class-llms-txt.php — Jetpack – WP Security, Backup, Speed, & Growth 16.1-beta

jetpack / jetpack_vendor / automattic / jetpack-seo / src / class-llms-txt.php

jetpack / jetpack_vendor / automattic / jetpack-seo / src Last commit date

class-llms-txt.php

296 lines

1	<?php
2	/**
3	* Generation and serving of the site's llms.txt.
4	*
5	* Serves a root-level `/llms.txt` — the emerging convention (llmstxt.org) for
6	* giving AI assistants a curated, Markdown map of a site's key content. Gated on
7	* the durable option `jetpack_seo_llms_txt_enabled` (round-tripped through the
8	* existing `/jetpack/v4/settings` endpoint by the AI tab).
9	*
10	* Serving approach: this hooks `template_redirect` and inspects the raw request
11	* path rather than registering a rewrite rule, so it needs no rewrite flush and
12	* works under any permalink structure. The tradeoff — it matches a root-level
13	* request only, and like robots.txt won't apply on a multisite subdirectory or
14	* a site fronted by a static file. Hardening the serving path (a rewrite rule,
15	* caching) is tracked separately in JETPACK-1830.
16	*
17	* @package automattic/jetpack-seo-package
18	*/
19
20	namespace Automattic\Jetpack\SEO;
21
22	/**
23	* Builds and serves the site's llms.txt.
24	*/
25	class Llms_Txt {
26
27	/**
28	* Option recording whether llms.txt generation is enabled. Mirrored as a
29	* literal in the plugin's settings endpoint whitelist (`jp_group =>
30	* 'seo-tools'`).
31	*
32	* @var string
33	*/
34	const OPTION = 'jetpack_seo_llms_txt_enabled';
35
36	/**
37	* Max items listed per supported content type, so a large site doesn't dump
38	* its whole tree.
39	*/
40	const MAX_POSTS_PER_TYPE = 50;
41
42	/**
43	* Wire the front-end request handler.
44	*
45	* @return void
46	*/
47	public static function init() {
48	add_action( 'template_redirect', array( __CLASS__, 'maybe_serve' ) );
49	}
50
51	/**
52	* Whether llms.txt generation is currently switched on.
53	*
54	* @return bool
55	*/
56	public static function is_enabled() {
57	return (bool) get_option( self::OPTION, false );
58	}
59
60	/**
61	* Whether WordPress can actually serve the dynamic `/llms.txt` on this site.
62	*
63	* Two ways it can't, and the toggle then silently does nothing:
64	* - A physical `llms.txt` sits at the site root, so the web server delivers
65	* that file directly and WordPress never runs for the request. Detected
66	* here.
67	* - The host fronts WordPress for root-level paths (some managed hosts /
68	* static-file setups). PHP can't see that, so hosts can signal it via the
69	* `jetpack_seo_llms_txt_can_serve` filter.
70	*
71	* When this returns false the GEO tab shows an honest "can't take effect"
72	* notice instead of letting the admin believe the file is live.
73	*
74	* @return bool
75	*/
76	public static function can_serve() {
77	$has_static_file = defined( 'ABSPATH' ) && file_exists( ABSPATH . 'llms.txt' );
78
79	/**
80	* Filters whether WordPress can serve the dynamic `/llms.txt`. Hosts that
81	* intercept root-level paths before WordPress runs can return false to
82	* surface the honest "can't take effect" state in the SEO dashboard.
83	*
84	* @since 0.7.0
85	*
86	* @param bool $can_serve Whether WordPress can serve `/llms.txt`.
87	*/
88	return (bool) apply_filters( 'jetpack_seo_llms_txt_can_serve', ! $has_static_file );
89	}
90
91	/**
92	* The site-root-relative request path, normalized without surrounding
93	* slashes (e.g. `llms.txt`).
94	*
95	* @return string
96	*/
97	private static function request_path() {
98	if ( empty( $_SERVER['REQUEST_URI'] ) ) {
99	return '';
100	}
101	$uri = esc_url_raw( wp_unslash( $_SERVER['REQUEST_URI'] ) );
102	$path = (string) wp_parse_url( $uri, PHP_URL_PATH );
103	return trim( $path, '/' );
104	}
105
106	/**
107	* Serve llms.txt when enabled and the request is for it; otherwise no-op.
108	*
109	* @return void
110	*/
111	public static function maybe_serve() {
112	if ( ! self::is_enabled() \|\| 'llms.txt' !== self::request_path() ) {
113	return;
114	}
115
116	// Respect the site's "discourage search engines" setting. When indexing is
117	// discouraged, WordPress emits `Disallow: /` in robots.txt — serving an AI
118	// content map would contradict that signal, and would expose a content map
119	// on staging/private sites, so we don't serve it.
120	if ( ! get_option( 'blog_public' ) ) {
121	return;
122	}
123
124	nocache_headers();
125	status_header( 200 );
126	header( 'Content-Type: text/plain; charset=utf-8' );
127
128	// Content is plain-text Markdown; not HTML, so it's emitted as-is.
129	echo self::generate(); // phpcs:ignore WordPress.Security.EscapeOutput.OutputNotEscaped
130	exit;
131	}
132
133	/**
134	* Build the llms.txt Markdown document from site identity and published
135	* supported content types.
136	*
137	* @return string
138	*/
139	public static function generate() {
