Salesforce Docs Portal

Search Index Reference

Data 360 search supports many different file formats with unstructured data, embedding models, and pre-filter operations on search results.

Required Editions

Available in: All Editions supported by Data 360. See Data 360 edition availability.

Supported Embedding Models

Data 360 supports these embedding models.

Model Modality Max Sequence/Input length
E5-large-v2 Text 512 tokens
Multilingual-E5-large Text 512 tokens
SFR-v2-small Text 8,000 tokens

Supported Large Language Models

With LLM-based parsing and LLM-based Visual Data Preprocessing you can override the default GPT-4o model to use any supported LLM that best fits your business needs and your content. You can only select LLMs that are approved for use by your admin. If your admin turns off a model that is currently in use, your previously processed content remains accessible. Data 360 supports these Large Language Models (LLMs).

  • Anthropic Claude Opus 4.6 Opus on Amazon
  • Anthropic Claude Sonnect 4.6 on Amazon
  • Google Gemini 3.5 Flash
  • GPT 4.o Omni
  • GPT 5.4

Supported File Formats for Unstructured Data

Data 360 supports several file formats for unstructured data and search index configurations. See Unstructured Data File Formats and Connectors for details.

Supported Operators

Enhance your search queries with pre-filtering fields. Use logical operators and manage case sensitivity for field names and values to ensure precise data retrieval.

Data Type Supported Operators
Number and Percent
  • <  (Less than)
  • <= (Less than or equal to)
  • =  (Equal to)
  • > (Greater than)
  • >= (Greater than or equal to)
Date and DateTime
  • <  (Less than)
  • <= (Less than or equal to)
  • =  (Equal to)
  • > (Greater than)
  • >= (Greater than or equal to)

Examples:

  • DateTimeField=''2024-05-21T00:00:00.000Z''
  • DateField=''2024-05-21''

Timezone can be any standard timezone, such as UTC, PST, or EST, or it can be an offset, such as +05:30 or -08:00. If you specify no timezone, the system assumes UTC.
Text, URL, Email, and Phone
  • =  (Equal to)
  • != (Not Equal to)
  • in (‘option1’, ‘option2’, ‘option3’) -> To match any of the listed options. A maximum of 10 options can be provided.
  • not in (‘option1’, ‘option2’, ‘option3’) -> To match none of the listed options. A maximum of 10 options can be provided.
  • LIKE (example: LIKE ‘d/%’)
Boolean

= (Equal to)

Example: <BooleanField>=true

Search Index Considerations

Note these limitations when working with search indexes.

  • Search index supports only UTF-8 encoding. Multilingual content must use UTF-8 encoding.
  • Search indexes are not supported in Developer Edition orgs if namespaces are enabled in those orgs.
  • If you’re creating a search index for unstructured data that you’ve connected from an external blob store to Data 360, ensure that you have set up a file notification pipeline on your blob store.
  • A DMO that has mappings from one or more external DLOs can only be used as the data source if caching is enabled for all those external DLOs. For more information see, Caching in Data Federation.
  • Search Indexes don't support date and datetime fields for chunking. Only Text data type fields are available for chunking data.
  • Modifying data space filters on a Data Lake Object (DLO) after creating a search index on the associated Data Model Object (DMO) doesn't automatically update the search index. The index keeps data indexed under the old filter criteria, even if that data no longer qualifies. Only new or modified records that the system processes after the filter change reflect the updated criteria. To make sure that your search index accurately reflects the current data space filters, rebuild the index. To reprocess all records with the updated filter criteria, from the search index detail page, click Rebuild.
  • Civilian Government Cloud environments support enriched indexes, but defense Government Cloud environments don't.