Positioning Jev as listed in the Cloudflare Workers AI model catalog: the model is served directly inside Workers AI rather than being reached through an external gateway, and it appears alongside the catalog's other models.
What it does Integration is a single env.AI.run() call that sends the state and questions directly to the model, with no separate gateway or API key to wire up. As a catalog model, it shares the same invocation path as the rest of Workers AI. The model page carries four complete request/response examples covering scenarios such as refund review, department routing, and account risk scoring, among others; each shows the full round trip from request body to typed answer, so the examples can serve as starting points rather than something to reverse-engineer. Each example shows the request body, the questions, and the typed answer for its scenario.
Characteristics Specs per the Cloudflare model page: a 32k context window, zero data retention, and input pricing of $0.042 per million tokens. One distinction to keep in mind: Cloudflare also maintains Clef, its own open-source decision model, which is covered in a separate entry in this collection — typesafe/jev on Workers AI is the Jev model as served by Cloudflare, not Clef.
When to use Fits applications already running inside the Cloudflare Workers ecosystem — a Worker calls the model the same way it calls the catalog's other models — and the worked examples double as references for putting the request format together. The access path is tied to Workers AI's invocation style, so deployments outside that environment need a different route to Jev.