
Official references
What to confirm
Explain chunking as input-document preparation and distinguish retrieval granularity, context size, and information lost at boundaries.
Key points
- Retrieval unit
- Context size
- Relevant passages
Notes
- Environment: Reference review of the official AIF-C01 guide and AWS documentation; no AWS resources or inference are required.
- Task 2.1 includes the concept definition. RAG application design is related to Task 3.1, but this question asks for the purpose of chunking.
- Smaller chunks are not always better. More overlap may preserve context while increasing duplicated retrieved text.
Foundation review
Boundary loss
Chunks that are too small can separate a question from the details needed to answer it.
Retrieval precision
Very large or poorly scoped chunks can add irrelevant text and consume context capacity.