Search Index

Specify the name of the search index that should be made available. This index must be configured and populated in advance. The AI Assistant will use this index to retrieve documents or data in response to user queries.

Usage

• Select from available search indexes configured in your environment.
• The index name must match the identifier used during index creation.

Impact of Changes

• Changing the index name will alter the set of documents and schema available to the AI Assistant.
• If the index does not exist or is not populated, search operations will fail.

Example

• Enter invoices to expose an index containing invoice documents.

Enable this option to allow the AI Assistant to use semantic search, which retrieves results based on the meaning of the query rather than just keyword matches. Semantic search is ideal for natural language questions, context-based retrieval, and scenarios where users may not know the exact terms present in the documents.

Requires that 'Vector Search' is enabled on the Indexing Behavior associated with the selected search index.

Usage

• Enable for indexes where users will ask open-ended or natural language questions.
• Use in combination with standard search for maximum flexibility.

Guidance When Both Search Types Are Enabled

• When both vector and standard search are enabled, the AI Assistant will prefer vector search for most user inquiries.
• Standard (keyword) search should only be used when the user clearly requests a field-based or exact keyword lookup.
• Example: Use vector search for "Show me invoices with late fees" and standard search for "Find documents where Status = 'Paid'".

Impact of Changes

• Disabling this option will prevent the AI Assistant from using semantic search, limiting retrieval to keyword-based queries.
• Enabling without proper index support will result in validation errors.

Enable this option to allow the AI Assistant to use standard keyword-based search, similar to what users can perform from the Search Page. Standard search is best for precise lookups, field-based queries, or when searching for specific values or terms.

Usage

• Enable for indexes where users need to perform exact or fielded searches.
• Use in combination with vector search to support both natural language and structured queries.

Guidance When Both Search Types Are Enabled

• When both search types are enabled, provide clear instructions to users on when to use each:
  • Use vector search for general questions or when unsure of the exact terms.
  • Use standard search for field-specific or exact value queries.
• The AI Assistant will default to vector search unless the user's intent is clearly for a keyword or fielded search.

Impact of Changes

• Disabling this option will prevent keyword-based lookups, limiting retrieval to semantic queries.

Define one or more Subset Filters to expose logical partitions of the search index, such as by document type or field value. Subsets allow users to target specific groups of documents for retrieval, improving relevance and performance.

Usage

• Add subsets to represent common categories, such as document types, departments, or statuses.
• Each subset is defined by a filter expression and a display name.

Impact of Changes

• Adding or removing subsets changes the options available to users when specifying which part of the index to search.
• If no subsets are defined, the entire index is available for search.

Example

• Create a subset named Contracts with a filter TypeName eq 'Contract' to allow targeted searches for contract documents.

Specify an OData filter expression to restrict all search queries to a subset of documents. This can be used to enforce security, limit results to a specific status, or apply other business rules.

Usage

• Enter a valid OData filter expression to apply to every search.
• Use to restrict access to sensitive documents or to focus searches on relevant content.

Impact of Changes

• Changing the filter will immediately affect all search results returned by the AI Assistant.
• Invalid filter expressions may cause search errors.

Example

• To limit searches to approved documents: Status eq 'Approved'
• To restrict by date: CreatedDate ge 2024-01-01

The 'Resource Name' is shown in the user interface and is used to reference this resource in configuration and during chat interactions. If an 'Alias' is provided, it will be used as the display name; otherwise, the default 'Name' is used.

Provide a clear and concise explanation of what this resource is for, what information or capabilities it provides, and how it should be used in relation to other available resources. This description is used by the AI Assistant to determine when and how to utilize the resource during a conversation.

If two resources would otherwise have the same 'Resource Name', you can specify an 'Alias' to ensure each resource can be uniquely identified. The alias will be used as the display name in the user interface and in chat interactions.