140	$name = wp_strip_all_tags( get_bloginfo( 'name' ) );
141	$description = wp_strip_all_tags( get_bloginfo( 'description' ) );
142
143	$blocks = array();
144
145	$blocks[] = '# ' . ( '' !== $name ? $name : home_url() );
146	if ( '' !== $description ) {
147	$blocks[] = '> ' . $description;
148	}
149
150	foreach ( self::get_llms_post_type_objects() as $post_type ) {
151	$list = self::link_list( self::get_posts_for_type( $post_type->name ) );
152	if ( '' !== $list ) {
153	$blocks[] = '## ' . self::section_title( $post_type ) . "\n\n" . $list;
154	}
155	}
156
157	return implode( "\n\n", $blocks ) . "\n";
158	}
159
160	/**
161	* Return supported post types in llms.txt reading order.
162	*
163	* @return \WP_Post_Type[]
164	*/
165	private static function get_llms_post_type_objects() {
166	$post_types = Post_Types::get_supported_content_type_objects();
167	$ordered = array();
168
169	foreach ( array( 'page', 'post' ) as $post_type ) {
170	if ( isset( $post_types[ $post_type ] ) ) {
171	$ordered[ $post_type ] = $post_types[ $post_type ];
172	unset( $post_types[ $post_type ] );
173	}
174	}
175
176	return array_merge( $ordered, $post_types );
177	}
178
179	/**
180	* Fetch published content for a supported post type.
181	*
182	* @param string $post_type Post type slug.
183	* @return \WP_Post[]
184	*/
185	private static function get_posts_for_type( $post_type ) {
186	$args = array(
187	'post_type' => $post_type,
188	'post_status' => 'publish',
189	'has_password' => false,
190	'posts_per_page' => self::MAX_POSTS_PER_TYPE,
191	'suppress_filters' => false,
192	'meta_query' => array(
193	// WP_Query's meta_query takes its `relation` as a keyed entry alongside
194	// the positional clauses below; the mixed shape is required, not a slip.
195	// @phan-suppress-next-line PhanPluginMixedKeyNoKey
196	'relation' => 'OR',
197	array(
198	'key' => Content_Coverage::META_NOINDEX,
199	'compare' => 'NOT EXISTS',
200	),
201	array(
202	'key' => Content_Coverage::META_NOINDEX,
203	'value' => '1',
204	'compare' => '!=',
205	),
206	),
207	);
208
209	if ( 'page' === $post_type ) {
210	$args['orderby'] = 'menu_order title';
211	$args['order'] = 'ASC';
212	} else {
213	$args['orderby'] = 'date';
214	$args['order'] = 'DESC';
215	}
216
217	return get_posts( $args );
218	}
219
220	/**
221	* Section title for a supported post type.
222	*
223	* @param \WP_Post_Type $post_type Post type object.
224	* @return string
225	*/
226	private static function section_title( $post_type ) {
227	if ( 'page' === $post_type->name ) {
228	return __( 'Pages', 'jetpack-seo' );
229	}
230	if ( 'post' === $post_type->name ) {
231	return __( 'Posts', 'jetpack-seo' );
232	}
233	return wp_strip_all_tags( $post_type->label );
234	}
235
236	/**
237	* Render a list of posts as llms.txt link lines:
238	* `- [Title](url): summary`.
239	*
240	* @param \WP_Post[] $posts Posts to render.
241	* @return string Newline-joined link lines, or '' when there are none.
242	*/
243	private static function link_list( $posts ) {
244	$lines = array();
245	foreach ( $posts as $post ) {
246	// Password-protected posts are `publish` status but their content is
247	// intentionally gated; keep them out of the public llms.txt entirely.
248	if ( ! empty( $post->post_password ) ) {
249	continue;
250	}
251
252	// Collapse whitespace (incl. newlines) so each entry stays on one line.
253	$title = trim( preg_replace( '/\s+/', ' ', wp_strip_all_tags( get_the_title( $post ) ) ) );
254	if ( '' === $title ) {
255	$title = __( '(untitled)', 'jetpack-seo' );
256	}
257	// Escape brackets so a title can't break the `[title](url)` link syntax.
258	$title = str_replace( array( '[', ']' ), array( '\[', '\]' ), $title );
259	$url = esc_url_raw( (string) get_permalink( $post ) );
260	$line = '- [' . $title . '](' . $url . ')';
261
262	$summary = self::summary( $post );
263	if ( '' !== $summary ) {
264	$line .= ': ' . $summary;
265	}
266	$lines[] = $line;
267	}
268	return implode( "\n", $lines );
269	}
270
271	/**
272	* A short, single-line summary for a post: its excerpt when set, otherwise a
273	* trimmed slice of the content.
274	*
275	* @param \WP_Post $post Post to summarize.
276	* @return string
277	*/
278	private static function summary( $post ) {
279	if ( has_excerpt( $post ) ) {
280	$raw = get_the_excerpt( $post );
281	} elseif (
282	get_post_meta( $post->ID, '_jetpack_memberships_contains_paid_content', true )
283	\|\| get_post_meta( $post->ID, '_jetpack_memberships_contains_paywalled_content', true )
284	\|\| has_block( 'premium-content/container', $post )
285	\|\| has_block( 'jetpack/paywall', $post )
286	) {
287	return '';
288	} else {
289	$raw = wp_trim_words( wp_strip_all_tags( strip_shortcodes( $post->post_content ) ), 30, '' );
290	}
291
292	// Collapse whitespace so a link line stays on one row.
293	return trim( preg_replace( '/\s+/', ' ', wp_strip_all_tags( $raw ) ) );
294	}
295	}
296