Jev’s decision workflow now has an LLM command-line plugin

Simon Willison’s llm-typesafe plugin brings TypeSafe’s Jev decision model into the LLM command-line tool. Instead of requesting a prose answer, a script can ask for a yes/no decision, choose among supplied options or score an input against a rubric. The new integration makes it easier to pipe files into small classification and routing experiments.

Simon Willison · Sep 22, 2026

Why it matters

A concrete CLI follow-up to Jev’s launch and the evaluation concerns covered previously.

More context

Try version 0.1a0 with the LLM tool and a TypeSafe API key; the article points to a waitlist for access. Start with labeled examples and compare decisions against a simple baseline before putting the output into an automated workflow.

Keep in mind

The plugin is alpha and requires external model access. Structured answers can still be wrong, and usage terms or costs apply. 1min has read the examples, not executed them.

Original read