llm-html-compressor

A specialized HTML compressor designed to optimize HTML content for use as context with Large Language Models (LLMs). Removes unnecessary whitespace, comments, and other "noise" from HTML documents to make them more suitable for LLM processing while preserving semantically important content.

Installation

npm install llm-html-compressor
# or
yarn add llm-html-compressor

Why use llm-html-compressor?

When using HTML content as context for LLMs, unnecessary elements like whitespace, comments, and certain attributes can:

Consume token quota without adding value
Add noise that makes it harder for the LLM to focus on important content
Increase the chance of context truncation

This library provides targeted optimizations specifically designed for LLM context usage, distinct from traditional HTML minifiers which focus on network transmission size.

Usage

Basic Usage

import { compress } from 'llm-html-compressor';

const html = `
<!DOCTYPE html>
<html>
  <!-- This is a comment -->
  <head>
    <title>Example</title>
    <style>
      body { font-family: Arial, sans-serif; }
    </style>
  </head>
  <body>
    <div class="container" id="main">
      <h1>Hello, World!</h1>
      <p style="color: blue; font-size: 16px;">This is an example.</p>
    </div>
    <script>
      console.log('Hello');
    </script>
  </body>
</html>
`;

const compressed = compress(html);
console.log(compressed);
// Output: <!DOCTYPE html><html><head><title>Example</title><style>body { font-family: Arial, sans-serif; }</style></head><body><div class="container" id="main"><h1>Hello, World!</h1><p style="color:blue;font-size:16px;">This is an example.</p></div><script>console.log('Hello');</script></body></html>

Advanced Usage with Custom Options

import { createCompressor } from 'llm-html-compressor';

const compressor = createCompressor({
  removeComments: true,
  collapseWhitespace: true,
  removeEmptyAttributes: true,
  removeStyleTags: true,
  removeScriptTags: true,
  preserveLineBreaks: false,
  removeDataAttributes: true,
  removeHiddenElements: true,
  minifyInlineCSS: true,
  removeClassAttributes: true,
  removeIdAttributes: false
});

const html = `... your HTML here ...`;
const compressed = compressor.compress(html);

API

Functions

`compress(html: string): string`

Compresses HTML using default options.

`createCompressor(options?: Partial<CompressionOptions>): HtmlCompressor`

Creates a compressor instance with custom options.

Classes

`HtmlCompressor`

The main compressor class that can be instantiated directly.

import { HtmlCompressor } from 'llm-html-compressor';

const compressor = new HtmlCompressor(options);
const result = compressor.compress(html);

CompressionOptions

Option	Type	Default	Description
`removeComments`	boolean	`true`	Removes HTML comments
`collapseWhitespace`	boolean	`true`	Collapses multiple whitespace characters into a single space
`removeEmptyAttributes`	boolean	`true`	Removes attributes with empty values
`removeStyleTags`	boolean	`false`	Removes `<style>` tags and their content
`removeScriptTags`	boolean	`false`	Removes `<script>` tags and their content
`preserveLineBreaks`	boolean	`false`	Preserves line breaks when collapsing whitespace
`removeDataAttributes`	boolean	`false`	Removes data-* attributes
`removeHiddenElements`	boolean	`false`	Removes elements with display:none or hidden attribute
`minifyInlineCSS`	boolean	`false`	Minifies inline CSS in style attributes
`removeClassAttributes`	boolean	`false`	Removes class attributes
`removeIdAttributes`	boolean	`false`	Removes id attributes

Use Cases

Preprocessing HTML for RAG (Retrieval-Augmented Generation) systems
Optimizing web page content for chatbots and assistants
Reducing token usage when working with HTML documentation
Making HTML content more digestible for code analysis with LLMs

License

MIT