models Package

The multi-region account of an Azure AI service resource that's attached to a skillset.

The account key of an Azure AI service resource that's attached to a skillset, to be used with the resource's subdomain.

The result of testing an analyzer on text.

Specifies some text and analysis components used to break that text into tokens.

Information about a token returned by an analyzer.

Converts alphabetic, numeric, and symbolic Unicode characters which are not in the first 127 ASCII characters (the "Basic Latin" Unicode block) into their ASCII equivalents, if such equivalents exist. This token filter is implemented using Apache Lucene.

Credentials of a registered application created for your search service, used for authenticated access to the encryption keys stored in Azure Key Vault.

Configuration for Azure Blob Storage knowledge source.

Parameters for Azure Blob Storage knowledge source.

Specifies the properties for connecting to an AML vectorizer.

Specifies an Azure Machine Learning endpoint deployed via the Azure AI Foundry Model Catalog for generating the vector embedding of a query string.

Allows you to generate a vector embedding for a given text input using the Azure OpenAI resource.

Specifies the Azure OpenAI resource used to vectorize a query string.

Specifies the parameters for connecting to the Azure OpenAI resource.

Ranking function based on the Okapi BM25 similarity algorithm. BM25 is a TF-IDF-like algorithm that includes length normalization (controlled by the 'b' parameter) as well as term frequency saturation (controlled by the 'k1' parameter).

Contains configuration options specific to the binary quantization compression method used during indexing and querying.

Base type for character filters.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: MappingCharFilter, PatternReplaceCharFilter

Common language model parameters for Chat Completions. If omitted, default values are used.

Determines how the language model's response should be serialized. Defaults to 'text'.

Object defining the custom schema the model will use to structure its output.

Properties for JSON schema response format.

A skill that calls a language model via Azure AI Foundry's Chat Completions endpoint.

Forms bigrams of CJK terms that are generated from the standard tokenizer. This token filter is implemented using Apache Lucene.

Legacy similarity algorithm which uses the Lucene TFIDFSimilarity implementation of TF-IDF. This variation of TF-IDF introduces static document length normalization as well as coordinating factors that penalize documents that only partially match the searched queries.

Grammar-based tokenizer that is suitable for processing most European-language documents. This tokenizer is implemented using Apache Lucene.

Base type for describing any Azure AI service resource attached to a skillset.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: AIServicesAccountIdentity, AIServicesAccountKey, CognitiveServicesAccountKey, DefaultCognitiveServicesAccount

The multi-region account key of an Azure AI service resource that's attached to a skillset.

Construct bigrams for frequently occurring terms while indexing. Single terms are still indexed too, with bigrams overlaid. This token filter is implemented using Apache Lucene.

A skill that enables scenarios that require a Boolean operation to determine the data to assign to an output.

A skill that leverages Azure AI Content Understanding to process and extract structured insights from documents, enabling enriched, searchable content for enhanced document indexing and retrieval.

Controls the cardinality for chunking the content.

Defines options to control Cross-Origin Resource Sharing (CORS) for an index.

Resources created by the knowledge source. Keys represent resource types (e.g., 'datasource', 'indexer', 'skillset', 'index') and values represent resource names.

Allows you to take control over the process of converting text into indexable/searchable tokens. It's a user-defined configuration consisting of a single predefined tokenizer and one or more filters. The tokenizer is responsible for breaking text into tokens, and the filters for modifying tokens emitted by the tokenizer.

An object that contains information about the matches that were found, and related metadata.

A complex object that can be used to specify alternative spellings or synonyms to the root entity name.

A skill looks for text from a custom, user-defined list of words and phrases.

Allows you to configure normalization for filterable, sortable, and facetable fields, which by default operate with strict matching. This is a user-defined configuration consisting of at least one or more filters, which modify the token that is stored.

Base type for data change detection policies.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: HighWaterMarkChangeDetectionPolicy, SqlIntegratedChangeTrackingPolicy

Base type for data deletion detection policies.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: NativeBlobSoftDeleteDeletionDetectionPolicy, SoftDeleteColumnDeletionDetectionPolicy

Represents credentials that can be used to connect to a datasource.

An empty object that represents the default Azure AI service resource for a skillset.

Decomposes compound words found in many Germanic languages. This token filter is implemented using Apache Lucene.

Defines a function that boosts scores based on distance from a geographic location.

Provides parameter values to a distance scoring function.

A skill that extracts content from a file within the enrichment pipeline.

