Azure AI Search for TypeScript
Build TypeScript search apps with vector, hybrid, and semantic retrieval patterns.
The skill shows Entra ID, managed identity, and environment-variable authentication, and recommends document-level security filters. It does not clearly specify least-privilege RBAC, secret handling, data flows, user confirmation, external write effects, or rollback, so points are deducted.
It covers core TypeScript workflows for indexes, uploads, full-text, vector, hybrid, and semantic search. However, it has no skill-specific tests, input validation, or diagnosable failure handling, and the main document and references may differ on semantic-configuration field syntax; static scoring therefore stays at or below 10.
The target SDK, TypeScript audience, and major invocation scenarios are reasonably clear. Non-fit boundaries, service-version prerequisites, input/output contracts, regional reachability, and Chinese-language considerations are not stated, so points are deducted.
The documentation is readable and layered, with installation, authentication, core examples, best practices, MIT licensing, author, version 1.0.0, and two references. It lacks a changelog, explicit maintenance owner/update path, and sufficiently detailed limitations and troubleshooting.
The examples provide a useful starting point for search, indexing, and semantic retrieval and cover several common patterns. Embedding generation, existing-index state, service configuration, permissions, and production error handling remain for the user to complete, and some API details require verification, so only moderate static effectiveness is awarded.
The references point to Azure SDK for JS and Microsoft Learn sources, and the repository shows general CI and test infrastructure. There are no skill-specific auditable tests or execution results, so evidence coverage is limited.
- Examples include external writes such as index creation, document upload, and deletion, but provide no confirmation, permission scoping, or recovery steps; review RBAC, secret storage, filter security boundaries, and rollback before production use.
- Semantic-configuration field naming appears potentially inconsistent across references, and SDK/API version prerequisites are not pinned; compile-check each example against the target package version and official documentation.
- Core functionality depends on Azure AI Search, Entra ID, and potentially embedding/semantic services; mainland-China reachability and regional limitations are not documented.
What does this skill do, and when should you use it?
This skill helps coding agents use the Azure AI Search SDK for JavaScript and TypeScript. It covers index creation and management, document indexing, full-text, vector, hybrid, and semantic search, plus filters, facets, autocomplete, suggestions, and batch operations. It also demonstrates DefaultAzureCredential and managed identity authentication. It is a focused SDK implementation guide rather than a complete application or infrastructure package.
It shows how to install @azure/search-documents and @azure/identity, configure Azure Search environment variables, create SearchClient and SearchIndexClient instances, define vector fields and HNSW profiles, upload and batch-process documents, run full-text, vector, hybrid, and semantic queries, apply filters and facets, and configure autocomplete and suggestions. The examples read endpoint, index-name, credential, and token-related environment variables and print search results, scores, facets, captions, or reranker scores. The embedding function remains application-provided.
- A TypeScript developer needs to create a product index with searchable text, filters, facets, and embeddings.
- A search application must combine keyword matching with vector similarity in one query.
- A natural-language search experience needs semantic ranking, extractive captions, or answers.
- A catalog application needs category or price filtering with facet counts.
- An agent application needs Azure AI Search patterns for knowledge-base or agentic retrieval scenarios.
What are this skill's strengths and limitations?
- Covers the core Azure AI Search TypeScript workflow from index definition through querying.
- Includes full-text, vector, hybrid, and semantic search examples.
- Shows Entra ID-oriented authentication with DefaultAzureCredential and managed identity.
- Includes filtering, facets, autocomplete, suggestions, and batch document operations.
- Released under the MIT license in a Microsoft Agent Skills repository.
- The vector examples require an application-supplied getEmbedding function.
- The material consists primarily of focused snippets and practices, not a complete runnable application.
- It does not document Azure resource provisioning, deployment, or service pricing.
- Semantic search requires a semantic configuration on the index first.
How do you install this skill?
Install the SDK dependencies with: npm install @azure/search-documents @azure/identity. To install the complete skill collection, the README provides: npx skills add microsoft/skills, followed by selecting skills in the wizard. The source does not document a command for selecting only azure-search-documents-ts.
How do you use this skill?
Place the skill where the coding agent can load it, then use a request such as: “Using @azure/search-documents, create a product vector index and implement full-text, vector, and hybrid search.” Configure AZURE_SEARCH_ENDPOINT and AZURE_SEARCH_INDEX_NAME. Add AZURE_SEARCH_ADMIN_KEY when using a key, or configure AZURE_TOKEN_CREDENTIALS=prod when using DefaultAzureCredential. Provide the embedding-generation function in the application.