AI Gateway Adds TypeSafe Client and HTTP API Support for Jev Models

AI Gateway has expanded the invocation methods for Jev, a probabilistic decision-making model. In addition to existing AI SDK support, users can now call Jev directly using existing TypeSafe clients or via the HTTP API.

Integration Methods and Features

• TypeSafe Client: Integration is possible by simply changing the Base URL and API Key, maintaining existing evaluation call code and response structures.
• HTTP API: Jev can be called by sending POST requests to the /v1/evaluate endpoint, regardless of language or framework.
• AI SDK: Evaluations can be executed identically in TypeScript applications via the evaluate function.

Jev returns typed answers with probabilities for input states without text generation. There are three request types: boolean (0–1 probability), choice (selection from specified options), and score (score on a defined scale). All calls are billed through AI Gateway and integrated into usage and observability data alongside other model calls.

Integration with Agent Framework 'eve'

eve, a durable agent building framework, uses Jev as its default evaluation model. This enables features such as automatic model selection, typed evaluation, and automatic tool approval.