A skill that extracts content and layout information, via Azure AI Services, from files within the enrichment pipeline.

Controls the cardinality for chunking the content.

The type of the keysOrIds.

Generates n-grams of the given size(s) starting from the front or the back of an input token. This token filter is implemented using Apache Lucene.

Generates n-grams of the given size(s) starting from the front or the back of an input token. This token filter is implemented using Apache Lucene.

Tokenizes the input from an edge into n-grams of the given size(s). This tokenizer is implemented using Apache Lucene.

Removes elisions. For example, "l'avion" (the plane) will be converted to "avion" (plane). This token filter is implemented using Apache Lucene.

Using the Text Analytics API, extracts linked entities from text.

Using the Text Analytics API, extracts entities of different types from text.

Contains configuration options specific to the exhaustive KNN algorithm used during querying, which will perform brute-force search across the entire vector index.

Contains the parameters specific to exhaustive KNN algorithm.

Defines a mapping between a field in a data source and a target field in an index.

Represents a function that transforms a value from a data source before indexing.

Defines a function that boosts scores based on the value of a date-time field.

Provides parameter values to a freshness scoring function.

Statistics for a given index. Statistics are collected periodically and are not guaranteed to always be up-to-date.

Defines a data change detection policy that captures changes based on the value of a high water mark column.

Contains configuration options specific to the HNSW approximate nearest neighbors algorithm used during indexing and querying. The HNSW algorithm offers a tunable trade-off between search speed and accuracy.

Contains the parameters specific to the HNSW algorithm.

A skill that analyzes image files. It extracts a rich set of visual features based on the image content.

Configuration for OneLake knowledge source.

Parameters for OneLake knowledge source.

Represents the result of an individual indexer execution.

Request body for resync indexer operation.

Represents parameters for indexer execution.

A dictionary of indexer-specific configuration properties. Each name is the name of a specific property. Each value must be of a primitive type.

Represents a schedule for indexer execution.

Input field mapping for a skill.

A token filter that only keeps tokens with text contained in a specified list of words. This token filter is implemented using Apache Lucene.

A skill that uses text analytics for key phrase extraction.

Marks terms as keywords. This token filter is implemented using Apache Lucene.

Emits the entire input as a single token. This tokenizer is implemented using Apache Lucene.

Emits the entire input as a single token. This tokenizer is implemented using Apache Lucene.

Represents a knowledge base definition.

Specifies the Azure OpenAI resource used to do query planning.

Specifies the connection parameters for the model to use for query planning.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: KnowledgeBaseAzureOpenAIModel

Represents a knowledge source definition.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: AzureBlobKnowledgeSource, IndexedOneLakeKnowledgeSource, SearchIndexKnowledgeSource, WebKnowledgeSource

Reference to a knowledge source.

A skill that detects the language of input text and reports a single language code for every document submitted on the request. The language code is paired with a score indicating the confidence of the analysis.

Removes words that are too long or too short. This token filter is implemented using Apache Lucene.

Base type for analyzers.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: CustomAnalyzer, PatternAnalyzer, LuceneStandardAnalyzer, StopAnalyzer

Base type for normalizers.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: CustomNormalizer

Base type for tokenizers.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: ClassicTokenizer, EdgeNGramTokenizer, KeywordTokenizer, KeywordTokenizerV2, MicrosoftLanguageStemmingTokenizer, MicrosoftLanguageTokenizer, NGramTokenizer, PathHierarchyTokenizerV2, PatternTokenizer, LuceneStandardTokenizer, LuceneStandardTokenizerV2, UaxUrlEmailTokenizer

Limits the number of tokens while indexing. This token filter is implemented using Apache Lucene.

Standard Apache Lucene analyzer; Composed of the standard tokenizer, lowercase filter and stop filter.

Breaks text following the Unicode Text Segmentation rules. This tokenizer is implemented using Apache Lucene.

Breaks text following the Unicode Text Segmentation rules. This tokenizer is implemented using Apache Lucene.

Defines a function that boosts scores based on the magnitude of a numeric field.

Provides parameter values to a magnitude scoring function.

A character filter that applies mappings defined with the mappings option. Matching is greedy (longest pattern matching at a given point wins). Replacement is allowed to be the empty string. This character filter is implemented using Apache Lucene.

A skill for merging two or more strings into a single unified string, with an optional user-defined delimiter separating each component part.

Divides text using language-specific rules and reduces words to their base forms.

