Build Azure Search Applications

This skill builds search applications with full-text, vector, semantic, and hybrid search using Azure.Search.Documents (.NET, stable v11.7.0), authenticated via DefaultAzureCredential (preferred) or an AzureKeyCredential API key. It distinguishes three client types: SearchClient for querying indexes and uploading/updating/deleting documents, SearchIndexClient for creating and managing indexes and synonym maps, and SearchIndexerClient for managing indexers, skillsets, and data sources. Index creation is shown two ways: the recommended FieldBuilder approach, decorating a POCO's properties with [SimpleField] (filterable/sortable/facetable, non-searchable), [SearchableField] (full-text searchable, with an optional analyzer like LexicalAnalyzerName.EnLucene), and [VectorSearchField] (a ReadOnlyMemory<float> embedding field with a declared dimension count and vector-search profile) - or manual field definition building the same SearchField objects directly, both requiring a VectorSearch configuration with a named profile and an HNSW algorithm. Document operations cover UploadDocumentsAsync (add new), MergeDocumentsAsync (update existing), MergeOrUploadDocumentsAsync (upsert), DeleteDocumentsAsync by key, and batched mixed operations via IndexDocumentsBatch/IndexDocumentsAction. Search patterns cover basic filtered/sorted/paged/projected search with SearchOptions (Filter, OrderBy, Select, Size, Skip, IncludeTotalCount), faceted search (aggregating result counts per field value), autocomplete and fuzzy-matched suggestions via a named suggester, pure vector search (a VectorizedQuery with a k-nearest-neighbors count against a named vector field), semantic search (a named semantic configuration producing extractive captions and answers alongside ranked results), and hybrid search combining vector, keyword, and semantic ranking in a single query for the best relevance - with detailed vector-search and semantic-search patterns deferred to companion reference files. A field-attribute reference table covers SimpleField/SearchableField/VectorSearchField and modifiers (IsKey, IsFilterable, IsSortable, IsFacetable, IsHidden, AnalyzerName). Error handling distinguishes a 404 (index not found) via RequestFailedException. Best practices: prefer DefaultAzureCredential over API keys in production, use FieldBuilder with model attributes for type-safe index definitions, use CreateOrUpdateIndexAsync for idempotent index creation, batch document operations for throughput, project only needed fields with Select, configure semantic search for natural-language queries, and combine vector, keyword, and semantic search for the best relevance.

Use this skill when building search functionality in a .NET application on Azure AI Search - full-text search with filters/facets, vector/embedding-based semantic similarity search, natural-language semantic ranking, or a hybrid combination of all three for best relevance. It is not an indexer/skillset orchestration deep-dive (that's SearchIndexerClient's domain, only named here) and detailed vector/semantic search implementation patterns live in companion reference files rather than fully inline in this overview. Connection details are read from SEARCH_ENDPOINT, SEARCH_INDEX_NAME, and, for API-key auth, SEARCH_API_KEY.

Input is documents to index (typed POCOs or raw field data) and search queries (keyword text, filters, vector embeddings, or a combination). Output is ranked search results with scores, optional facet counts, autocomplete/suggestion candidates, and for semantic search, extractive answers and captions.

Built on Azure AI Search via Azure.Search.Documents, authenticated with Azure Identity's DefaultAzureCredential or an API key, with SearchIndexerClient for indexer/skillset/data-source orchestration.

dotnet add package Azure.Search.Documents

.NET developers building search experiences - full-text, vector/embedding, semantic, or hybrid - on Azure AI Search who need type-safe index definitions and production-ready query patterns.