Home › Key concepts › Chunking
Key concept · Walk (intermediate) · the same on every platform
Chunking
How you cut documents into pieces decides what search can find.
What it is
Documents are split into chunks before embedding. Too small and a chunk loses its meaning; too big and it mixes topics. Good chunking follows the document's own structure, such as headings and clauses, and keeps a little overlap.
Why it matters
Most RAG quality problems are chunking problems.
An example
A policy is split by clause, each chunk keeping its clause number and section title.
The common mistake
Fixed 500-character cuts that split a table or sentence in half.
Where you meet it on each platform
The idea stays the same; only the product name changes.
| Job | AWS | Azure | Google Cloud | Databricks | Open source |
|---|---|---|---|---|---|
| Vector searchfind passages by meaning, not exact words | OpenSearch Serverless / Aurora pgvector | Azure AI Search | Vertex AI Vector Search / AlloyDB | Vector Search | pgvector or Qdrant |
Related ideas
Where it shows up
Part of the Walk level on these paths: AI Engineering (knowledge).