<?php
/**
 * AI access fixes: the things the AI Readiness report flags that a WordPress
 * site can fix from inside WordPress.
 *
 *  1. /llms.txt   The report offers a one-click "Publish an llms.txt file" fix.
 *                 The content is built on our server from the Entity Profile
 *                 and the synced page list (GET /sites/:id/llms-txt), and this
 *                 class serves it at the site root. On by default. A real
 *                 llms.txt file on disk always wins, because the web server
 *                 answers before WordPress runs.
 *  2. robots.txt  An opt-in switch that adds explicit "Allow: /" groups for
 *                 the main AI crawlers to WordPress's own robots.txt, plus a
 *                 status check of which AI crawlers the live robots.txt blocks.
 *                 It cannot change a physical robots.txt file or a firewall
 *                 rule (Cloudflare's "Block AI bots"), and says so.
 *  3. A warning when Settings > Reading > "Discourage search engines" is on,
 *     which hides the whole site from search and AI crawlers alike, with a
 *     one-click fix and a way to dismiss it for staging sites.
 *
 * Settings live in schema_ai_settings: llms_txt_enabled (missing = on) and
 * allow_ai_crawlers (missing = off). Turning either off restores WordPress's
 * normal behaviour at once.
 *
 * @package Schema_AI
 * @since   1.0.36
 */

if ( ! defined( 'ABSPATH' ) ) {
	exit;
}

class Schema_AI_AI_Access {

	const LLMS_TRANSIENT      = 'schema_ai_llms_txt';
	const LLMS_LAST_GOOD      = 'schema_ai_llms_txt_last';
	const ROBOTS_STATUS_CACHE = 'schema_ai_ai_crawler_status';
	const HIDDEN_DISMISSED    = 'schema_ai_hidden_notice_dismissed';

	/**
	 * The AI crawlers we check and, when switched on, allow. User-agent token => plain name.
	 *
	 * @return array<string,string>
	 */
	public static function crawlers() {
		return array(
			'GPTBot'           => __( 'ChatGPT (training)', 'ai-schema-gen' ),
			'OAI-SearchBot'    => __( 'ChatGPT search', 'ai-schema-gen' ),
			'ChatGPT-User'     => __( 'ChatGPT (pages a user asks it to open)', 'ai-schema-gen' ),
			'ClaudeBot'        => __( 'Claude (training)', 'ai-schema-gen' ),
			'Claude-SearchBot' => __( 'Claude search', 'ai-schema-gen' ),
			'Claude-User'      => __( 'Claude (pages a user asks it to open)', 'ai-schema-gen' ),
			'PerplexityBot'    => __( 'Perplexity search', 'ai-schema-gen' ),
			'Perplexity-User'  => __( 'Perplexity (pages a user asks it to open)', 'ai-schema-gen' ),
			'Google-Extended'  => __( 'Google Gemini (training)', 'ai-schema-gen' ),
		);
	}

	/**
	 * Hook everything up. Called once from Schema_AI_Gen::define_public_hooks().
	 */
	public function register() {
		add_action( 'init', array( $this, 'maybe_serve_llms_txt' ), 1 );
		add_filter( 'robots_txt', array( $this, 'filter_robots_txt' ), 99, 2 );

		add_action( 'admin_notices', array( $this, 'hidden_site_notice' ) );
		add_action( 'admin_post_schema_ai_make_visible', array( $this, 'handle_make_visible' ) );
		add_action( 'admin_post_schema_ai_dismiss_hidden', array( $this, 'handle_dismiss_hidden' ) );

		// A settings save may have switched either feature, so re-read fresh.
		add_action( 'update_option_schema_ai_settings', array( __CLASS__, 'clear_caches' ) );
	}

	/** @return array */
	private static function settings() {
		$settings = get_option( 'schema_ai_settings', array() );
		return is_array( $settings ) ? $settings : array();
	}

	/** @return bool */
	public static function llms_enabled() {
		$settings = self::settings();
		return ! array_key_exists( 'llms_txt_enabled', $settings ) || ! empty( $settings['llms_txt_enabled'] );
	}

	/** @return bool */
	public static function crawlers_allowed() {
		$settings = self::settings();
		return ! empty( $settings['allow_ai_crawlers'] );
	}

	/** @return bool True when a real file sits where we would serve one. */
	public static function physical_file_exists( $name ) {
		return file_exists( trailingslashit( ABSPATH ) . $name );
	}

	public static function clear_caches() {
		delete_transient( self::LLMS_TRANSIENT );
		delete_transient( self::ROBOTS_STATUS_CACHE );
	}

	// ── 1. llms.txt ─────────────────────────────────────────────────────────

	/**
	 * Serve /llms.txt when that is what was asked for. Runs on init, before
	 * WordPress decides the URL is a 404, and before any theme output.
	 */
	public function maybe_serve_llms_txt() {
		if ( is_admin() || ! self::llms_enabled() || empty( $_SERVER['REQUEST_URI'] ) ) {
			return;
		}
		$path = wp_parse_url( esc_url_raw( wp_unslash( $_SERVER['REQUEST_URI'] ) ), PHP_URL_PATH );
		if ( ! is_string( $path ) || strtolower( $path ) !== strtolower( self::llms_path() ) ) {
			return;
		}

		$content = self::get_llms_content();
		if ( '' === $content ) {
			return; // Nothing useful to publish yet: let WordPress answer 404 as before.
		}

		status_header( 200 );
		header( 'Content-Type: text/plain; charset=utf-8' );
		header( 'Cache-Control: public, max-age=3600' );
		echo $content; // phpcs:ignore WordPress.Security.EscapeOutput.OutputNotEscaped -- plain-text file body, served as text/plain.
		exit;
	}

