Редагувати

Поділитися через


What's Azure AI Search?

Azure AI Search (formerly known as "Azure Cognitive Search") is an enterprise-ready search and retrieval system, with a comprehensive set of advanced search technology, built for high-performance applications at any scale.

Azure AI Search is the primary recommended retrieval system when building RAG-based applications on Azure, with native LLM integrations between Azure OpenAI Service and Azure Machine Learning.

Azure AI Search can be used in both traditional and GenAI scenarios. Common use cases include knowledge base insights (catalog or document search), information discovery (data exploration), retrieval-augmented generation (RAG), and automation.

When you create a search service, you work with the following capabilities:

Architecturally, a search service sits between the external data stores that contain your un-indexed data, and your client app that sends query requests to a search index and handles the response.

Azure AI Search architecture

In your client app, the search experience is defined using APIs from Azure AI Search, and can include relevance tuning, semantic ranking, autocomplete, synonym matching, fuzzy matching, pattern matching, filter, and sort.

Across the Azure platform, Azure AI Search can integrate with other Azure services in the form of indexers that automate data ingestion/retrieval from Azure data sources, and skillsets that incorporate consumable AI from Azure AI services, such as image and natural language processing, or custom AI that you create in Azure Machine Learning or wrap inside Azure Functions.

Inside a search service

On the search service itself, the two primary workloads are indexing and querying.

  • Indexing is an intake process that loads content into your search service and makes it searchable. Internally, inbound text is processed into tokens and stored in inverted indexes, and inbound vectors are stored in vector indexes. The document format that Azure AI Search can index is JSON. You can upload JSON documents that you've assembled, or use an indexer to retrieve and serialize your data into JSON.

    Applied AI through a skillset extends indexing with image and language models. If you have images or large unstructured text in source document, you can attach skills that perform OCR, analyze and describe images, infer structure, translate text and more. Output is text that can be serialized into JSON and ingested into a search index.

    Skillsets can also perform data chunking and vectorization during indexing. Skills that attach to Azure OpenAI, the model catalog in Azure AI Studio, or custom skills that attach to any external chunking and embedding model can be used during indexing to create vector data. Output is chunked vector content that can be ingested into a search index.

  • Querying can happen once an index is populated with searchable content, when your client app sends query requests to a search service and handles responses. All query execution is over a search index that you control.

    Semantic ranking is an extension of query execution. It adds secondary ranking, using language understanding to reevaluate a result set, promoting the most semantically relevant results to the top.

    Integrated vectorization is also an extension of query execution. If you have vector fields in your search index, you can submit raw vector queries or text that's vectorized at query time.

Azure AI Search is well suited for the following application scenarios:

  • Use it for traditional full text search and next-generation vector similarity search. Back your generative AI apps with information retrieval that leverages the strengths of both keyword and similarity search. Use both modalities to retrieve the most relevant results.

  • Consolidate heterogeneous content into a user-defined and populated search index composed of vectors and text. You maintain ownership and control over what's searchable.

  • Integrate data chunking and vectorization for generative AI and RAG apps.

  • Apply granular access control at the document level.

  • Offload indexing and query workloads onto a dedicated search service.

  • Easily implement search-related features: relevance tuning, faceted navigation, filters (including geo-spatial search), synonym mapping, and autocomplete.

  • Transform large undifferentiated text or image files, or application files stored in Azure Blob Storage or Azure Cosmos DB, into searchable chunks. This is achieved during indexing through AI skills that add external processing from Azure AI.

  • Add linguistic or custom text analysis. If you have non-English content, Azure AI Search supports both Lucene analyzers and Microsoft's natural language processors. You can also configure analyzers to achieve specialized processing of raw content, such as filtering out diacritics, or recognizing and preserving patterns in strings.

For more information about specific functionality, see Features of Azure AI Search

How to get started

Functionality is exposed through the Azure portal, simple REST APIs, or Azure SDKs like the Azure SDK for .NET. The Azure portal supports service administration and content management, with tools for prototyping and querying your indexes and skillsets.

Use the Azure portal

An end-to-end exploration of core search features can be accomplished in four steps:

  1. Decide on a tier and region. One free search service is allowed per subscription. All quickstarts can be completed on the free tier. For more capacity and capabilities, you'll need a billable tier.

  2. Create a search service in the Azure portal.

  3. Start with Import data wizard. Choose a built-in sample or a supported data source to create, load, and query an index in minutes.

  4. Finish with Search Explorer, using a portal client to query the search index you just created.

Use APIs

Alternatively, you can create, load, and query a search index in atomic steps:

  1. Create a search index using the portal, REST API, .NET SDK, or another SDK. The index schema defines the structure of searchable content.

  2. Upload content using the "push" model to push JSON documents from any source, or use the "pull" model (indexers) if your source data is of a supported type.

  3. Query an index using Search explorer in the portal, REST API, .NET SDK, or another SDK.

Use accelerators

Or, try solution accelerators:

  • Chat with your data solution accelerator helps you create a custom RAG solution over your content.

  • Conversational Knowledge Mining solution accelerator helps you create an interactive solution to extract actionable insights from post-contact center transcripts.

  • Build your own copilot solution accelerator, leverages Azure OpenAI Service, Azure AI Search and Microsoft Fabric, to create custom copilot solutions.

    • Generic copilot helps you build your own copilot to identify relevant documents, summarize unstructured information, and generate Word document templates using your own data.

    • Client Advisor all-in-one custom copilot empowers Client Advisor to harness the power of generative AI across both structured and unstructured data. Help our customers to optimize daily tasks and foster better interactions with more clients

    • Research Assistant helps build your own AI Assistant to identify relevant documents, summarize and categorize vast amounts of unstructured information, and accelerate the overall document review and content generation.

Tip

For help with complex or custom solutions, contact a partner with deep expertise in Azure AI Search technology.

Compare search options

Customers often ask how Azure AI Search compares with other search-related solutions. The following table summarizes key differences.

Compared to Key differences
Microsoft Search Microsoft Search is for Microsoft 365 authenticated users who need to query over content in SharePoint. Azure AI Search pulls in content across Azure and any JSON dataset.
Bing Bing APIs query the indexes on Bing.com for matching terms. Azure AI Search searches over indexes populated with your content. You control data ingestion and the schema.
Database search Azure SQL has full text search and vector search. Azure Cosmos DB also has text search and vector search. Azure AI Search becomes an attractive alternative when you need features like relevance tuning, or content from heterogeneous sources. Resource utilization is another inflection point. Indexing and queries are computationally intensive. Offloading search from the DBMS preserves system resources for transaction processing.
Dedicated search solution Assuming you've decided on dedicated search with full spectrum functionality, a final categorical comparison is between search technologies. Among cloud providers, Azure AI Search is strongest for vector, keyword, and hybrid workloads over content on Azure, for apps that rely primarily on search for both information retrieval and content navigation.

Key strengths include:

  • Support for vector and nonvector (text) indexing and queries. With vector similarity search, you can find information that’s semantically similar to search queries, even if the search terms aren’t exact matches. Use hybrid search for the best of keyword and vector search.
  • Ranking and relevance tuning through semantic ranking and scoring profiles. Query syntax supports term boosting and field prioritization.
  • Azure data integration (crawlers) at the indexing layer.
  • Azure AI integration for transformations that make content text and vector searchable.
  • Microsoft Entra security for trusted connections, and Azure Private Link for private connections in no-internet scenarios.
  • Full search experience: Linguistic and custom text analysis in 56 languages. Faceting, autocomplete queries and suggested results, and synonyms.
  • Azure scale, reliability, and global reach.