Retrieve

kendra_retrieve

R Documentation

Retrieves relevant passages or text excerpts given an input query¶

Description¶

Retrieves relevant passages or text excerpts given an input query.

This API is similar to the query API. However, by default, the query API only returns excerpt passages of up to 100 token words. With the retrieve API, you can retrieve longer passages of up to 200 token words and up to 100 semantically relevant passages. This doesn't include question-answer or FAQ type responses from your index. The passages are text excerpts that can be semantically extracted from multiple documents and multiple parts of the same document. If in extreme cases your documents produce zero passages using the retrieve API, you can alternatively use the query API and its types of responses.

You can also do the following:

Override boosting at the index level
Filter based on document fields or attributes
Filter based on the user or their group access to documents
View the confidence score bucket for a retrieved passage result. The confidence bucket provides a relative ranking that indicates how confident Amazon Kendra is that the response is relevant to the query.

Confidence score buckets are currently available only for English.

You can also include certain fields in the response that might provide useful additional information.

The retrieve API shares the number of query capacity units that you set for your index. For more information on what's included in a single capacity unit and the default base capacity for an index, see Adjusting capacity.

If you're using an Amazon Kendra Gen AI Enterprise Edition index, you can only use ATTRIBUTE_FILTER to filter search results by user context. If you're using an Amazon Kendra Gen AI Enterprise Edition index and you try to use USER_TOKEN to configure user context policy, Amazon Kendra returns a ValidationException error.

Usage¶

kendra_retrieve(IndexId, QueryText, AttributeFilter,
  RequestedDocumentAttributes, DocumentRelevanceOverrideConfigurations,
  PageNumber, PageSize, UserContext)

Arguments¶

IndexId

[required] The identifier of the index to retrieve relevant passages for the search.

QueryText

[required] The input query text to retrieve relevant passages for the search. Amazon Kendra truncates queries at 30 token words, which excludes punctuation and stop words. Truncation still applies if you use Boolean or more advanced, complex queries. For example, ⁠Timeoff AND October AND Category:HR⁠ is counted as 3 tokens: timeoff, october, hr. For more information, see Searching with advanced query syntax in the Amazon Kendra Developer Guide.

Filters search results by document fields/attributes. You can only provide one attribute filter; however, the AndAllFilters, NotFilter, and OrAllFilters parameters contain a list of other filters.

The AttributeFilter parameter means you can create a set of filtering rules that a document must satisfy to be included in the query results.

For Amazon Kendra Gen AI Enterprise Edition indices use AttributeFilter to enable document filtering for end users using ⁠_email_id⁠ or include public documents (⁠_email_id=null⁠).

RequestedDocumentAttributes

A list of document fields/attributes to include in the response. You can limit the response to include certain document fields. By default, all document fields are included in the response.

DocumentRelevanceOverrideConfigurations

Overrides relevance tuning configurations of fields/attributes set at the index level.

If you use this API to override the relevance tuning configured at the index level, but there is no relevance tuning configured at the index level, then Amazon Kendra does not apply any relevance tuning.

If there is relevance tuning configured for fields at the index level, and you use this API to override only some of these fields, then for the fields you did not override, the importance is set to 1.

PageNumber

Retrieved relevant passages are returned in pages the size of the PageSize parameter. By default, Amazon Kendra returns the first page of results. Use this parameter to get result pages after the first one.

PageSize

Sets the number of retrieved relevant passages that are returned in each page of results. The default page size is 10. The maximum number of results returned is 100. If you ask for more than 100 results, only 100 are returned.

UserContext

The user context token or user and group information.

Value¶

A list with the following syntax:

list(
  QueryId = "string",
  ResultItems = list(
    list(
      Id = "string",
      DocumentId = "string",
      DocumentTitle = "string",
      Content = "string",
      DocumentURI = "string",
      DocumentAttributes = list(
        list(
          Key = "string",
          Value = list(
            StringValue = "string",
            StringListValue = list(
              "string"
            ),
            LongValue = 123,
            DateValue = as.POSIXct(
              "2015-01-01"
            )
          )
        )
      ),
      ScoreAttributes = list(
        ScoreConfidence = "VERY_HIGH"|"HIGH"|"MEDIUM"|"LOW"|"NOT_AVAILABLE"
      )
    )
  )
)

Request syntax¶

svc$retrieve(
  IndexId = "string",
  QueryText = "string",
  AttributeFilter = list(
    AndAllFilters = list(
      list()
    ),
    OrAllFilters = list(
      list()
    ),
    NotFilter = list(),
    EqualsTo = list(
      Key = "string",
      Value = list(
        StringValue = "string",
        StringListValue = list(
          "string"
        ),
        LongValue = 123,
        DateValue = as.POSIXct(
          "2015-01-01"
        )
      )
    ),
    ContainsAll = list(
      Key = "string",
      Value = list(
        StringValue = "string",
        StringListValue = list(
          "string"
        ),
        LongValue = 123,
        DateValue = as.POSIXct(
          "2015-01-01"
        )
      )
    ),
    ContainsAny = list(
      Key = "string",
      Value = list(
        StringValue = "string",
        StringListValue = list(
          "string"
        ),
        LongValue = 123,
        DateValue = as.POSIXct(
          "2015-01-01"
        )
      )
    ),
    GreaterThan = list(
      Key = "string",
      Value = list(
        StringValue = "string",
        StringListValue = list(
          "string"
        ),
        LongValue = 123,
        DateValue = as.POSIXct(
          "2015-01-01"
        )
      )
    ),
    GreaterThanOrEquals = list(
      Key = "string",
      Value = list(
        StringValue = "string",
        StringListValue = list(
          "string"
        ),
        LongValue = 123,
        DateValue = as.POSIXct(
          "2015-01-01"
        )
      )
    ),
    LessThan = list(
      Key = "string",
      Value = list(
        StringValue = "string",
        StringListValue = list(
          "string"
        ),
        LongValue = 123,
        DateValue = as.POSIXct(
          "2015-01-01"
        )
      )
    ),
    LessThanOrEquals = list(
      Key = "string",
      Value = list(
        StringValue = "string",
        StringListValue = list(
          "string"
        ),
        LongValue = 123,
        DateValue = as.POSIXct(
          "2015-01-01"
        )
      )
    )
  ),
  RequestedDocumentAttributes = list(
    "string"
  ),
  DocumentRelevanceOverrideConfigurations = list(
    list(
      Name = "string",
      Relevance = list(
        Freshness = TRUE|FALSE,
        Importance = 123,
        Duration = "string",
        RankOrder = "ASCENDING"|"DESCENDING",
        ValueImportanceMap = list(
          123
        )
      )
    )
  ),
  PageNumber = 123,
  PageSize = 123,
  UserContext = list(
    Token = "string",
    UserId = "string",
    Groups = list(
      "string"
    ),
    DataSourceGroups = list(
      list(
        GroupId = "string",
        DataSourceId = "string"
      )
    )
  )
)