jev-align (Sutro)
evaluates CSV, Parquet, and JSONL rows with Jev `Choice`, `Score`, or `Boolean` decisions, sends ambiguous and audit samples to a human, and uses accepted human labels to optimize the saved definition with GEPA.
何を判断するか
CHOICE
Picks one option from a fixed set
SCORE
Rates against ordered levels
NOUL
Answers a yes / no question
Inferred from this project's own one-line summary in the community list — it reads as a SCORE + NOUL decision. Read the source to confirm.
SCORE 呼び出しの形
この質問型の一般的な骨格であって、このプロジェクトの実際のコードではない。
from typesafe import TypeSafe
ts = TypeSafe()
result = ts.evaluate(
state=candidate,
questions={
"relevance": {
"type": "score",
"levels": ["low", "medium", "high"],
"instructions": "How relevant is this to the query?",
}
},
)
if result["relevance"] >= 0.7:
keep(candidate)データのラベリングと整備の他のプロジェクト
jev-curatesifts synthetic JSONL and Parquet rows using Jev Noul checks and calibrated confidence scores, streaming passed records and rejections straight to disk.typeful-triagemultiplayer triage dashboard where Jev answers a fixed set of typed questions per issue — kind, severity, urgency, duplicate, and next step — and every human correction is kept and shown back to the model on later runs.jlinklinks records under a plain-English match rule using Jev Noul pair judgments, with local candidate blocking and match resolution.jgrepfilters text, structured records, functions, and diff hunks against plain-English descriptions using Jev Noul judgments.