Glin Profanity
ML-powered profanity filter with TensorFlow.js toxicity detection, 23-language support, and leetspeak/Unicode obfuscation resistance. 21M+ ops/sec, 69k+ npm downloads a month.
Basic profanity filters break the moment someone types "pr0fan1ty" instead of the real word. Glin Profanity is built to catch that: it pairs a fast wordlist-based filter with TensorFlow.js toxicity detection, and normalizes leetspeak and Unicode obfuscation before checking, so the workarounds people use to dodge simple filters mostly don't work.
It supports 23 languages, runs at 21M+ operations per second, and ships customizable word lists, severity scoring, and context-aware filtering with LRU caching to keep repeated checks cheap. Available as an npm package with React hooks, a PyPI package for Python, an MCP server for AI assistants, and a web tool on GLINCKER.
Averages 69,000+ downloads a month on npm, by far the most-used package I maintain.
Related Projects
theauth
Auth for AI agents and humans (rebranded from Kavachos). OAuth 27+ providers, enterprise SAML/OIDC, MCP 2.1 auth server, scoped permissions, and delegation chains.
theSVG
The open SVG brand library with 2,700+ GitHub stars. 5,600+ brand icons with multi-variant support (color, mono, light, dark, wordmark). npm packages, CLI, API, and MCP server.
FeaturedDrop
Open-source product adoption toolkit: feature announcements, guided tours, checklists, and in-app feedback, driven by a 3kB client-side behavioral engine.