{"id":4329,"date":"2025-03-19T07:53:01","date_gmt":"2025-03-19T14:53:01","guid":{"rendered":"https:\/\/devblogs.microsoft.com\/azure-sql\/?p=4329"},"modified":"2025-03-19T07:59:53","modified_gmt":"2025-03-19T14:59:53","slug":"vector-search-with-azure-sql-semantic-kernel-and-entity-framework-core","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/azure-sql\/vector-search-with-azure-sql-semantic-kernel-and-entity-framework-core\/","title":{"rendered":"Vector Search with Azure SQL, Semantic Kernel and Entity Framework Core"},"content":{"rendered":"

<\/i>Guest Post<\/strong><\/p>Marco Minerva has been using .NET since its first introduction. Today he works as a freelance consultant and is involved in designing and developing solutions for the Microsoft ecosystem. He is a speaker at technical events, writes books and articles about developing and makes live streaming about coding on Twitch. He is a Microsoft MVP since 2013. Thanks Marco! <\/div><\/p>\n

Vector databases like Qdrant and Milvus are specifically designed to efficiently store, manage, and retrieve embeddings. However, many applications already use relational databases like SQL Server or SQL Azure. In such cases, installing and managing another database can be challenging, especially since these vector databases may not offer all the features of a modern relational database like Azure SQL or SQL Server.<\/p>\n

This is where the new vector data type in SQL Azure becomes extremely useful. It allows you to create Retrieval-Augmented Generation (RAG) solutions by storing embeddings in the same place as your other data, eliminating the need for an external vector database.<\/p>\n

Database Vector Search Sample App<\/h2>\n

For a practical example, check out the GitHub repository https:\/\/github.com\/marcominerva\/SqlDatabaseVectorSearch<\/a> showcases how to use the native vector type in Azure SQL Database to perform embeddings and RAG with Azure OpenAI.<\/p>\n

\"SQL<\/p>\n

The application enables users to load documents, generate embeddings, and save them into the database as vectors. It also supports performing searches using Vector Search and Retrieval-Augmented Generation (RAG). Currently, the application supports PDF, DOCX, TXT, and MD file formats. Vectors are saved and retrieved using Entity Framework Core with the EFCore.SqlServer.VectorSearch<\/a> library. Additionally, Embedding and Chat Completion functionalities are integrated with Semantic Kernel<\/a>.<\/p>\n

This repository includes a Blazor Web App and a Minimal API, which allow for programmatic interaction with embeddings and RAG.<\/p>\n

Supported features<\/strong><\/h3>\n