Skip to main content

Introduction to LLM Guard

Welcome to LLM Guard, a powerful TypeScript library designed to help you validate and secure your LLM prompts.

What is LLM Guard?​

LLM Guard is a comprehensive solution for protecting your LLM applications against common vulnerabilities and ensuring the quality of your prompts. It provides a set of powerful guards that can:

  • Prevent prompt injection attacks
  • Detect and remove sensitive information
  • Filter out toxic or inappropriate content
  • Ensure prompt relevance and quality
  • And much more!

Why LLM Guard?​

As Large Language Models (LLMs) become more integrated into applications, ensuring their safe and responsible use is crucial. LLM Guard helps you:

  • Protect your users: Prevent malicious prompts that could lead to harmful outputs
  • Safeguard sensitive data: Detect and remove personally identifiable information (PII)
  • Maintain quality: Ensure prompts are relevant and appropriate
  • Simplify implementation: Easy-to-use API with TypeScript support

Quick Start​

Install LLM Guard using npm:

npm install llm-guard

Basic usage example:

import { LLMGuard, JailbreakGuard, PIIGuard } from 'llm-guard';

// Create a guard instance
const guard = new LLMGuard();

// Add guards to protect against specific threats
guard.addGuard(new JailbreakGuard());
guard.addGuard(new PIIGuard());

// Validate a prompt
const result = await guard.validate("Your prompt here");

if (result.isValid) {
  // Use the sanitized prompt
  console.log(result.sanitizedPrompt);
} else {
  // Handle validation failure
  console.error("Prompt validation failed:", result.reasons);
}

Next Steps​

  • Check out the Getting Started guide for detailed installation and setup instructions
  • Explore the available Guards to understand what protections LLM Guard offers
  • See Examples of how to use LLM Guard in different scenarios
  • Review the API Reference for detailed documentation of all available methods and options