jev-suite
Four Java tools on one kernel that ask Jev whether something was delivered as required, with deterministic gates keeping the final say.
Overview
jev-suite is a set of four decision-quality apps sharing one Java kernel. jev-proof checks whether a sponsored video delivered the brief; jev-fit finds where a candidate falls short of job requirements; jev-fidelity checks whether an edit preserved the information in the original; jev-rental lists what must be confirmed with a listing agent before viewing. Each feeds Jev sanitised evidence and asks typed Score, Choice or Noul questions.
The shared jev-kit provides a typed TypeSafe HTTP client aware of rate limits and timeouts, an offline mock, a three-way NoulGate (confident pass, confident fail, everything else to human review) and a sanitiser for external text against prompt injection. Design rules: Jev judges and code decides, degrade toward caution when the model is unreachable, calibration runs refuse mock data, and checks fail closed without evidence. The README reports per-app calibration results, for example jev-proof 68 of 68 gated decisions correct with 12 designed abstentions and 0 of 10 injection flips.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Did the sponsored video actually deliver the brief? | — |
| Yes / no | Did the edit preserve this fact from the original? | — |
| Other | Where does this candidate fall short of the job requirements? | — |
How it works
- Gather and sanitise external evidence
- Ask Jev typed questions over the evidence
- NoulGate: confident pass, confident fail, or human review
- Code applies thresholds, vetoes and routing; unreachable model escalates
Strengths
- Fails closed and degrades toward review
- Built-in prompt-injection sanitiser with measured injection resistance
- Calibration runners that refuse mock data
- README reports high gated accuracy with designed abstentions
Caveats
- Calibration sets are small (50-60 samples per app)
- Requires JDK 17+, Maven and a TypeSafe key
Getting started
mvn package (JDK 17+; jev-proof also needs yt-dlp).At a glance
- Type
- Project
- Status
- beta
- Language
- Java
- Runtime
- JVM CLI
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Four decision-quality tools on Jev (TypeSafe System One): Jev answers…
wh000wh000/awesome-jev-liveFour decision-quality tools on Jev (TypeSafe System One): Jev answers structured questions, deterministic code keeps the final say.
hellogumbo/awesome-jevFour decision-quality apps on one kernel, each asking Jev a structured question about whether something was delivered as required, with deterministic code keeping the final say.
MrJev/awesome-jev
Frequently asked questions
What are the four apps?
jev-proof (sponsored video brief delivery), jev-fit (candidate gaps), jev-fidelity (edit preservation) and jev-rental (what to confirm before a viewing).
What happens when Jev is unsure?
The three-way gate sends anything that is not a confident pass or fail to human review.
How is injection handled?
External text passes through a sanitiser, and each app reports injection-flip tests with zero flips.
Related projects
DeepSeek Harness plugin where Jev prunes long tool output, screens fetched pages for injected instructions and gates completion claims.
Agent skillbetaTypeScriptExample of using Jev as a webhook guardrail inside the agentgateway proxy to inspect model requests and responses.
ProxyproductionRustLarge catalog of Claude Code agents, commands, hooks and MCPs, listed with community Jev mods for guardrails and subagent models.
CLIproductionJavaScriptx402 pre-payment risk-check provider combining sanctions lists, phishing feeds and simulation with Jev checks for injected instructions.
ProjectproductionAuditable Zig terminal agent pairing Codex for coding with Jev typed preflight and postflight checks under deterministic policy gates.
CLIproductionZigCisco's security scanner for agent skills, with a System One analyzer as an advisory tier that cannot emit findings or change severity.
Agent skillproductionPython