Microsoft has released new T-SQL AI functions for Azure SQL and SQL database in Microsoft Fabric to public preview. The features enable developers to generate embeddings from relational data using Azure OpenAI models and perform vector similarity searches directly within SQL databases. Key capabilities include AI_GENERATE_EMBEDDINGS for creating vector representations of text data, VECTOR_DISTANCE for similarity searches using cosine distance, and AI_GENERATE_CHUNKS for breaking large text into manageable pieces. The implementation requires setting up database scoped credentials, creating external models pointing to Azure OpenAI endpoints, and adding vector data type columns to existing tables.
Table of contents
Prerequisites Copy linkSet up your environment Copy linkSet up the database Copy linkUse VECTOR_DISTANCE Copy linkSort: