spam-humanizer

spam-humanizer

A lightweight JavaScript library for normalizing obfuscated text and detecting spam across English, Russian, and Ukrainian languages using OpenAI or DeepSeek AI providers.

Features

• Text Humanization: Convert garbled or obfuscated characters (e.g., ೧ⲢӤβεT) into readable text by mapping Unicode symbols to their Latin or Cyrillic equivalents.
• Spam Detection: Leverage GPT-based models to classify whether a piece of text is spam or advertisement, with support for single-language or multi-language checks.
• Flexible AI Providers: Choose between OpenAI and DeepSeek APIs with a unified interface.
• Multi-Language Support: Built-in dictionaries for English (en.json), Russian (ru.json), and Ukrainian (ua.json).

Installation

npm install spam-humanizer

Quick Start

import { SpamHumanizer } from "spam-humanizer";

// Initialize with your API key and preferred provider
const humanizer = new SpamHumanizer({
  apiKey: process.env.OPENAI_KEY,
  provider: "openai"  // or "deepseek"
});

(async () => {
  // Humanize garbled text
  const readable = await humanizer.humanize("೧ⲢӤβεT", "ru");
  console.log(readable); // "Привет"

  // Spam detection (single language)
  const result = await humanizer.isSpam("Buy cheap Viagra now!", "en");
  console.log(result);
  // → { spam: true, lang: 'en', reason: 'yes: promotional content detected' }

  // Spam detection (multi-language fallback)
  const multiCheck = await humanizer.isSpam("Купи підписку вже зараз!");
  console.log(multiCheck);
  // → { spam: true, lang: 'ua', reason: 'yes: рекламний текст' }
})();

API Reference

new SpamHumanizer({ apiKey, provider })

• `` (string, required): Your API key for the AI provider.
• `` (string, optional): Either "openai" (default) or "deepseek".

.humanize(text, lang?)

Normalize obfuscated text:

• `` (string, required): The input string to normalize.
• `` ("en"|"ru"|"ua", optional): Primary language. Defaults to "en".

Returns a Promise<string> with the humanized text.

.isSpam(text, lang?)

Detects whether text is spam:

• `` (string, required): The input to classify.
• `` ("en"|"ru"|"ua", optional): If omitted, checks across all three languages.

Returns a Promise<Object>:

• spam (boolean)
• lang (string) – the language in which spam was detected
• reason (string) – the raw model response

License

MIT © @radonl

Donate

BTC: bc1qzjjffns84rxcfg9hqd4se76p5k3f0y5urv4q52

ETH: 0x7c625921BF29337fB91231d0df78c88B97cA85a7

USDT(ETH): 0x7c625921BF29337fB91231d0df78c88B97cA85a7