Azure AI Vector Search

 Vector Search Overview - Azure AI Search | Microsoft Learn

What it is: An information retrieval approach in Azure AI Search that indexes and queries numeric vector representations of content, allowing matching based on conceptual/semantic similarity, multiple languages, or different data types (like text and images).

Key Scenarios Supported:

Similarity search: Retrieves documents using vector-encoded queries.

Hybrid search: Runs vector search and keyword search in parallel and merges the results.

Multimodal search: Queries across embedding spaces containing both text and images (e.g., using OpenAI CLIP).

Multilingual search: Matches concepts across different languages.

Filtered vector search: Combines vector queries with metadata filters on text or numeric fields.

Vector database: Acts as a pure vector index for RAG architectures and AI apps.

How It Works:

Indexing: Uses a nearest neighbors algorithm to organize vectors into vector indexes. Embeddings can be generated internally via integrated vectorization or externally and pushed directly.

Querying: Converts user input into a vector and retrieves the nearest neighbors ($k$NN) from the index.

Pricing & Availability: Available in all regions and on all tiers at no extra charge (though generating embeddings via model providers like Azure OpenAI may incur separate fees).