Build Azure Search Applications

This skill teaches an agent to build search applications with the Azure.Search.Documents .NET SDK, covering three clients: SearchClient for querying and uploading documents, SearchIndexClient for creating and managing indexes and synonym maps, and SearchIndexerClient for indexers, skillsets, and data sources. It defines search indexes either with a FieldBuilder reading attributes (SimpleField, SearchableField, VectorSearchField) off a model class, or with manually constructed field objects, and covers full-text, vector, semantic, and hybrid search - combining keyword search, vector search, and semantic ranking in one query. Vector fields are configured through a VectorSearch object referencing a VectorSearchProfile and an HNSW algorithm configuration, with the field's VectorSearchDimensions and VectorSearchProfileName set on the model.

Use it when writing .NET code against Azure AI Search: defining an index schema, uploading or updating documents, running full-text, vector, semantic, or hybrid queries, or building autocomplete and suggestions. It is not a guide to standing up the Azure AI Search service itself or configuring indexers and skillsets in depth beyond client usage - those get their own reference files. Prefer DefaultAzureCredential over an API key for production authentication, prefer FieldBuilder over hand-built field lists for maintainability, call CreateOrUpdateIndexAsync for idempotent index setup, batch document writes with IndexDocumentsBatch rather than one call per document, use Select to limit returned fields, and combine vector, keyword, and semantic search in one query for the best relevance.

Inputs: SEARCH_ENDPOINT and SEARCH_INDEX_NAME as environment variables (plus SEARCH_API_KEY only if not using Azure AD auth), a credential (DefaultAzureCredential or AzureKeyCredential), a SearchIndex definition with Fields and optional VectorSearch profiles and algorithms, and per-call SearchOptions (filter, order-by, select, facets, vector queries, semantic configuration). A field can carry IsKey, IsFilterable, IsSortable, IsFacetable, IsHidden, and AnalyzerName attributes to control how it is indexed and returned. Outputs: SearchResults<T> with a total count and an async-enumerable of scored SearchResult<T> documents, facet counts, autocomplete and suggestion results, semantic answers and captions read off SemanticSearch.Answers and .Captions, and RequestFailedExceptions carrying a status code, such as 404 for a missing index, handled by branching on the status code.

Azure AI Search (full-text, vector, semantic, and hybrid search, indexers, skillsets, synonym maps), Azure Identity (DefaultAzureCredential), and vector search algorithms and profiles such as HNSW (HnswAlgorithmConfiguration) for embedding fields queried with VectorizedQuery and its KNearestNeighborsCount. Document writes go through UploadDocumentsAsync, MergeDocumentsAsync, MergeOrUploadDocumentsAsync, or DeleteDocumentsAsync, and can be batched with IndexDocumentsBatch and IndexDocumentsAction. Semantic search is configured through SemanticSearchOptions with a QueryCaption and QueryAnswer setting. Bundled reference files, references/vector-search.md and references/semantic-search.md, cover those two areas in more depth than the main skill body.

dotnet add package Azure.Search.Documents
dotnet add package Azure.Identity

.NET developers building search experiences on Azure AI Search who need type-safe index definitions, document CRUD, and full-text, vector, semantic, or hybrid query patterns in one SDK.