Divides text using language-specific rules.

Generates n-grams of the given size(s). This token filter is implemented using Apache Lucene.

Generates n-grams of the given size(s). This token filter is implemented using Apache Lucene.

Tokenizes the input into n-grams of the given size(s). This tokenizer is implemented using Apache Lucene.

Defines a data deletion detection policy utilizing Azure Blob Storage's native soft delete feature for deletion detection.

A skill that extracts text from image files.

Output field mapping for a skill.

Using the Text Analytics API, extracts personal information from an input text and gives you the option of masking it.

Tokenizer for path-like hierarchies. This tokenizer is implemented using Apache Lucene.

Flexibly separates text into terms via a regular expression pattern. This analyzer is implemented using Apache Lucene.

Uses Java regexes to emit multiple tokens - one for each capture group in one or more patterns. This token filter is implemented using Apache Lucene.

A character filter that replaces characters in the input string. It uses a regular expression to identify character sequences to preserve and a replacement pattern to identify characters to replace. For example, given the input text "aa bb aa bb", pattern "(aa)\s+(bb)", and replacement "$1#$2", the result would be "aa#bb aa#bb". This character filter is implemented using Apache Lucene.

A character filter that replaces characters in the input string. It uses a regular expression to identify character sequences to preserve and a replacement pattern to identify characters to replace. For example, given the input text "aa bb aa bb", pattern "(aa)\s+(bb)", and replacement "$1#$2", the result would be "aa#bb aa#bb". This token filter is implemented using Apache Lucene.

Tokenizer that uses regex pattern matching to construct distinct tokens. This tokenizer is implemented using Apache Lucene.

Create tokens for phonetic matches. This token filter is implemented using Apache Lucene.

Contains the options for rescoring.

Represents a resource's usage and quota.

Contains configuration options specific to the scalar quantization compression method used during indexing and querying.

Contains the parameters specific to Scalar Quantization.

Base type for functions that can modify document scores during ranking.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: DistanceScoringFunction, FreshnessScoringFunction, MagnitudeScoringFunction, TagScoringFunction

Defines parameters for a search index that influence scoring in search queries.

Represents an index alias, which describes a mapping from the alias name to an index. The alias name can be used in place of the index name for supported operations.

Represents a field in an index definition, which describes the name, data type, and search behavior of a field.

This class adds backward compatibility support for the 'hidden' property, which is the inverse of 'retrievable'.

Represents a search index definition, which describes the fields and search behavior of an index.

Field reference for a search index.

Knowledge Source targeting a search index.

Parameters for search index knowledge source.

Represents an indexer.

Represents information about the entity (such as Azure SQL table or CosmosDB collection) that will be indexed.

Abstract base type for data identities.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: SearchIndexerDataNoneIdentity, SearchIndexerDataUserAssignedIdentity

Clears the identity property of a datasource.

Represents a datasource definition, which can be used to configure an indexer.

This class adds an additional overload to support passing connection_string directly instead of credentials.

Specifies the identity for a datasource to use.

Represents an item- or document-level indexing error.

Definition of additional projections to secondary search indexes.

Description for what data to store in the designated search index.

A dictionary of index projection-specific configuration properties. Each name is the name of a specific property. Each value must be of a primitive type.

Definition of additional projections to azure blob, table, or files, of enriched data.

Abstract class to share properties between concrete selectors.

Projection definition for what data to store in Azure Files.

Projection definition for what data to store in Azure Blob.

Container object for various projection selectors.

Abstract class to share properties between concrete selectors.

Description for what data to store in Azure Tables.

Represents the limits that can be applied to an indexer.

Base type for skills.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: ChatCompletionSkill, WebApiSkill, AzureOpenAIEmbeddingSkill, CustomEntityLookupSkill, KeyPhraseExtractionSkill, LanguageDetectionSkill, MergeSkill, PIIDetectionSkill, SplitSkill, TextTranslationSkill, EntityLinkingSkill, EntityRecognitionSkillV3, SentimentSkillV3, ConditionalSkill, ContentUnderstandingSkill, DocumentExtractionSkill, DocumentIntelligenceLayoutSkill, ShaperSkill, ImageAnalysisSkill, OcrSkill

A list of skills.

Represents the current status and execution history of an indexer.

Represents an item-level warning.

A customer-managed encryption key in Azure Key Vault. Keys that you create and manage can be used to encrypt or decrypt data-at-rest, such as indexes and synonym maps.

