Jev by TypeSafe AI

Jev is TypeSafe AI’s model for making structured decisions inside software. A developer provides a state and asks typed questions; Jev returns values and probability distributions that code can use directly, rather than a block of generated prose that needs to be parsed. Some question types also return confidence information that can help an application decide whether to act or request review. The available question types include choosing among options, scoring a state against a rubric, and judging whether a statement is true. Multiple questions can be sent in one API call and evaluated independently against the same state. TypeSafe recommends keeping each question focused, then combining results in application code when a workflow needs to weigh several factors. That approach lets developers adjust decision logic without rewriting a large prompt. Jev is aimed at builders adding classification, scoring, or routing decisions to software and automation workflows. Its outputs are designed for branching and sorting in code, with confidence signals available for handling uncertain cases. TypeSafe offers enterprise conversations for teams that need higher limits or specific data requirements.