DataObjectSearchIndexConf in SFDX

Architectural Overview#

The DataObjectSearchIndexConf metadata type configures vector search indexing, semantic embedding pipelines, and hybrid retrieval parameters within Salesforce Data Cloud. In the era of autonomous AI agents and Agentforce, generative language models require accurate, real-time grounding in enterprise knowledge and unstructured customer interaction histories.

DataObjectSearchIndexConf defines how Data Model Objects (DMOs) and Data Lake Objects (DLOs)—such as customer support transcripts, knowledge articles, product manuals, or PDF claims—are chunked, vectorized using embedding models, and indexed. When an Agentforce agent or Prompt Template executes a semantic query, the Data Cloud vector database relies on this configuration to perform high-precision hybrid retrieval.

Directory and File Suffix Conventions#

In SFDX source format, DataObjectSearchIndexConf metadata files reside in the dataObjectSearchIndexConfs directory:

Key Metadata Fields#

Declarative XML and Manifest Examples#

Source Format XML (Knowledge_Vector_Index.dataObjectSearchIndexConf-meta.xml)#

<?xml version="1.0" encoding="UTF-8"?>
<DataObjectSearchIndexConf xmlns="http://soap.sforce.com/2006/04/metadata">
    <developerName>Knowledge_Vector_Index</developerName>
    <masterLabel>Knowledge Base Vector Search Index</masterLabel>
    <sourceDataObject>Knowledge_Article_DMO</sourceDataObject>
    <searchIndexType>Hybrid</searchIndexType>
    <chunkSize>512</chunkSize>
    <chunkOverlap>64</chunkOverlap>
</DataObjectSearchIndexConf>

Manifest (package.xml)#

<?xml version="1.0" encoding="UTF-8"?>
<Package xmlns="http://soap.sforce.com/2006/04/metadata">
    <types>
        <members>Knowledge_Vector_Index</members>
        <name>DataObjectSearchIndexConf</name>
    </types>
    <version>68.0</version>
</Package>

Retrieval and Deployment Commands#

Use the Salesforce CLI to retrieve and deploy DataObjectSearchIndexConf metadata:

# Retrieve all search index configurations
sf project retrieve start --metadata DataObjectSearchIndexConf

# Retrieve a specific vector search index configuration
sf project retrieve start --metadata DataObjectSearchIndexConf:Knowledge_Vector_Index

# Deploy vector index metadata to target environment
sf project deploy start --metadata DataObjectSearchIndexConf:Knowledge_Vector_Index

Senior Architect Best Practices and Gotchas#

• Chunk Sizing Tradeoffs: Tuning chunkSize and chunkOverlap directly affects retrieval precision. Smaller chunks (256–512 tokens) excel at specific factual question-answering, while larger chunks (1024 tokens) preserve holistic narrative context for Agentforce agents.
• Index Rebuild Latency: Deploying metadata updates to DataObjectSearchIndexConf (such as changing embedding models or chunk sizes) triggers a background re-indexing job across the target data lake. Plan major changes during low-utilization windows.
• Field-Level Security and Filtering: Ensure indexed fields do not contain raw unmasked PII unless access policies restrict Agentforce grounding prompts to authorized administrative audiences.

Frequently asked questions

What is the purpose of DataObjectSearchIndexConf?

It configures semantic vector search indexes and text chunking on Data Cloud objects to ground Agentforce autonomous agents and generative AI.

Where are DataObjectSearchIndexConf files stored in SFDX?

They are stored in force-app/main/default/dataObjectSearchIndexConfs/ with the file suffix .dataObjectSearchIndexConf-meta.xml.

Which search types can DataObjectSearchIndexConf define?

It configures vector embedding search for unstructured data as well as hybrid search combining semantic similarity and keyword filtering.

How do you retrieve DataObjectSearchIndexConf via CLI?

Execute sf project retrieve start with the metadata flag DataObjectSearchIndexConf.