gg-friggin-ez launches as an open-source, low-cost Node.js library for evasion-aware multilingual profanity and toxicity screening.
gg-friggin-ez is an open-source Node.js package designed for high-speed, evasion-resilient profanity and toxicity moderation. Powered by TypeSafe AI's System 1 models (Jev and Laya), the library identifies nuanced online chat evasion techniques such as leetspeak, ASCII art, deliberate character spacing, and romanized profanity across languages including Kannada, Telugu, Tamil, Hindi, and Bengali. Operating at sub-500ms latency and roughly $0.000004 per message, it translates calibrated toxicity and obfuscation probabilities into configurable deterministic policies to trigger actions like allowing, reviewing, censoring, or banning content.
General-purpose LLMs are too sluggish and costly for real-time game chat and streaming moderation, while static regex filters fail against leetspeak and romanized Indic slang—making lightweight System 1 models the practical sweet spot for automated moderation.
• Purpose-built System 1 models drastically undercut full LLMs on cost and latency while delivering deterministic, actionable scores.
• Robust evasion detection closes common moderation loopholes around ASCII art, character substitution, and non-standard spacing.
• Native support for romanized Indic languages solves a major blind spot present in traditional, English-centric moderation suites.
• Drop-in Node.js integration with configurable thresholds simplifies deployment into existing chat and multiplayer game backends.
DISCOVERED
1d ago
2026-09-22
PUBLISHED
1d ago
2026-09-22
RELEVANCE
AUTHOR
Shikhar Srivastava