LLMs Helper

Writes static Markdown copies of WordPress content and an llms.txt to uploads, served at the llmstxt.org URLs (/page/index.html.md and /llms.txt). See readme.txt for what the plugin does.

Requirements

• PHP 8.2+
• WordPress 6.0+

Development

composer install
npm install

PHPUnit

Runs against a MySQL database, using wp-phpunit and the WordPress copy roots/wordpress installs into wordpress/.

Copy tests/.env_sample to tests/.env and fill in a database the tests can empty. Every table with the wpphpunittests_ prefix is dropped.

composer test
composer coverage           # HTML report in tests/.cache/coverage

Mutation testing

Infection. Slow, so it is run occasionally, one thread at a time:

composer mutation -- --threads=1    # reports in tests/.cache/infection

Linting

composer lint:php           # PHPCS (src + tests) and PHPStan
composer format:php         # PHPCBF

End-to-end

Playwright against a wp-env site on PHP 8.3 (ports 57891 and 57892). Needs Node 20 (.nvmrc), Docker, and vendor/ from composer install.

nvm use
npx playwright install chromium
npm run wp-env:start
npm run test:e2e

Copy e2e/.env.example to e2e/.env to change the site URL or login.

Copies are rendered by a request from the site to its own page (Loopback_Content_Renderer). Inside the wp-env containers the site's address, localhost and its port, does not reach the web server, so e2e/mu-plugins/llms-e2e-loopback.php sends that request to the compose service (wordpress or tests-wordpress) with the site's Host header.

seo-noindex.spec.js runs the same tests against Yoast SEO, Rank Math, SEOPress, AIOSEO and Slim SEO, one at a time. The global setup installs them from wordpress.org, inactive. e2e/mu-plugins/llms-e2e-seo.php adds wp llms-e2e-seo, which sets noindex where each plugin stores it, and each test checks the copy against the robots tag the plugin prints.

membership.spec.js does the same for Paid Memberships Pro, Members, Content Control, Ultimate Member, Restrict Content, LifterLMS and WooCommerce Memberships, through wp llms-e2e-access, checking the copy against whether a logged out visitor sees the post's content. Paid Memberships Pro is installed from its GitHub tag, as it has left wordpress.org. WooCommerce Memberships is paid, so e2e/plugins/woocommerce-memberships stands in for it, as on CI: it reads the real plugin's wc_memberships_rules and hides what the real plugin hides from a visitor with no membership. To test against the real plugin, map a local copy over it in .wp-env.override.json, which is not committed:

{
    "mappings": {
        "wp-content/mu-plugins": "./e2e/mu-plugins",
        "wp-content/plugins/woocommerce-memberships": "/path/to/woocommerce-memberships"
    }
}

cache.spec.js turns on the page cache of WP Super Cache, W3 Total Cache and WP Fastest Cache, and Autoptimize, one at a time, through each plugin's own commands and settings (e2e/fixtures/cache-plugins.js). Each first checks the plugin is serving a page from its cache, then that llms.txt and a copy show each change, redirected and sent by PHP. WP Rocket is paid, so e2e/mu-plugins/llms-e2e-cache.php stands in for its rocket_clean_files() and records the URLs it is given. LiteSpeed Cache needs a LiteSpeed web server, so only its litespeed_purge_url action is covered, by PHPUnit.

sitemap.spec.js turns on Yoast SEO and Rank Math, one at a time, and checks the sitemap index lists llms-sitemap.xml, with llms.txt in it, only while "Let search engines index llms.txt" is on, and that each plugin's cached sitemaps change when the setting changes or llms.txt is rebuilt.

GitHub workflows

• tests.yml PHPUnit on PHP 8.2 to 8.4.
• quality.yml PHPCS and PHPStan.
• e2e.yml Playwright in 4 shards, each on its own runner, on pull requests.
• mutation.yml Infection, one thread, started by hand.
• release.yml builds llms-helper.zip when a GitHub release is published.

Layout

• src/Container small autowiring DI container.
• src/Hooks the Hook_Loader and Hookable interface.
• src/App.php builds each controller through the container and registers its hooks.
• src/Plugin.php the DI rules binding each interface to its class.
• src/Interfaces the interfaces each part is swapped through.
• src/Controllers the hooked classes.
• src/Content the converter, content renderers, file store, copy and llms.txt builders, and the SEO and membership plugin checks.
• src/Sitemap the llms sitemap and its Yoast SEO and Rank Math providers.
• src/Http routes .md and /llms.txt requests before the rest of WordPress loads.
• src/Queue and src/Cron background jobs, through Action Scheduler when installed, otherwise WP-Cron.
• src/Settings, src/Cache, src/CLI the settings option, cache purging and WP-CLI commands.