faradailab/guard-laya

FaradAiGuard Laya-Guard Security SLM 🛡️

Deep heuristic security evaluation model with agent-to-agent prompt injection defense. Part of the FaradAiGuard Security SLM Factory.

🚀 Instant Ollama Run

Run this model directly without local configuration or compilation:

ollama run hf.co/faradailab/guard-laya:latest

⚙️ Model Details

• Distribution: Hugging Face Hub (faradailab/guard-laya)
• Format: GGUF (guard-laya.Q4_K_M.gguf)
• System Prompt & Strict JSON Grammar: Pre-configured via Modelfile.
• Repository: https://github.com/kurtfi/FaradAiGuard

📐 Strict Output Schema (GBNF Grammar)

FaradAiGuard models return strictly validated JSON:

{
  "verdict": "ALLOW" | "BLOCK",
  "risk_score": 0.0 - 1.0,
  "reason": "Brief technical security explanation",
  "cwe": "CWE-ID or empty string"
}