Microsoft Azure · AI Search

Knowledge for
Intelligent Agents

Empower your AI agents with enterprise knowledge retrieval. Connect data sources, build RAG pipelines, and deliver answers — not just search results.

0
Regions Worldwide
0
Compliance Certs
0
File Formats
02 — Core Capabilities
Built for Enterprise RAG

Click any card to explore the details of each capability.

🔗

Knowledge Sources

Connect to multiple data sources with minimal ETL, indexing, or infrastructure management.

Supports Azure Blob Storage, Cosmos DB, SQL Database, SharePoint, and public web sources. Automated indexers keep your search index synchronized with source data changes.
🧩

Knowledge Bases

One unified API for cross-source agentic RAG with multi-hop query planning.

Knowledge bases break down complex questions using multi-hop reasoning and synthesize answers across sources. Agents can query a single endpoint that federates across your entire data estate.
🔀

Hybrid Search

Combine keyword, vector, and semantic search in a single request for optimal results.

Executes both BM25 text ranking and HNSW / KNN vector similarity in parallel. Results are fused using Reciprocal Rank Fusion (RRF) to combine the best of both approaches.
🛡️

Enterprise Security

Encryption, secure auth, network isolation, and 100+ compliance certifications.

Data encrypted at rest and in transit. Supports RBAC, managed identity, private endpoints, and customer-managed keys. Compliant with SOC 2, HIPAA, ISO 27001, and more.
⚙️

Platform Integrations

SDKs, Microsoft Foundry, Copilot Studio, and open-source framework support.

Native SDKs for .NET, Python, Java, JavaScript. Integrates with LangChain, Semantic Kernel, and the Foundry Agent Service for end-to-end AI agent workflows.
📊

Data Import Wizard

Automate your RAG pipeline with built-in parsing, chunking, and embedding.

The portal wizard handles document cracking, text chunking, AI enrichment, and vectorization in a single flow. Supports PDF, DOCX, PPTX, XLSX, HTML, JSON, and image formats.
03 — RAG Pipeline
From Data to Answers

Click any step to understand how data flows through the retrieval-augmented generation pipeline.

STEP 01
📁

Ingest

Connect data sources

STEP 02
✂️

Chunk

Split documents

STEP 03
🧬

Embed

Generate vectors

STEP 04
🗄️

Index

Store in AI Search

STEP 05
🔍

Retrieve

Query & rank

STEP 06
💡

Generate

LLM synthesizes

04 — Comparison
Search Modes at a Glance

Understanding when to use each search approach for your RAG application.

Capability Full-Text Vector Hybrid Semantic
Exact keyword matching ✦ — ✦ ✦
Meaning-based similarity — ✦ ✦ ✦
Handles typos & synonyms ◐ ✦ ✦ ✦
SKU / barcode lookup ✦ — ✦ ◐
Natural language queries ◐ ✦ ✦ ✦
Multi-language support ✦ ✦ ✦ ✦
Best for RAG pipelines ◐ ✦ ✦ ✦
05 — FAQ
Frequently Asked Questions
What is Azure AI Search?+
Azure AI Search is an enterprise knowledge system that powers retrieval-augmented generation (RAG) applications and search engines. It combines vector, full-text, and hybrid search over indexed data, integrating with Azure OpenAI and Foundry models for end-to-end AI agent workflows.
What is RAG (Retrieval-Augmented Generation)?+
RAG is an AI technique that combines retrieval-based methods with generative models. It retrieves relevant information from external sources like databases and documents, then feeds that context to a large language model to produce accurate, grounded responses enriched with up-to-date information.
How does hybrid search work?+
Hybrid search executes both text search (BM25 keyword ranking) and vector search (semantic similarity) in a single request. Results are merged using Reciprocal Rank Fusion (RRF), combining keyword precision with semantic understanding for superior retrieval quality.
What data sources does it support?+
Built-in indexers support Azure Cosmos DB, Azure SQL Database, Azure Blob Storage, SharePoint, and public web sources. Azure Data Factory provides 80+ additional connectors. The Push API allows you to index data from any source in any format.
Is Azure AI Search a vector database?+
Azure AI Search supports vector storage and similarity search alongside keyword and hybrid search. It can function as a vector store for applications requiring long-term memory, knowledge bases, or grounding data, making it a versatile choice that goes beyond traditional vector databases.
What algorithms are used for search?+
For vector search, it uses HNSW (Hierarchical Navigable Small World) for approximate nearest neighbor search and exhaustive KNN. For text search, it employs BM25 relevance scoring. Semantic ranking uses deep learning models to rerank results for contextual relevance.