Represents service-level resource counters and quotas.

Represents various service level limits.

Response from a get service statistics request. If successful, it includes service level counters and limits.

Defines how the Suggest API should apply to a group of fields in the index.

Defines a specific configuration to be used in the context of semantic capabilities.

A field that is used as part of the semantic configuration.

Describes the title, content, and keywords fields to be used for semantic ranking, captions, highlights, and answers.

Defines parameters for a search index that influence semantic capabilities.

Using the Text Analytics API, evaluates unstructured text and for each record, provides sentiment labels (such as "negative", "neutral" and "positive") based on the highest confidence score found by the service at a sentence and document-level.

A skill for reshaping the outputs. It creates a complex type to support composite fields (also known as multipart fields).

Creates combinations of tokens as a single token. This token filter is implemented using Apache Lucene.

Base type for similarity algorithms. Similarity algorithms are used to calculate scores that tie queries to documents. The higher the score, the more relevant the document is to that specific query. Those scores are used to rank the search results.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: BM25SimilarityAlgorithm, ClassicSimilarityAlgorithm

The type of the skill names.

A filter that stems words using a Snowball-generated stemmer. This token filter is implemented using Apache Lucene.

Defines a data deletion detection policy that implements a soft-deletion strategy. It determines whether an item should be deleted based on the value of a designated 'soft delete' column.

A skill to split a string into chunks of text.

Defines a data change detection policy that captures changes using the Integrated Change Tracking feature of Azure SQL Database.

Provides the ability to override other stemming filters with custom dictionary-based stemming. Any dictionary-stemmed terms will be marked as keywords so that they will not be stemmed with stemmers down the chain. Must be placed before any stemming filters. This token filter is implemented using Apache Lucene. See http://lucene.apache.org/core/4_10_3/analyzers-common/org/apache/lucene/analysis/miscellaneous/StemmerOverrideFilter.html.

Language specific stemming filter. This token filter is implemented using Apache Lucene. See https://learn.microsoft.com/rest/api/searchservice/Custom-analyzers-in-Azure-Search#TokenFilters.

Divides text at non-letters; Applies the lowercase and stopword token filters. This analyzer is implemented using Apache Lucene.

Removes stop words from a token stream. This token filter is implemented using Apache Lucene. See http://lucene.apache.org/core/4_10_3/analyzers-common/org/apache/lucene/analysis/core/StopFilter.html.

Represents a synonym map definition.

Matches single or multi-word synonyms in a token stream. This token filter is implemented using Apache Lucene.

Defines a function that boosts scores of documents with string values matching a given list of tags.

Provides parameter values to a tag scoring function.

A skill to translate text from one language to another.

Defines weights on index fields for which matches should boost scoring in search queries.

Base type for token filters.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: AsciiFoldingTokenFilter, CjkBigramTokenFilter, CommonGramTokenFilter, DictionaryDecompounderTokenFilter, EdgeNGramTokenFilter, EdgeNGramTokenFilterV2, ElisionTokenFilter, KeepTokenFilter, KeywordMarkerTokenFilter, LengthTokenFilter, LimitTokenFilter, NGramTokenFilter, NGramTokenFilterV2, PatternCaptureTokenFilter, PatternReplaceTokenFilter, PhoneticTokenFilter, ShingleTokenFilter, SnowballTokenFilter, StemmerOverrideTokenFilter, StemmerTokenFilter, StopwordsTokenFilter, SynonymTokenFilter, TruncateTokenFilter, UniqueTokenFilter, WordDelimiterTokenFilter

Truncates the terms to a specific length. This token filter is implemented using Apache Lucene.

Tokenizes urls and emails as one token. This tokenizer is implemented using Apache Lucene.

Filters out tokens with same text as the previous token. This token filter is implemented using Apache Lucene.

Contains configuration options related to vector search.

Contains configuration options specific to the algorithm used during indexing or querying.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: ExhaustiveKnnAlgorithmConfiguration, HnswAlgorithmConfiguration

Contains configuration options specific to the compression method used during indexing or querying.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: BinaryQuantizationCompression, ScalarQuantizationCompression

Defines a combination of configurations to use with vector search.

Specifies the vectorization method to be used during query time.

You probably want to use the sub-classes and not this class directly. Known sub-classes are: AzureMachineLearningVectorizer, AzureOpenAIVectorizer, WebApiVectorizer

A dictionary of http request headers.

A skill that can call a Web API endpoint, allowing you to extend a skillset by having it call your custom code.

