Azure/azure-sdk-for-java com.azure+azure-search-documents_12.2.0

on GitHub

12.2.0 (2026-11-13)

These release notes compare 12.2.0 with the previous GA release, 12.1.1. Changes that apply only when upgrading
from a preview release are not included.

Features Added

• Added support for the 2026-10-01 GA service version, which is now the default service version.
• Added SearchOptions.moreLikeThis to find documents similar to an existing document.
• Added GetDocumentOptions and three-argument getDocumentWithResponse overloads on SearchClient and
  SearchAsyncClient to group document lookup authorization, elevated-read, and selected-field options while
  preserving existing overloads, including getDocument(key, null).
• Added query-source authorization and elevated-read controls to SearchOptions and document lookup overloads, and
  query-source authorization overloads and a header-only KnowledgeBaseRetrievalOptions.querySourceAuthorization
  option to knowledge base retrieval. Non-null explicit authorization arguments override the option; null falls
  back to the option.
• Added document-level permission configuration through SearchField.permissionFilter, organizationAccessExpiration,
  and sharepointSiteUrl, SearchIndex.permissionFilterOption, and SharePointConnectorAppRegistration.
• Added SearchIndexerDataSourceConnection.indexerPermissionOptions,
  IndexingParametersConfiguration.refreshAllAcls, and
  SearchIndexerDataUserAssignedIdentity.federatedIdentityClientId for permission ingestion and federated identity.
• Added SearchIndexerClient.resync and SearchIndexerAsyncClient.resync to resynchronize indexed permissions.
• Added SearchIndexerStatus.runtime and SearchServiceStatistics.indexersRuntime, including used and remaining
  runtime and the reporting period.
• Added service-driven pagination and name-prefix filtering for resource listings, with ListingSearchType and
  search, searchType, and pageSize query parameters. Added typed request-options overloads for synonym maps,
  indexers, data sources, skillsets, and indexes with selected properties.
• Added pageable index statistics through listIndexStatsSummary and IndexStatisticsSummary.
• Added file knowledge sources and multipart file upload, update, listing, and deletion APIs on SearchIndexClient
  and SearchIndexAsyncClient, with file metadata, directory-like prefixes, parsing and extraction modes, and
  retrieval parameter, reference, and activity models.
• Added KnowledgeSourceFileCapacity and KnowledgeSourceStatus.fileCapacity for maximum file size, maximum file
  count, and remaining file count.
• Added indexed SQL and indexed SharePoint knowledge sources, their retrieval parameter, reference, and activity
  models, SQL content and embedding column mappings, and SharePoint container selection.
• Added knowledge source ingestion permission options and network access mode, and
  SearchIndexKnowledgeSourceParameters.baseFilter.
• Added low and medium retrieval reasoning effort, answer-synthesis and extractive-data output modes, retrieval and
  answer instructions, and CORS configuration on knowledge bases.
• Added model-based createOrUpdateKnowledgeBase and typed knowledge base/source update response overloads on
  SearchIndexAsyncClient.
• Added conversation messages, reasoning effort, output mode, and maximum output document counts to knowledge
  retrieval requests. Added constructors for message content and image URLs, and a writable message role.
• Added alwaysQuerySource, failOnError, and maxOutputDocuments to knowledge source retrieval parameters, and
  language, market, count, and freshness to web knowledge source configuration.
• Added typed source, query-planning, and answer-synthesis activity records, activity timestamps and warnings, model
  and deployment details, and citation URLs on search-index, Azure Blob, and indexed OneLake references.
• Added typed retrieveStream APIs returning CloseableIterableStream<KnowledgeBaseRetrievalStreamEvent> on the
  synchronous retrieval client and Flux<KnowledgeBaseRetrievalStreamEvent> on the asynchronous client. Events
  include retrieval start, activity start and completion, answers, references, errors, and response completion.
  Response-completion events expose their numeric status code as Integer, including unknown status values.
• Added modelName and modelDeployment to ContentUnderstandingSkill, fixed-size and semantic chunking method
  selection, and token-based chunking units.
• Added GPT54, GPT55, GPT56SOL, GPT56TERRA, and GPT56LUNA to AzureOpenAIModelName.

Breaking Changes

These removals correct preview-only APIs accidentally exposed by the GA specification. The corresponding features
were not supported by the previous GA service API. The specification correction was signed off as a bug fix in
Azure/azure-rest-api-specs#47112.
Removing these published Java APIs remains a source and binary compatibility break, despite the lack of supported
service behavior.

The following APIs were public in GA 12.1.1 and have been removed in 12.2.0:

• Removed AIFoundryModelCatalogName, AzureMachineLearningParameters, and AzureMachineLearningVectorizer.
• Removed VectorizableImageBinaryQuery and VectorizableImageUrlQuery, along with VectorQueryKind.IMAGE_BINARY
  and VectorQueryKind.IMAGE_URL.
• Removed VectorFilterMode.STRICT_POST_FILTER.
• Removed KnowledgeBaseAgenticReasoningActivityRecord.getRetrievalReasoningEffort(). For streaming retrieval,
  the effective reasoning effort is available through
  KnowledgeBaseRetrievalStartedStreamEvent.getValue().getReasoningEffort().

Bugs Fixed

• Fixed knowledge source status deserialization for compact synchronization intervals such as 1d, while
  preserving the existing Java Duration API.