LLM Guard: Security Toolkit for LLM Interactions

Purpose

Comprehensive security toolkit by Protect AI for fortifying LLM interactions through sanitization, harmful language detection, data leakage prevention, and prompt injection resistance.

Prompt Input Scanners (15 types)

Anonymize, BanCode, BanCompetitors, BanSubstrings, BanTopics, Code detection, Gibberish detection, InvisibleText, Language detection, PromptInjection, Regex, Secrets, Sentiment analysis, TokenLimit, Toxicity

Output Scanners (20 types)

BanCode, BanCompetitors, BanSubstrings, BanTopics, Bias detection, Code detection, Deanonymize, JSON validation, Language checks, MaliciousURLs, NoRefusal, ReadingTime, FactualConsistency, Gibberish, Regex, Relevance, Sensitive content, Sentiment, Toxicity, URLReachability

Key Quote

"By offering sanitization, detection of harmful language, prevention of data leakage, and resistance against prompt injection attacks" -- ensures secure LLM interactions.

Technical Details

• Python (99%), MIT licensed
• Python 3.9+
• pip install llm-guard
• 2.9k stars, 391 forks
• Documentation, Playground, API deployment options available