Site directory

Community projects by category.

Each entry names the decision and its evidence.

The next batch of coverage cases, each day.

Public posts, each with its source.

Notes on model and site updates.

Choice, Noul, and Score together.

Routing, fan-out, and composite scoring.

HTTP, Python, and JavaScript.

Send a state and read the answer.

Tetris, six rapid judgments, and a sort. Copy the call and run it again.

Support triage, release review, compliance routing — edit and run on the spot.

clavue-jev and jev-1.13.0 answer the same question.

Judgment examples you can edit.

Input is $0.042 per million tokens. Output is free.

Call POST /v1/judge with a key.

What this model is for.

The smallest call that runs.

Written here: a state plus one typed question. The model is clavue-jev.

The text you send in.

How sure the answer is, and when to act on it.

Pick one mutually exclusive option.

A yes/no question, returned as the probability of yes.

Rate along a scale.

One Choice call dispatches the request.

Act when confidence is high. Hand off when it is low.

Ask several questions of one state.

Fold several scores into one ranking.

The request shape without an SDK.

Call a judgment from Python.

Call a judgment from JavaScript.

Which primitive people use for which job.

A yes/no check before an action runs.

Pick which function to call.

Do not act when the class is unsure.

Find the passages that match.

Ask the whole set in one call.

Coarse class first, then the fine one.

Whether a sentence is supported by the source.

Whether two names are the same thing.

Choose which skill fits the moment.