Crate hanzo_guard 

Hanzo Guard

LLM I/O sanitization and safety layer - the “condom” for AI.

Hanzo Guard sits between your application and LLM providers, sanitizing all inputs and outputs to prevent:

• PII Leakage: Detects and redacts personal identifiable information
• Prompt Injection: Detects jailbreak and manipulation attempts
• Unsafe Content: Filters harmful content via Zen Guard models
• Rate Abuse: Prevents excessive API usage
• Audit Violations: Logs all requests for compliance

Quick Start

use hanzo_guard::{Guard, GuardConfig, SanitizeResult};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let guard = Guard::new(GuardConfig::default());

    // Sanitize input before sending to LLM
    let input = "My SSN is 123-45-6789, can you help me?";
    let result = guard.sanitize_input(input).await?;

    match result {
        SanitizeResult::Clean(text) => {
            // Safe to send to LLM
            println!("Clean: {}", text);
        }
        SanitizeResult::Redacted { text, redactions } => {
            // PII was redacted
            println!("Redacted: {} ({} items)", text, redactions.len());
        }
        SanitizeResult::Blocked { reason, category } => {
            // Content blocked
            println!("Blocked: {} ({:?})", reason, category);
        }
    }

    Ok(())
}

Architecture

┌─────────────┐     ┌──────────────┐     ┌─────────────┐
│ Application │ ──► │ Hanzo Guard  │ ──► │ LLM Provider│
└─────────────┘     │              │     └─────────────┘
                    │ ┌──────────┐ │
                    │ │ PII      │ │
                    │ │ Detector │ │
                    │ └──────────┘ │
                    │ ┌──────────┐ │
                    │ │ Injection│ │
                    │ │ Detector │ │
                    │ └──────────┘ │
                    │ ┌──────────┐ │
                    │ │ Content  │ │
                    │ │ Filter   │ │
                    │ └──────────┘ │
                    │ ┌──────────┐ │
                    │ │ Rate     │ │
                    │ │ Limiter  │ │
                    │ └──────────┘ │
                    │ ┌──────────┐ │
                    │ │ Audit    │ │
                    │ │ Logger   │ │
                    │ └──────────┘ │
                    └──────────────┘

Re-exports

pub use config::GuardConfig;

pub use error::GuardError;

pub use error::Result;

pub use guard::Guard;

pub use types::*;

Modules

audit

Audit logging for Guard

config

Configuration for Hanzo Guard

content

Content filtering via Zen Guard API

error

Error types for Hanzo Guard

guard

Main Guard implementation

injection

Prompt injection and jailbreak detection

pii

PII (Personally Identifiable Information) detection and redaction

prelude

Prelude for convenient imports

rate_limit

Rate limiting for Guard

types

Core types for Hanzo Guard