Using jev-cli

jev-cli asks Jev, a "System One" evaluation model, to judge one shared state against a set of typed questions. It returns structured JSON — a choice, a score, or a probability — never prose.

Reach for it when you need a decision your code can branch on. It is far faster and cheaper than a full LLM call (priced per input token, with zero output tokens), so checking ten properties of a diff, a log, or a support transcript is cheap.

Do not use it to write, summarize, translate, or rewrite anything. It cannot produce text.

Just run the evaluation. There is no setup check to perform first — eval fails fast with an actionable message and a distinct exit code when something is wrong, and jev-cli doctor is the diagnostic you reach for then, not before every call.