Jev directory
Organization repository on GitHub
Agent evals and guardrails as Jev decisions: one request per trace, a fraction of a cent, fast enough for the agent loop. Runs locally with Kev or Laya.
Repo stats from the GitHub API, cached Oct 2026.
Evals and guardrails for agents, using Jev-style decision models instead of an LLM judge. All the evals for a trace go out as one request that costs a few thousandths of a cent and comes back in a few hundred milliseconds, so you can run them on every trace and inside the agent loop. Works with Jev through the TypeSafe or Vercel APIs, with Kev or Laya running locally on a Mac, with Eikos on your own GPU, or with a regular chat LLM if that's all you have (slower, costs more).
Quoted from the project's README. Read the full README.
from jevals import evaluate
from jevals.agent import Grounded, StayedInScope
from jevals.quality import AnswerRelevancy, Completeness
report = evaluate(
{"messages": messages, "tools": tools},
[Grounded(), StayedInScope(), AnswerRelevancy(), Completeness()],
)Watch out
Alpha software: the TypeSafe direct backend has only been tested against a mock and the adapters against fakes, and probabilities differ per backend, so calibrate thresholds yourself.
Short answers drawn from the linked project or page.
Posts, threads, and videos about this entry from around the web.
X: Agent Handoff Gate: agents verify what they hand off
X: Jev as an LLM guardrail inside Agentgateway
X: Vercel's fx safety reviewer, measured at 18× faster
Reddit: Early experiments wiring Jev into Pi as a safety layer and model router.
Forks and copies of this project in the directory. This entry is the canonical one.
duberblock/jevalsTypeScript
AntonioCoppe/jev-harnessTypeScript
allebee/pytest-jevPython
somoore/interlockPython
anpicasso/hermes-jev-approvalsPython
Kevthetech143/super-jevPython
GiesN/typesafe-jev-workflowPythonOriginal write-ups on what this entry is for.
Posts from builders shipping with Jev right now.