Jev in production › Moderation and content filtering

jev-edge

OpenResty, APISIX and edge-worker module in which Jev scores each incoming natural-language request for prompt injection or abuse and blocks it above a threshold, with caching and fail-open. A deployment context cut misses from 37% to 5% (author).

Open on GitHub ↗

from 37% to 5%measured against a baseline, as published by the source
Use
Moderation and content filtering
Industry
Security
Form
Open-source tool
Stage
Beta
Plugs into
nginx/OpenResty
Listed
2026-09-23
Found via
github
Repository
kiwi0719/jev-edge
Stars
41
Forks
1
Last push
2026-10-09
Language
Lua
License
Apache-2.0

The README opens with

jev-edge sits in nginx / OpenResty or Apache APISIX, behind Envoy, Istio, HAProxy, Traefik, Caddy and plain nginx, in a Cloudflare Worker, a Next.js or Node middleware or Lambda@Edge, or inside LiteLLM proxy, and asks one question about incoming requests: what is this request trying to do to my service? It uses TypeSafe Jev, a System One model that returns probabilities instead of prose, to catch prompt injection and abuse at the entry point of LLM-backed applications, before the request reaches your backend.

Badge

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

Listed in Jev in production

Also used for moderation and content filtering