Jev in production › Agent guardrails and approvals

JevGuard (SakshiKumari271)

Jev arbitrates streaming RAG output against retrieved context in a Streamlit harness; author claims sub-150ms execution per check (author).

Open on GitHub ↗

sub-150msmeasured, as published by the source
Use
Agent guardrails and approvals
Industry
AI infrastructure
Form
Open-source tool
Stage
Beta
Listed
2026-09-29
Found via
github
Repository
SakshiKumari271/JevGuard
Stars
2
Forks
0
Last push
2026-09-29
Language
Python
License
none stated

The README opens with

🛡️ JevGuard: Real-Time Arbitration Harness for Jev (TypeSafe AI)

JevGuard is a low-latency, real-time System-1 circuit-breaker runtime built explicitly around Jev — the specialized non-generative AI decision model developed by TypeSafe AI.

While conversational LLMs are designed to generate text (introducing 2–5 seconds of latency when used as evaluators), Jev is architected purely for software systems to output structured decision vectors and probability scores. JevGuard operationalizes Jev to arbitrate streaming RAG outputs against retrieved ground-truth context with sub-150ms execution times.

Badge

For the project's own README, linking back here:

Listed in Jev in production

Also used for agent guardrails and approvals