AI Guardrail Gateway

Enterprise AI content safety gateway. Screens every prompt and response through 7 guardrails running in parallel on CPU with ONNX INT8 quantization. Supports English and Chinese (Traditional/Simplified).

v0.6.4

Checking...

Open Console →

Swagger UI · API Health · ML Services

47ms

7 Guardrails

21K

Test Cases

99.99%

Pass Rate

F1 1.0

All Guardrails

Guardrails (Parallel Execution)

Keyword<1ms

Toxicity~21ms

Injection~33ms

PII~13ms

Intent~47ms

Semantic~22ms

Output Safety~2ms

Architecture

ONNX INT8 Quantization

All 5 ML models quantized. CPU-only, no GPU required.

8 Microservices on K8s

Each ML model in its own pod, independently scalable.

Chinese NER (CKIP)

Traditional/Simplified Chinese PII detection. Unique in market.

21K Test Suite

3-tier (L0/L1/L2), 9 guardrails, auto-calibrated. F1=1.000.

Spring Boot 3 FastAPI ONNX Runtime React Ant Design Pro K3s PostgreSQL Redis Elasticsearch Qdrant

Elasticsearch Qdrant

More Projects

Coming soon...

Planned