1 link tagged with all of: decision model + caching + snowflake + jevflake + structured prediction
Click any tag below to further narrow down your results
Links
Jevflake is a dbt package that lets you query Snowflake data using TypeSafe AI's Jev decision model, which returns typed answers (yes/no probabilities, category picks, or scored ratings) that you can filter and join like regular SQL. It handles API setup, creates SQL functions, and manages answer caching so you don't re-query the same data.
- Jev returns structured answers (probabilities for yes/no, picks with confidence for categories, weighted scores for ratings) rather than text, making answers usable directly in SQL operations
- The package stores answers in an incremental table keyed by row ID and question, so rows are only sent to the API again if their content changes or you modify the questions
- Setup splits between admin (network/integration) and dbt role (function creation), requiring Snowflake external access enabled and a TypeSafe API key