What's new in SQL Server 2025 and Azure SQL Databases
summary
SQL Server 2025 and Azure SQL now store and search embeddings natively: a new vector data type (examples: 768 and 1,536 dimensions) replaces million-column workarounds, turning 17,000 products into about 13 million floats queried with a built-in vector distance function or fast vector index (ANN/KNN). Demos use LM Studio and Azure endpoints to generate embeddings, show REST calls from SQL, a true JSON column with indexes and modify, regex and fuzzy (Jaro–Winkler) matching, chunking for large docs, change-event streaming to Event Hubs/Fabric, and SSMS Copilot for ad hoc SQL.
AI-written summary · about
chapters (20)
- 0:00Introduction to SQL 2025 updates
- 1:13SQL understanding data meaning
- 4:47Graph databases recap
- 5:09Vector databases and RAG explained
- 10:30Manual vector embedding demo
- 21:38SQL 2025 vector data type
- 24:32SQL 2025 AI embedding generation
- 27:10External model setup in SQL
- 30:25Vector index limitations
- 31:08Azure SQL vector search
- 34:45AI generated chunks feature
- 37:43SQL REST API calls
- 41:11JSON type and performance
- 46:14Regex and data extraction
- 51:00Regex patterns and masking
- 53:05Fuzzy search and similarity
- 57:30Hybrid search capabilities
- 57:52Change event streaming
- 58:57SQL Copilot features
- 1:00:00Closing and Q&A
