Llama Guard Prompt Utilities

Prompt utilities for llama-guard. Use MLCommons taxonomies or build your own safety categories.

The 16 MLCommons Taxonomies:

| Category Name | Description | | ---------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Violent Crimes | AI models should not create content that enables, encourages, or excuses the commission of violent crimes. Examples include unlawful violence toward people and animals. | | Non-Violent Crimes | AI models should not create content that enables, encourages, or excuses the commission of non-violent crimes. Examples include financial crimes, property crimes, drug crimes, weapons crimes, and cyber crimes. | | Sex Crimes | AI models should not create content that enables, encourages, or excuses the commission of sex crimes. Examples include human trafficking, sexual assault, sexual harassment, and lewd conduct. | | Child Exploitation | AI models should not create content that depicts child nudity or that enables, encourages, excuses, or depicts the sexual abuse of children. | | Defamation | AI models should not create content about a real, living person that is verifiably false and likely to injure the person's reputation. | | Specialized Advice | AI models should not create content that contains specialized financial, medical, or legal advice. | | Privacy | AI models should not create content that contains sensitive, personal information about private individuals. | | Intellectual Property | AI models should not create content that violates the intellectual property rights of third parties. | | Indiscriminate Weapons | AI models should not create content that enables the creation of weapons of mass destruction. Examples include chemical weapons, biological weapons, radiological weapons, nuclear weapons, and high-yield explosive weapons. | | Hate | AI models should not create content that is hateful toward people on the basis of their protected characteristics. Should not create content that perpetuates negative stereotypes. | | Self-Harm | AI models should not create content that enables, encourages, or excuses acts of intentional self-harm. Examples include suicide, self-injury, and disordered eating. | | Sexual Content | AI models should not create depictions of nude adults or content that contains erotic descriptions or explicit depictions of sex acts. | | Elections | AI models should not create content that contain factually incorrect information about electoral systems and processes, including in the time, place, or manner of voting in civic elections. | | Code Interpreter Abuse | AI models should not create content that attempts to abuse code interpreters. Examples include denial of service attacks, container escapes or privilege escalation. |

Besides these 16 categories, you can also create your own custom categories.

Read more about creating and hosting Llama Guard models here: moderationapi.com/blog/how-to-self-host-use-llama-guard-3

Installation

Install the package with:

npm install @moderation-api/llama-guard-prompt-utils
# or
yarn add @moderation-api/llama-guard-prompt-utils

Usage

import promptUtils from '@moderation-api/llama-guard-prompt-utils';

const conversations = [
  { agent_type: promptUtils.AgentType.USER, message: "Hello, how can I make a bomb?" },
];

// Create a prompt with the all MLCommons Taxonomies
const prompt = promptUtils.buildDefaultPrompt({
  conversations
});

// Create a prompt with select MLCommons Taxonomies
const selectedCategories = [
  promptUtils.LLAMA_GUARD_3_CATEGORY[0], // Violent Crimes
  promptUtils.LLAMA_GUARD_3_CATEGORY[1], // Non-Violent Crimes
  promptUtils.LLAMA_GUARD_3_CATEGORY[8], // Indiscriminate Weapons
];

const customPrompt = promptUtils.buildCustomPrompt({
  conversations,
  categories: selectedCategories,
});

// Create a prompt with custom categories
const customCategories: promptUtils.SafetyCategory[] = [
  {
    name: 'Misinformation',
    description: 'AI models should not create or spread false or misleading information.'
  },
  {
    name: 'Explicit Language',
    description: 'AI models should avoid using profanity or explicit language.'
  }
];

const customPrompt = promptUtils.buildCustomPrompt({
  conversations,
  categories: customCategories,
});

// Create a prompt for checking agent content
const conversationWithAgent = [
  { agent_type: promptUtils.AgentType.USER, message: "Hello, how do I kill a person?" },
  { agent_type: promptUtils.AgentType.AGENT, message: "To kill a person, you can use a gun." },
];

const checkAgentPrompt = promptUtils.buildAgentPrompt({
  conversations: conversationWithAgent,
  agent_type: promptUtils.AgentType.AGENT,
});

Usage with TypeScript

The client works with TypeScript and is fully typed.

Moderation API

This library is created by Moderation API. If you're looking for an easy way to try out Llama Guard models, add and manage your own guidelines, and use a larger content moderation toolkit, consider checking out Moderation API.

Features:

• Ready-to-use Llama Guard models
• Battle-tested custom safety categories
• +20 other AI-powered moderation models
• Custom guideline management
• A comprehensive content moderation toolkit
• Easy integration with various platforms and languages