Skip to content

kendra.md

Latest commit

 

History

History
151 lines (114 loc) · 5.66 KB

kendra.md

File metadata and controls

151 lines (114 loc) · 5.66 KB
Raw

Kendra

AWS's kendra service · portable interface driver.Kendra · AWS index

Operations (13)

Operation Description
CreateDataSource
CreateIndex
DeleteDataSource
DeleteIndex
DescribeDataSource
DescribeIndex
ListDataSources
ListIndices
ListTagsForResource
TagResource
UntagResource
UpdateDataSource
UpdateIndex

Optional capabilities

Discovered by type assertion; only some providers implement these.

AccessControls

AccessControls is the optional access control configuration capability.

Operation Description
CreateAccessControlConfiguration
DeleteAccessControlConfiguration
DescribeAccessControlConfiguration
ListAccessControlConfigurations
UpdateAccessControlConfiguration

BlockLists

BlockLists is the optional query suggestions block list capability.

Operation Description
CreateQuerySuggestionsBlockList
DeleteQuerySuggestionsBlockList
DescribeQuerySuggestionsBlockList
ListQuerySuggestionsBlockLists
UpdateQuerySuggestionsBlockList

Documents

Documents is the optional document-ingestion and search capability: documents

Operation Description
BatchDeleteDocument
BatchGetDocumentStatus
BatchPutDocument
Query
Retrieve

Experiences

Experiences is the optional search experience capability.

Operation Description
CreateExperience
DeleteExperience
DescribeExperience
ListExperiences
UpdateExperience

Faqs

Faqs is the optional FAQ capability (Kendra has no UpdateFaq operation).

Operation Description
CreateFaq
DeleteFaq
DescribeFaq
ListFaqs

FeaturedResults

FeaturedResults is the optional featured results capability.

Operation Description
BatchDeleteFeaturedResultsSet
CreateFeaturedResultsSet
DescribeFeaturedResultsSet
ListFeaturedResultsSets
UpdateFeaturedResultsSet

PrincipalMappings

PrincipalMappings is the optional user-to-group principal mapping capability.

Operation Description
DeletePrincipalMapping
DescribePrincipalMapping
ListGroupsOlderThanOrderingID
PutPrincipalMapping

QuerySuggestions

QuerySuggestions is the optional query suggestions capability.

Operation Description
ClearQuerySuggestions
DescribeQuerySuggestionsConfig
GetQuerySuggestions
UpdateQuerySuggestionsConfig

SyncJobs

SyncJobs is the optional data source synchronization capability.

Operation Description
ListDataSourceSyncJobs
StartDataSourceSyncJob
StopDataSourceSyncJob

Thesauri

Thesauri is the optional thesaurus capability.

Operation Description
CreateThesaurus
DeleteThesaurus
DescribeThesaurus
ListThesauri
UpdateThesaurus

Not in scope

  • Search is a simple term-matching engine over the documents added with BatchPutDocument: results are ranked by how many query terms match (title weighs three times the body) and the confidence bucket (VERY_HIGH to LOW) comes from the share of terms matched. There is no semantic ranking, synonym expansion, spelling correction, relevance tuning, collapsing, nested facets or user-context filtering. FAQ and suggested-answer result types are never produced, so a QueryResultTypeFilter of QUESTION_ANSWER or ANSWER returns nothing.
  • Text formats (PLAIN_TEXT, HTML, MD, CSV, JSON, XML, XSLT, RTF) are indexed from the inline Blob; binary formats (PDF, MS_WORD, PPT, MS_EXCEL) and documents given only an S3Path are indexed by title and attributes only, because the emulator does not parse binary files or read S3.
  • FAQ, thesaurus and query suggestions block list files are registered by reference: the S3 object is not read, so FileSizeBytes, TermCount, SynonymRuleCount, ItemCount and the FAQ question count in IndexStatistics are 0 and no block list is applied to suggestions.
  • Data source sync jobs never crawl a repository: a job succeeds with zero counters (under async settling it reports SYNCING first), and DocumentsDeleted only counts BatchDeleteDocument calls that name the job in DataSourceSyncJobMetricTarget.
  • Query suggestions are learned from the queries the index served that had a result and a word longer than four characters; MinimumNumberOfQueryingUsers is stored but not applied (users are not tracked) and DOCUMENT_ATTRIBUTES suggestions are not generated. The settings a fresh index starts with besides ENABLED and 180 days (5 minimum queries, 8 minimum users, queries without user information included) are the emulator's choice.
  • Experience Endpoints are synthesized and not reachable; experience entity and persona operations, SubmitFeedback and GetSnapshots are not implemented.
  • Index and child resources are created ACTIVE at once so IaC waiters complete; with --async-settle they report CREATING/UPDATING for a short window and operations on them raise ConflictException. An index is never FAILED, so ErrorMessage is always empty. The FeaturedResultsSet default status (ACTIVE) and the exact error message texts are not documented by AWS and are the emulator's choice; the exception types are the documented ones.