ainvirion/aiproxyguard repository overview

⁠AIProxyGuard

Drop-in security layer for LLM applications. Intercepts requests to OpenAI, Anthropic, and other providers to block prompt injections before they reach the model.

⁠Why AIProxyGuard?

LLMs are vulnerable to adversarial inputs that bypass safety guardrails. AIProxyGuard acts as a reverse proxy, scanning every request in real-time and blocking:

• Prompt injection ("ignore previous instructions...")
• Jailbreak attempts (DAN, persona exploits)
• Encoding bypasses (base64, hex, ROT13)
• Delimiter attacks (JSON/XML injection)
• Unicode evasion (homoglyphs, fullwidth)

⁠30-Second Setup

docker run -d -p 8080:8080 ainvirion/aiproxyguard:latest

Change your base URL from api.openai.com to localhost:8080 — that's it.

client = OpenAI(base_url="http://your-aiproxyguard:8080/openai/v1⁠")

⁠Features

• Multi-Provider — OpenAI, Anthropic, OpenRouter, Ollama
• Detection-Only Mode — /check endpoint for scanning without forwarding
• ML + Signatures — Layered detection with low false positives
• Policy Engine — Block, warn, or log per category
• Metrics — Prometheus endpoint at /metrics
• Multi-Arch — linux/amd64 and linux/arm64

⁠Links

• Docs: https://ainvirion.github.io/aiproxyguard⁠
• GitHub: https://github.com/AInvirion/aiproxyguard⁠
• Cloud: https://aiproxyguard.com⁠