llm-typesafe 0.1a0 brings Jev's yes/no, choice and score answer types to Simon Willison's llm CLI
Simon Willison·low signal
Simon Willison released an llm plugin for TypeSafe AI's Jev decision model on 09-22. `llm -m jev` returns typed JSON such as {"type": "noul", "noul": 0.99}, and `-o answer_type choice` or `score` takes a criteria JSON object for routing or rubric grading. That makes Jev scriptable from shell pipelines, for example `cat message.txt | llm -m jev`. The waitlist for API keys reportedly moves quickly.