index / Souprem/Relay
Relay
Evaluation-first, confidence-aware workflow engine for synthetic prior-authorization cases: narrow AI judgments with probabilities, deterministic policy gates, replay, a CI regression gate, and shadow mode.
Research and EvalsVerificationListedPython
Repository
Stars0
LanguagePython
LicenseMIT
Created2026-09-25
Last pushed2026-09-30
Officialno
Judgment
Genuine Jev project0.95
gate 0.5
Uses Jev at runtime0.84
Is a meta list0.03
Is a reimplementation0.07
Quality scales
Substance2.37 / 3
gate 0.5
Docs2.92 / 3
Novelty2.71 / 3
Composite score0.87
Category distribution
Research and Evals0.70
Applications0.20
Agent and Dev Tooling0.10
Learning0.00
SDKs and Clients0.00
Games and Simulation0.00
Other0.00
Integrations0.00
Assigned categoryResearch and Evals
Category probability0.70
Category confidence0.65
PatternVerification
Pattern confidence0.85
Curation
StatusListed
Reasongate passed
README pickno
Overridenone
Sources
- code:"typesafe_sdk"
Provenance
Modeljev-1.13.0
Question setv2
Judged at2026-09-30 11:54 UTC