ReleaseBytes
databricks Databricks Blog ·

Databricks Adds SQL AI Functions for Integrated Data Warehouse Analytics

blogaidatabricksgaengineer
feature

Databricks has added new AI Functions that enable direct invocation of AI models within standard SQL queries in their data warehouses. This integration simplifies AI workloads by eliminating data movement to external environments, enhances security and governance through Unity Catalog, and provides unified billing. Data professionals can now perform tasks like document intelligence, sentiment analysis, and data translation natively in SQL. The functions include `ai_classify`, `ai_extract`, `ai_translate`, `ai_parse_document`, and the general `ai_query`, leveraging specialized models for cost-effective results.

  • →SQL-native AI Functions integrated into Databricks data warehouses
  • →Best practices for production use of Databricks AI Functions
Features (1) ›
  • SQL-native AI Functions integrated into Databricks data warehouses

    Databricks has integrated AI Functions directly into its SQL data warehouses, allowing users to invoke AI models via standard SQL queries. This capability brings AI processing to data, reducing data movement and leveraging Unity Catalog for governance, with specialized functions like `ai_classify`, `ai_extract`, `ai_translate`, `ai_parse_document`, and the general `ai_query`.

Notes (1) ›
  • Best practices for production use of Databricks AI Functions

    Recommendations for using AI Functions in production include tagging jobs for cost attribution, prioritizing task-specific functions, and requesting structured outputs. Users are advised to be intentional about model choice, sample data before scaling, and manage prompts as versioned code for better reliability.

Read the original announcement →

https://www.databricks.com/blog/using-aifunctions-your-data-warehouse-top-use-cases

Related releases