This repository is for active development of the Azure SDK for JavaScript (NodeJS & Browser). For consumers of the SDK we recommend visiting our public developer docs at https://docs.microsoft.com/javascript/azure/ or our versioned developer docs at https://azure.github.io/azure-sdk-for-js.
Feature: Hierarchical Aggregation and Facet Filtering
API changes
On request, there are no API changes. Facets are exposed as Strings in the API and this feature is an update to the syntax used to allow for hierarchical aggregation and facet filtering. On response, FacetResult will now be a recursive data structure where in addition to the previous count: int64 there will now be a facets: Map<String, FacetResult>.
Caption and Answer length in Semantic Search
API changes
On request, this is an update to the custom classes added to support Semantic Search Answers and Captions to expose the configuration settings as strongly typed values rather than the pipe (|) delimited format the wire request uses. This will be an update to those classes to add maxcharlength: int32 and have those classes create the pipe delimited string correctly.
Runtime notes
This feature is only available where the query type is semantic.
Query rewriting for improved L1 retrieval
API changes
On request, this is an update to both VectorQuery and SearchRequest to support a new pipe delimited string that indicates how many additional queries the Search service will generate based on the given query to improve L1 retrieval. For this feature we should create a custom helper class similar to what was added for Semantic Search Answers and Captions that provides a strongly typed way to configure this feature. Name proposal is QueryRewrite with configuration values of type: QueryRewritesType and count: int32.
On response, this adds a new field for SemanticQueryRewriteResultType which is metadata about how the Search service ran the search request and updates to DebugInfo to include QueryRewritesDebugInfo to include debugging data about how the query rewrites modified the search request.
Runtime notes
This feature is only available where the query type is semantic.
Vector Search HNSW storage optimization
API changes
When configuring vector compressions there is a new configuration to determine how the Search service should manage the original vectors (the vector before compression). This is a new field on the configuration.
Runtime notes
If the vector compression is configured to discard the original vector this means the Search service is unable to rescore results with the original vector. This means the enablement of rescore cannot happen with discard is chosen. Additionally, the storing options for the original vectors cannot be changed after creation.
Markdown parsing mode
API changes
Two new settings for IndexingParametersConfiguration were added for MarkdownParsingSubmode and MarkdownHeaderDepth. Historically, indexing configurations was just additional properties, then typeness was added, this may require updates to pull these new values out of additional properties used in older versions.
Azure Document Intelligence skill
API changes
This is a new skill added for Document Intelligence. Should be simple regeneration with no manual changes.
AI Services bill to sub-domain
API changes
New subtypes AIServicesAccountKey and AIServicesAccountIdentity for CognitiveServicesAccount have been added which can tell the Search service which subdomain to bill. Should be a simple regeneration with no manual changes.
In addition to the new API changes, there is one Swagger fix that needs to be applied with a Swagger transform:
This issue outlines the work to be done for the Search 2024-11-01-Preview release.
These are the links to generate from:
https://raw.githubusercontent.com/Azure/azure-rest-api-specs/14531a7cf6101c1dd57e7c1c83103a047bb8f5bb/specification/search/data-plane/Azure.Search/preview/2024-11-01-preview/searchindex.json https://raw.githubusercontent.com/Azure/azure-rest-api-specs/14531a7cf6101c1dd57e7c1c83103a047bb8f5bb/specification/search/data-plane/Azure.Search/preview/2024-11-01-preview/searchservice.json
Feature: Hierarchical Aggregation and Facet Filtering
API changes
On request, there are no API changes. Facets are exposed as Strings in the API and this feature is an update to the syntax used to allow for hierarchical aggregation and facet filtering. On response,
FacetResultwill now be a recursive data structure where in addition to the previouscount: int64there will now be afacets: Map<String, FacetResult>.Caption and Answer length in Semantic Search
API changes
On request, this is an update to the custom classes added to support Semantic Search Answers and Captions to expose the configuration settings as strongly typed values rather than the pipe (
|) delimited format the wire request uses. This will be an update to those classes to addmaxcharlength: int32and have those classes create the pipe delimited string correctly.Runtime notes
This feature is only available where the query type is
semantic.Query rewriting for improved L1 retrieval
API changes
On request, this is an update to both
VectorQueryandSearchRequestto support a new pipe delimited string that indicates how many additional queries the Search service will generate based on the given query to improve L1 retrieval. For this feature we should create a custom helper class similar to what was added for Semantic Search Answers and Captions that provides a strongly typed way to configure this feature. Name proposal isQueryRewritewith configuration values oftype: QueryRewritesTypeandcount: int32.On response, this adds a new field for
SemanticQueryRewriteResultTypewhich is metadata about how the Search service ran the search request and updates toDebugInfoto includeQueryRewritesDebugInfoto include debugging data about how the query rewrites modified the search request.Runtime notes
This feature is only available where the query type is
semantic.Vector Search HNSW storage optimization
API changes
When configuring vector compressions there is a new configuration to determine how the Search service should manage the original vectors (the vector before compression). This is a new field on the configuration.
Runtime notes
If the vector compression is configured to discard the original vector this means the Search service is unable to rescore results with the original vector. This means the enablement of rescore cannot happen with discard is chosen. Additionally, the storing options for the original vectors cannot be changed after creation.
Markdown parsing mode
API changes
Two new settings for
IndexingParametersConfigurationwere added forMarkdownParsingSubmodeandMarkdownHeaderDepth. Historically, indexing configurations was just additional properties, then typeness was added, this may require updates to pull these new values out of additional properties used in older versions.Azure Document Intelligence skill
API changes
This is a new skill added for Document Intelligence. Should be simple regeneration with no manual changes.
AI Services bill to sub-domain
API changes
New subtypes
AIServicesAccountKeyandAIServicesAccountIdentityforCognitiveServicesAccounthave been added which can tell the Search service which subdomain to bill. Should be a simple regeneration with no manual changes.In addition to the new API changes, there is one Swagger fix that needs to be applied with a Swagger transform: