Gangajal — Global Profanity Filtering Engine

Secure • Fast • Cross-language • No raw bad words in repo

Gangajal is a deterministic, high-performance profanity filter that works in every programming language via WebAssembly.
It uses Unicode normalization, Bloom filters, and SHA-256 hashing — keeping your dictionary 100% private while delivering lightning-fast filtering.

✨ Features

• Supports all world languages (Unicode NFKC + letters + combining marks)
• Zero raw profanity words ever in the repository
• Extremely fast (Bloom filter + binary search)
• One WASM core — works in Node.js, Python, Go, .NET, Java, Rust, browsers, etc.
• Deterministic & reproducible
• Async API with asset hot-reloading support

🚀 Quick Start

Install

npm install @saugatedith/gangajal

Usage

const { validate, reloadAssets } = require('@saugatedith/gangajal');

(async () => {
  // Full mask mode (0): masks entire word
  console.log(await validate("hello badword here", 0));  // "hello ******** here"

  // Partial mask mode (1): keeps first char
  console.log(await validate("hello badword here", 1));  // "hello b****** here"
})();

Optional (only if you changed assets/ on disk while your process is running):

reloadAssets();

Modes

• mode=0: Full mask - replaces entire word with * (e.g., badword → *******)
• mode=1: Partial mask - keeps first character (e.g., badword → b******)

ESM Support

If your file is *.js, either set "type": "module" in your project's package.json or rename your entry file to *.mjs to avoid Node's [MODULE_TYPELESS_PACKAGE_JSON] warning.

import { validate, reloadAssets } from '@saugatedith/gangajal';

const clean = await validate("text to filter", 1);
console.log(clean);

reloadAssets();

Architecture

• Binary Assets (public) → badwords.bloom + badwords.hash.bin
• WASM Core → same engine everywhere
• Language Bindings → JS, Python, Go, .NET, etc.

Full specification → SPEC.md

Also Available for Python

from gangajal import validate, reload_assets

print(validate("this is inappropriate", 1))

pip install gangajal

License

MIT © Saugat