Shadow Warden AI — Gateway v7.0 — Explore the API Reference
Filter pipeline
ALL SYSTEMS HEALTHY

Tune how Shadow Warden inspects every AI request before it leaves your perimeter.

SemanticBrain ML cosine
RECOMMENDED

Async ML detection layer using all-MiniLM-L6-v2 embeddings. Catches paraphrased jailbreaks that regex rules miss.

Adds ~15ms latency · +18% recall on adversarial set
EvolutionEngine
BETA

Auto-generate new detection rules from blocked attacks via Claude Opus. Hot-reloads without restart.

💡
BEST PRACTICE
Enable in staging first. Consumes API credits. Review the generated corpus weekly under Rules → Auto-generated.
GDPR strict mode
EU SOC 2

Never log request content. Only metadata (type, length, timing) is persisted. Required for EU deployments.

Endpoints /gdpr/export and /gdpr/purge rely on this
Sensitivity profile

Tradeoff between false-positive rate and recall. Balanced works for most workloads.

Rate limit per API key

Hard cap on requests per key per minute. Above 1,000 may require dedicated Redis.

req / min