html_processor

Clean up messy HTML from any source — extract the content you want, strip ads and noise, fix links, and sanitize — before you store, convert, or index it.

Feed a Markdown converter or search index a raw web page and you get navigation, ads, and boilerplate in the result. HTML Processor removes that noise first, so downstream output stays clean and token-efficient. It runs a configurable pipeline from a single service call or a saved default, and is standalone: no other Drupal modules required, just a few Symfony/League libraries that Composer installs for you.

What it does

Extracts the content you want with CSS selectors (article, #main-content) and drops the rest.
Removes ads and boilerplate — built-in patterns for common networks, plus your own.
Strips unwanted fragments with admin-trusted regex (guarded against ReDoS).
Rewrites relative links and images to absolute URLs so they keep working.
Sanitizes elements and attributes via the Symfony HTML Sanitizer.
Shapes output — wrap as a full document, or minify.

Pass options per call, or save a default pipeline in the admin form and have it applied automatically.

Use cases

Cleaning HTML before Markdown conversion, AI/RAG ingestion, migrations, or search indexing — anywhere you pull content from sources you don't control.

Requirements

Drupal core ^10.4 || ^11. A few small Symfony and League Composer libraries are installed automatically — no other Drupal modules required. (Exact packages are listed in the README.)

Recommended modules

Optional integrations — HTML Processor works on any HTML string on its own:

An HTML-to-Markdown loader — convert the cleaned HTML to Markdown for documentation or AI/RAG pipelines.
URL Fetcher (url_fetcher) — retrieve remote HTML to feed into the processor.

Install and configure

composer require drupal/html_processor
drush en html_processor -y

Set defaults at Configuration › Content authoring › HTML Processor (permission: Administer HTML Processor settings). Defaults are opt-in, and explicit options passed in code always win.

In code, inject HtmlProcessorInterface and call process():

$clean = $this-&gt;htmlProcessor-&gt;process([
  'content'    =&gt; $rawHtml,
  'container'  =&gt; 'article, #main-content',
  'remove_ads' =&gt; TRUE,
]);

The full service API, the Drush command, autowiring setup, and security notes are in the module's README.md.

Good to know

Container extraction uses explicit CSS selectors — precise, but review them if a source site changes its markup.
Headed for Markdown? Turn minify off (it breaks code blocks) and keep sanitization light to preserve structure.
Regex stripping and ad patterns are admin-trusted only — never safe for anonymous input.

Similar projects

The standalone successor to the cleaning services in Document Loader: HTML Processor (document_loader_html_processor). Pairs naturally with an HTML-to-Markdown loader downstream.