Specifies a user-defined vectorizer for generating the vector embedding of a query string. Integration of an external vectorizer is achieved using the custom Web API interface of a skillset.

Specifies the properties for connecting to a user-defined vectorizer.

Knowledge Source targeting web results.

Configuration for web knowledge source domain.

Domain allow/block configuration for web knowledge source.

Parameters for web knowledge source.

Splits words into subwords and performs optional transformations on subword groups. This token filter is implemented using Apache Lucene.

The name of the embedding model from the Azure AI Foundry Catalog that will be called.

The Azure Open AI model name that will be called.

Specifies the data to extract from Azure blob storage and tells the indexer which data to extract from image content when "imageAction" is set to a value other than "none". This applies to embedded image content in a .PDF or other application, or image files such as .jpg and .png, in Azure blobs.

Determines how to process embedded images and image files in Azure blob storage. Setting the "imageAction" configuration to any value other than "none" requires that a skillset also be attached to that indexer.

Determines algorithm for text extraction from PDF files in Azure blob storage.

Represents the parsing mode for indexing from an Azure blob data source.

Defines the names of all character filters supported by the search engine.

Specifies how 'extraParameters' should be handled by Azure AI Foundry. Defaults to 'error'.

Specifies how the LLM should format the response.

Scripts that can be ignored by CjkBigramTokenFilter.

Controls the cardinality of the chunk unit. Default is 'characters'.

Controls the cardinality of the content extracted from the document by the skill.

The language codes supported for input text by CustomEntityLookupSkill.

Controls the cardinality of the chunk unit. Default is 'characters'.

Controls the cardinality of the content extracted from the document by the skill.

The depth of headers in the markdown output. Default is h6.

Controls the cardinality of the output format. Default is 'markdown'.

Controls the cardinality of the output produced by the skill. Default is 'oneToMany'.

Specifies which side of the input an n-gram should be generated from.

A string indicating what entity categories to return.

The language codes supported for input text by EntityRecognitionSkill.

The language codes supported for input by ImageAnalysisSkill.

A string indicating which domain-specific details to return.

Defines behavior of the index projections in relation to the rest of the indexer.

Specifies the environment in which the indexer should execute.

Represents the status of an individual indexer execution.

Options with various types of permission data to index.

Represents the overall indexer status.

The language codes supported for input text by KeyPhraseExtractionSkill.

The AI model to be used for query planning.

Optional content extraction mode. Default is 'minimal'.

The kind of the knowledge source.

The current synchronization status of the knowledge source.

Defines the names of all text analyzers supported by the search engine.

Defines the names of all text normalizers supported by the search engine.

Defines the names of all tokenizers supported by the search engine.

Specifies the max header depth that will be considered while grouping markdown content. Default is h6.

Specifies the submode that will determine whether a markdown file will be parsed into exactly one search document or multiple search documents. Default is oneToMany.

Lists the languages supported by the Microsoft language stemming tokenizer.

Lists the languages supported by the Microsoft language tokenizer.

Defines the sequence of characters to use between the lines of text recognized by the OCR skill. The default value is "space".

The language codes supported for input by OcrSkill.

A string indicating what maskingMode to use to mask the personal information detected in the input text.

Identifies the type of phonetic encoder to use with a PhoneticTokenFilter.

Represents score to use for sort order of documents.

Defines a regular expression flag that can be used in the pattern analyzer and pattern tokenizer.

Defines the aggregation function used to combine the results of all the scoring functions in a scoring profile.

Defines the function used to interpolate score boosting across a range of documents.

Defines the data type of a field in a search index.

Defines the type of a datasource.

The language codes supported for input text by SentimentSkill.

The language to use for a Snowball token filter.

The language codes supported for input text by SplitSkill.

The language to use for a stemmer token filter.

Identifies a predefined list of language-specific stopwords.

A value indicating which split mode to perform.

The language codes supported for input text by TextTranslationSkill.

Represents classes of characters on which a token filter can operate.

Defines the names of all token filters supported by the search engine.

The encoding format for interpreting vector field contents.

The algorithm used for indexing and querying.

The similarity metric to use for vector comparisons. It is recommended to choose the same similarity metric as the embedding model was trained on.

The compression method used for indexing and querying.

The storage method for the original full-precision vectors used for rescoring and internal index operations.

The quantized data type of compressed vector values.

The vectorization method to be used during query time.

The strings indicating what visual feature types to return.