AWS::Bedrock::DataSource HierarchicalChunkingConfiguration

Settings for hierarchical document chunking for a data source. Hierarchical chunking splits documents into layers of chunks where the first layer contains large chunks, and the second layer contains smaller chunks derived from the first layer.

You configure the number of tokens to overlap, or repeat across adjacent chunks. For example, if you set overlap tokens to 60, the last 60 tokens in the first chunk are also included at the beginning of the second chunk. For each layer, you must also configure the maximum number of tokens in a chunk.

Syntax

To declare this entity in your AWS CloudFormation template, use the following syntax:

JSON

{
  "LevelConfigurations" : [ HierarchicalChunkingLevelConfiguration, ... ],
  "OverlapTokens" : Integer
}

YAML

  LevelConfigurations: 
    - HierarchicalChunkingLevelConfiguration
  OverlapTokens: Integer

Properties

LevelConfigurations

Token settings for each layer.

Required: Yes

Type: Array of HierarchicalChunkingLevelConfiguration

Minimum: 2

Maximum: 2

Update requires: Replacement

OverlapTokens

The number of tokens to repeat across chunks in the same layer.

Required: Yes

Type: Integer

Minimum: 1

Update requires: Replacement