Loading

Data 360: Overview and Key Considerations of the "Enrich chunks" Option in Search Index

Fecha de publicación: Mar 3, 2026
Descripción

This article explains the overview of the "Enrich chunks" option in Search Index, the types of metadata generated, and key considerations when using it — including Flex Credit consumption and data transfer to overseas regions.

Solución

Role of the "Enrich chunks" Option in Search Index

"Enrich chunks" is a feature that automatically generates and assigns "additional information for search (metadata)" using an LLM when vectorizing unstructured data in Data 360.
This enables AI applications such as Agentforce to reference more appropriate information, which is expected to improve the relevance of responses.
 
Previously, when creating a search index, text was mechanically split into chunks of a fixed length. With "Enrich chunks" enabled, Data 360 generates 3 chunks: one containing the original chunk text, one containing metadata text, and one containing a set of answerable questions. Retrievers access these chunks and use them in RAG and AI agent workflows.
 
 
Chunk TypeMetadata TypeDescription
Metadata
Keyword
Extracts keywords that uniquely identify the content.
Metadata
Entity
Progressively identifies and lists the following elements:

・location, organization, date, time 
・currency, amount, percentage 
・product, mentioned platforms
Metadata
Topic
Extracts the main topics discussed in the chunk. Topic names include important entities and keywords.
Metadata
Title
Generates a concise, informative title summarizing the content.
Metadata
Summary
An overview of the content with key topics and entities highlighted. Generated in 100+ and 250 or fewer words when possible.
Question
Question
Compiles a set of detailed and comprehensive questions that the text chunk content can answer.
 

Key Considerations

1. Flex Credits (Standard Prompt) are consumed with "Enrich chunks" in Search Index

The representative credit types for Data Services Cards used in Search Index processing are the following two:
  • Unstructured Data Processed
  • Intelligent Processing
 
However, when "Enrich chunks" is enabled, since the chunk splitting process involves analyzing content with an LLM on Amazon Bedrock, Flex Credits (Standard Prompt) are consumed in addition to the above Data Services Cards.
 

(Reference)

 

2. Data may be transferred to overseas regions during Search Index processing

"Enrich chunks" is supported in all regions where Data 360 is available, but the actual data processing is performed through Amazon Bedrock Managed Services limited to the following regions:
  • US: us-west-2 (Oregon), us-east-1 (Virginia)
  • Europe: eu-central-1 (Frankfurt), eu-central-2 (Zurich), eu-west-3 (Paris)
Therefore, even if you are using a Japanese region, when using Search Index, customer data is transferred to the above overseas regions and processed there.

(Reference)
Número del artículo de conocimiento

005299115

 
Cargando
Salesforce Help | Article