Top Repos· Jev-Based· Eval· Last updated 2026-09-21

openlayer-ai / jevals

“Agent evals and guardrails in one request. Built on Jev, Kev and Laya.”

Jev-Based rank #75In Eval #15

• 17Stars
• —Stars Δ / 1d
• —Stars Δ / 7d
• —Stars Δ / 30d
• 0Forks
• 0Watchers

GitHub ↗Official docs ↗Back to ranking

TL;DR

openlayer-ai/jevals is an independent decision engine or interface clone. Useful to study. Not Jev, not RLCD, not a calibration transfer.

The author describes it as: “Agent evals and guardrails in one request. Built on Jev, Kev and Laya.” That is their claim, not a bench this site ran.

Who it’s for

• Researchers comparing a local or open decision engine to hosted Jev — never as a silent substitute
• Agent / harness authors who need a typed gate, route, or prune step
• Anyone who needs the unofficial line: jev.pro does not issue keys and is not TypeSafe

What’s inside

Top-level entries on GitHub at generation time: .github, .gitignore, CONTRIBUTING.md, LICENSE, README.md, SECURITY.md, examples, pyproject.toml, src, tests, uv.lock. Read those before you trust a one-line install.

How to use

1. Clone and follow the author’s Python toolchain. Do not invent a package name.

git clone https://github.com/openlayer-ai/jevals.git
cd jevals

2. If you compare to Jev, keep a processor label so replica rows never read as Jev scores.
3. Send a complete state and typed questions. Code owns control flow after the probabilities.

Hosted Jev still looks like a function call — wrapper or not:

{
  model: "jev-1.13.0",
  state: { source: "openlayer-ai/jevals", brief: "..." },
  questions: {
    gate: { type: "noul", instructions: "Should we act on this state?" },
    route: { type: "choice", instructions: "Pick one next system.", criteria: ["code", "review", "human"] },
  },
}

Try this

Label a side-by-side, do not merge floors

Compare openlayer-ai/jevals and hosted jev-1.13.0 on my 50 labeled tickets. Report agreement. Do not copy thresholds. Tag every row with processor=replica or processor=jev.

Schema-safe ≠ correct

On a case openlayer-ai/jevals gets wrong, write down the schema-legal answer it returned and the human label. Do not “fix” it by asking Jev to generate an explanation.

FAQ

Can I treat these scores as Jev?

No. Label them as replica or local-engine results. Calibration does not transfer.

Does TypeSafe endorse it?

Their public docs index does not list it as Jev. We do not either.