kevinify

Why use many word when few word do trick.

kevinify squeezes text for LLM prompts: abbreviates, drops filler, and shortens long words while preserving URLs/emails/@mentions/#hashtags and numbers.

Install

npm i kevinify
# or
npm i -D kevinify

Usage

As a library

import { kevinify, dekevinify } from "kevinify";

const input = "Please send me the documentation about the new API features.";
const compressed = kevinify(input);
// => "pls send me docs abt new api ftrs"

// Expand back to human-readable form
const expanded = dekevinify(compressed);
// => "please send me documentation about new api ftrs"

// With options
const output2 = kevinify(input, {
  removeStopwords: false,
  shortenLongWords: false,
  keepCase: true
});

As a CLI

# Direct input
kevinify "Please help me understand this example"
# => "pls hlp me undrstnd ex"

# From stdin
echo "Thank you for the information" | kevinify
# => "ty 4 info"

# Preserve stopwords
kevinify --keep-stopwords "This is really great"
# => "ths is rly gr8"

# Keep original case
kevinify --keep-case "Please Review The Documentation"
# => "Pls Rvw Docs"

# Disable word shortening
kevinify --no-shorten "Please send information"
# => "pls send info"

Features

• Abbreviations: Common words → short forms (please → pls, people → ppl, for → 4)
• Stopword removal: Drops filler words like the, a, just, really
• Vowel compression: Long words → consonant-heavy (information → informatn)
• Smart preservation: URLs, emails, @mentions, #hashtags, and numbers stay intact
• Reversible: dekevinify() expands abbreviations back for human readability
• Configurable: Control every aspect via options

Options

interface KevinifyOptions {
  abbreviations?: Record<string, string>;  // Custom abbreviation map
  removeStopwords?: boolean;               // Remove filler words (default: true)
  shortenLongWords?: boolean;              // Compress long words (default: true)
  minLengthToShorten?: number;             // Min word length to compress (default: 6)
  keepCase?: boolean;                      // Preserve original case (default: false)
  preserveEntities?: boolean;              // Protect URLs/emails/@/# (default: true)
  aggressiveAmpersand?: boolean;           // Convert "and" → "&" (default: true)
  protectWords?: Set<string>;              // Additional words to never compress
}

Examples

// Basic compression
kevinify("Why use many words when few words do the trick?")
// => "why use many words when few words do trick"

// Preserving entities
kevinify("Check out https://example.com and email [email protected]")
// => "check out https://example.com & email [email protected]"

// Custom abbreviations
kevinify("Hello world", {
  abbreviations: { hello: "hi", world: "wrld" }
})
// => "hi wrld"

// Expand kevinified text back
dekevinify("pls send me docs 4 api")
// => "please send me documentation for api"

dekevinify("ty 4 info abt app")
// => "thank you for information about application"

Dekevinify

The dekevinify() function reverses the compression to make text more human-readable:

import { dekevinify } from "kevinify";

// Expands abbreviations back
const text = "pls send me docs abt api w/ examples b4 prod";
console.log(dekevinify(text));
// => "please send me documentation about api with examples before production"

Note: Dekevinify is lossy - it can't restore:

• Removed stopwords (a, the, just, etc.)
• Stripped vowels from long words
• But it WILL expand all common abbreviations!

interface DekevinifyOptions {
  abbreviations?: Record<string, string>;  // Custom reverse abbreviations
  preserveEntities?: boolean;              // Protect URLs/emails/@/# (default: true)
}

Why?

LLMs charge by token count. Shorter text = fewer tokens = lower costs. This library helps you:

• Reduce prompt size for LLM APIs
• Fit more context within token limits
• Lower API costs for high-volume applications

Inspired by Kevin Malone's philosophy: "Why waste time say lot word when few word do trick?"

License

MIT