AI Gateway now supports TypeSafe clients and an HTTP API for Jev

Vercel AI Gateway has been extended to accept calls from TypeSafe clients as well as a native HTTP endpoint that reaches the Jev probabilistic decision model. The update lets existing TypeSafe code point at the gateway by changing only the base URL and API key, while the HTTP API can be invoked from any language by posting a JSON payload to the /v1/evaluate path. The same model is also reachable through the AI SDK for TypeScript applications, so developers have three interchangeable ways to obtain typed answers with attached probabilities. Jev supplies deterministic‑looking results for software‑related decisions, returning a probability rather than free‑form text, which simplifies downstream processing. Because every request is billed through AI Gateway, usage of Jev appears alongside other model calls, giving teams unified observability and cost tracking. The ability to reuse existing TypeSafe integrations without code changes reduces migration effort and accelerates adoption of probabilistic evaluation in existing workflows. It is not yet clear how performance and latency will compare across the three access methods, as Vercel has not published benchmark data. The long‑term pricing impact of adding Jev calls to existing AI Gateway spend remains uncertain, and Vercel has not detailed any plans for additional question types beyond boolean, choice and score.