NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
NuGet · #487 most downloaded on NuGet
This is the Azure AI Search client library for developing .NET applications with rich search experiences. It enables you to query your indexes, update documents in your indexes, and manage indexes, synonym maps, indexers, data sources, or other service-level resources.
Last release 5 months ago
01 May 2026
Release timing varies
gaps range from 3 weeks to 1.2 years
Nearly every release is documented
notes for 11 of 12 stable releases
1 version withdrawn
withdrawn after publishing
127 years old
13 releases · first in 1900
Code | Docs Support: Active
Code | Docs
Support: Active
2026-04-01 service version.KnowledgeBaseRetrievalClient for agentic retrieval operations against knowledge bases.KnowledgeBase, KnowledgeBaseModel, KnowledgeBaseAzureOpenAIModel, and KnowledgeSourceReference.AzureBlobKnowledgeSource, SearchIndexKnowledgeSource, IndexedOneLakeKnowledgeSource, and WebKnowledgeSource with associated parameters.KnowledgeBaseRetrievalRequest, KnowledgeBaseRetrievalResponse, KnowledgeBaseMessage, KnowledgeBaseMessageContent, and KnowledgeBaseReference with typed variants (KnowledgeBaseAzureBlobReference, KnowledgeBaseSearchIndexReference, KnowledgeBaseIndexedOneLakeReference, KnowledgeBaseWebReference).KnowledgeBaseActivityRecord, KnowledgeBaseAgenticReasoningActivityRecord, and KnowledgeBaseModelWebSummarizationActivityRecord.KnowledgeBaseActivityRecordType.ModelWebSummarization value for LLM web summarization activity records.KnowledgeSourceIngestionParameters, KnowledgeSourceStatus, KnowledgeSourceStatistics, and SynchronizationState.ChatCompletionSkill, ContentUnderstandingSkill, and DocumentIntelligenceLayoutSkill with associated configuration models.AzureMachineLearningVectorizer and AzureMachineLearningParameters for AML-based vectorization.SearchAlias for index alias management.SearchIndexResponse model returned by index operations.NativeBlobSoftDeleteDeletionDetectionPolicy for native blob soft delete detection.VectorizableImageBinaryQuery and VectorizableImageUrlQuery for image-based vector search.DebugInfo property to SearchResults<T> and SearchResultsPage<T> for enhanced search debugging.VectorFilterMode.StrictPostFilter option for strict post-filtering in vector search.SearchServiceCounters.AliasCounter for alias resource tracking.SearchServiceLimits.MaxCumulativeIndexerRuntimeSeconds for runtime constraints.IndexerChangeTrackingState for indexer change tracking status.LookupDocument model for document lookup responses.SemanticErrorMode, SemanticErrorReason, SemanticSearchResultsType, and VectorFilterMode.AIServicesAccountIdentity and AIServicesAccountKey for AI services authentication.SearchClientSettings to support creating a SearchClient from IConfiguration, including configuration-based credential resolution and dependency injection registration.SearchIndexClientSettings to support creating a SearchIndexClient from IConfiguration, including configuration-based credential resolution and dependency injection registration.SearchIndexerClientSettings to support creating a SearchIndexerClient from IConfiguration, including configuration-based credential resolution and dependency injection registration.KnowledgeBaseRetrievalClientSettings to support creating a KnowledgeBaseRetrievalClient from IConfiguration, including configuration-based credential resolution and dependency injection registration.EntityRecognitionSkill.SkillVersion.V1 is now obsolete. Use V3 instead.EntityRecognitionSkill.IncludeTypelessEntities is now obsolete. Use V3 instead.SentimentSkill.SkillVersion.V1 is now obsolete. Use V3 instead.One column per quarter.
Added support for 2025-09-01 service version.
2025-09-01 service version.SemanticConfiguration.RankingOrder.VectorSearchCompression.RescoringOptions to configure how vector compression handles the original vector when indexing and how vectors are used during rescoring.SearchIndex.Description to provide a textual description of the index.LexicalNormalizer when defining SearchIndex, SimpleField, and SearchableField and the ability to use it when analyzing text with SearchIndexClient.AnalyzeText.DocumentIntelligenceLayoutSkill skillset skill and OneLake SearchIndexerDataSourceConnection data source.QueryDebugMode in searching to retrieve detailed information about search processing. Only vector is supported for QueryDebugMode.VectorSearchCompression.RerankWithOriginalVectors and VectorSearchCompression.DefaultOversampling don't work with
2025-09-01 and were replaced by VectorSearchCompression.RescoringOptions.EnabledRescoring and
VectorSearchCompression.RescoringOptions.DefaultOversampling. If using 2024-07-01 continue using the old properties,
otherwise if using 2025-09-01 use the new properties in RescoringOptions.Added support for HTTP status code 206 (Partial Content) in the Search method to handle partial results returned by the service.
Search method to handle partial results returned by the service.Added support for 2024-07-01 service version.
2024-07-01 service version.SemanticSearchOptions now supports SemanticQuery, which allows for specifying a semantic query that is only used
for semantic reranking.VectorQuery now supports Oversampling and Weight, which allows for specifying richer configurations on how
vector queries affect search results.VectorizableTextQuery, which allows for passing a text-based query that is vectorized service-side
by VectorSearchVectorizers configured on the index so that vectorization doesn't need to happen before querying.VectorSearchVectorizer, with implementations AzureOpenAIVectorizer
and WebApiVectorizer. This enables the service to use a user-provided configuration for vectorizing text, rather
than requiring all client-side calls to vectorize before querying, allowing for easier standardization of vectorization.VectorSearchCompression, with implementations BinaryQuantizationCompression
and ScalarQuantizationCompression. This allows for reducing the size of vectors in the index, which can reduce
storage costs and improve querying performance.VectorEncodingFormat, which allows for specifying the encoding format of the vector data.AzureOpenAIEmbeddingSkill, which is a skill that uses the Azure OpenAI service to create text
embeddings during indexing.SearchIndexerIndexProjection, which allows for specifying how indexed
documents are projected in the index (or indexes).SearchFieldDataType. This allows for specifying smaller types for vector fields
to reduce storage costs and improve querying performance.SearchIndexerDataIdentity, which allows for specifying the identity for the data source for the
indexer.SearchField and SearchableField now support IsStored and VectorEncodingFormat configurations. IsStored allows
for specifying behaviors on how the index will retain vector data (enabling the ability to reduce storage costs), and
VectorEncodingFormat allows for specifying the encoding format of the vector data.OcrSkill now supports LineEnding, which allows for specifying the line ending character used by the OCR skill.SplitSkill now supports MaximumPagesToTake and PageOverlapLength, which allows for specifying how the split
skill behaves when splitting documents into pages.SearchServiceLimits now supports MaxStoragePerIndexInBytes, which shows the maximum storage allowed per index.2024-07-01 GA have been removed. This
includes concepts such as index aliases, normalizers, Azure Machine Learning skills, hybrid search, and more.Fix paging issue for semantic and vector search (#40137).
Deprecated SentimentSkill.SkillVersion.V1 and EntityRecognitionSkill.SkillVersion.V1; use SentimentSkill.SkillVersion.V3 and EntityRecognitionSkill.Sk…
PiiDetectionSkill. It allows you extracts personal information from an input text and gives you the option of masking it using the Text Analytics API.OcrSkill and ImageAnalysisSkill as we have upgraded them to use Cognitive Services Computer Vision v3.2, which now includes support for additional languages. Refer to the language lists here. SplitSkill. Language lists can be found here.SentimentSkill.SkillVersion.V1 and EntityRecognitionSkill.SkillVersion.V1; use SentimentSkill.SkillVersion.V3 and EntityRecognitionSkill.SkillVersion.V3 instead.…11.3.0 GA release. Other preview features and breaking changes from the 11.4.0-beta.1 to 11.4.0-beta.9 releases are not included in this GA release.
TokenCredential when creating a SearchClient, SearchIndexClient or a SearchIndexerClient. For example, you can get started with new SearchClient(endpoint, new DefaultAzureCredential()) to authenticate via AAD using Azure.Identity. For more details see how to use role-based authentication in Azure Cognitive Search.SearchClientOptions.Audience to allow users to select the Azure cloud where the resource is located (#30306).SearchOptions properties by adding links to REST docs - https://github.com/Azure/azure-sdk-for-net/issues/22808SearchOptions.SessionId property was not passed to the service (#27549)Adds stable features and bug fixes from the 11.3.0-beta.1 and 11.3.0-beta.2 releases. Preview service features not generally available yet, like Seman
Nothing published for this version
Added setters for MaxLength and MinLength in LengthTokenFilter.
MaxLength and MinLength in LengthTokenFilter.SearchIndexingBufferedSender<T>.IndexActionEventArgs<T> to track indexing actions.IndexActionCompletedEventArgs<T> to track the completion of an indexing action.IndexActionFailedEventArgs<T> to track the failure of an indexing action.SearchIndexingBufferedSenderOptions<T>.MaxRetries to SearchIndexingBufferedSenderOptions<T>.MaxRetriesPerIndexAction.SearchIndexingBufferedSenderOptions<T>.MaxRetryDelay to SearchIndexingBufferedSenderOptions<T>.MaxThrottlingDelay.SearchIndexingBufferedSenderOptions<T>.RetryDelay to SearchIndexingBufferedSenderOptions<T>.ThrottlingDelay.SearchClient.CreateIndexingBufferedSender<T>(). Instead, callers are expected to use the public constructor of SearchIndexingBufferedSender<T>.Bug in TaskExtensions.EnsureCompleted method that causes it to unconditionally throw an exception in the environments with synchronization context
Added SearchClientOptions.Serializer to set which ObjectSerializer to use for serialization.
SearchClientOptions.Serializer to set which ObjectSerializer to use for serialization.FieldBuilder to easily create SearchIndex fields from a model type.$select from the query parameters logged by default. You can add it back via SearchClientOptions.Diagnostics.LoggedQueryParameters("$select"); if desired.Removed preview version SearchClientOptions.ServiceVersion.V2019_05_06_Preview and added version SearchClientOptions.ServiceVersion.V2020_06_30.
SearchClientOptions.ServiceVersion.V2019_05_06_Preview
and added version SearchClientOptions.ServiceVersion.V2020_06_30.Azure.Core.Experimental reference and features until they're moved
into Azure.Core.SearchServiceCounters.SkillsetCounter.new SearchOptions(string continuationToken) overload.Your coding agent can read these notes before it upgrades. Set up the MCP server →