	/** @return string The path this site's llms.txt lives at, subdirectory installs included. */
	public static function llms_path() {
		$home_path = wp_parse_url( home_url( '/' ), PHP_URL_PATH );
		return trailingslashit( is_string( $home_path ) && '' !== $home_path ? $home_path : '/' ) . 'llms.txt';
	}

	/**
	 * The llms.txt body, cached for 12 hours. Empty string means "nothing to
	 * serve". A failed fetch falls back to the last good copy so a network blip
	 * never takes a working file offline.
	 *
	 * @return string
	 */
	public static function get_llms_content() {
		$cached = get_transient( self::LLMS_TRANSIENT );
		if ( false !== $cached ) {
			return (string) $cached;
		}

		$settings = self::settings();
		$api_key  = ! empty( $settings['api_key'] ) ? $settings['api_key'] : '';
		$site_id  = (int) get_option( 'schema_ai_saas_site_id', 0 );
		if ( '' === $api_key || $site_id <= 0 ) {
			set_transient( self::LLMS_TRANSIENT, '', HOUR_IN_SECONDS );
			return '';
		}

		$response = wp_remote_get(
			SCHEMA_AI_API_BASE . '/sites/' . $site_id . '/llms-txt',
			array(
				'timeout' => 8,
				'headers' => array( 'X-API-Key' => $api_key ),
			)
		);

		if ( is_wp_error( $response ) || 200 !== wp_remote_retrieve_response_code( $response ) ) {
			$last = (string) get_option( self::LLMS_LAST_GOOD, '' );
			set_transient( self::LLMS_TRANSIENT, $last, 15 * MINUTE_IN_SECONDS );
			return $last;
		}

		$body    = json_decode( wp_remote_retrieve_body( $response ), true );
		$content = isset( $body['data']['content'] ) && is_string( $body['data']['content'] ) ? $body['data']['content'] : '';

		if ( '' !== $content ) {
			update_option( self::LLMS_LAST_GOOD, $content, false );
		}
		set_transient( self::LLMS_TRANSIENT, $content, '' !== $content ? 12 * HOUR_IN_SECONDS : HOUR_IN_SECONDS );
		return $content;
	}

	// ── 2. AI crawlers in robots.txt ────────────────────────────────────────

	/**
	 * Add explicit allow groups for the AI crawlers to WordPress's virtual
	 * robots.txt. Left alone when the switch is off, or when the site is set
	 * to discourage search engines (that choice is the owner's to undo).
	 *
	 * @param string $output Robots.txt body so far.
	 * @param bool|string $is_public Whether the site allows indexing (WordPress passes a bool).
	 * @return string
	 */
	public function filter_robots_txt( $output, $is_public ) {
		if ( ! self::crawlers_allowed() || ! $is_public ) {
			return $output;
		}
		$lines   = array( '', '# Added by AI Schema Gen: let AI assistants read this site' );
		foreach ( array_keys( self::crawlers() ) as $agent ) {
			$lines[] = 'User-agent: ' . $agent;
			$lines[] = 'Allow: /';
			$lines[] = '';
		}
		return rtrim( (string) $output ) . "\n" . implode( "\n", $lines );
	}

	/**
	 * Which AI crawlers the live robots.txt blocks, read the way a crawler
	 * reads it: its own group if it has one, otherwise the "*" group. Cached
	 * for 10 minutes. Only whole-site blocks ("Disallow: /") count.
	 *
	 * @return array{fetched:bool,physical:bool,crawlers:array<string,string>}
	 */
	public static function crawler_status() {
		$cached = get_transient( self::ROBOTS_STATUS_CACHE );
		if ( is_array( $cached ) ) {
			return $cached;
		}

		$result = array(
			'fetched'  => false,
			'physical' => self::physical_file_exists( 'robots.txt' ),
			'crawlers' => array(),
		);

		$response = wp_remote_get( home_url( '/robots.txt' ), array( 'timeout' => 8 ) );
		if ( ! is_wp_error( $response ) && 200 === wp_remote_retrieve_response_code( $response ) ) {
			$result['fetched'] = true;
			$groups            = self::parse_robots( (string) wp_remote_retrieve_body( $response ) );
			foreach ( array_keys( self::crawlers() ) as $agent ) {
				$result['crawlers'][ $agent ] = self::is_blocked( $groups, $agent ) ? 'blocked' : 'allowed';
			}
		}

		set_transient( self::ROBOTS_STATUS_CACHE, $result, 10 * MINUTE_IN_SECONDS );
		return $result;
	}

	/**
	 * @param string $body robots.txt text.
	 * @return array<int,array{agents:string[],allow_root:bool,disallow_root:bool}>
	 */
	private static function parse_robots( $body ) {
		$groups  = array();
		$current = null;
		$in_ua   = false;
		foreach ( preg_split( '/\r\n|\r|\n/', $body ) as $raw ) {
			$line = trim( preg_replace( '/#.*$/', '', $raw ) );
			if ( '' === $line || false === strpos( $line, ':' ) ) {
				continue;
			}
			list( $field, $value ) = array_map( 'trim', explode( ':', $line, 2 ) );
			$field                 = strtolower( $field );
			if ( 'user-agent' === $field ) {
				if ( ! $in_ua || null === $current ) {
					$groups[] = array( 'agents' => array(), 'allow_root' => false, 'disallow_root' => false );
					$current  = count( $groups ) - 1;
				}
				$groups[ $current ]['agents'][] = strtolower( $value );
				$in_ua                          = true;
				continue;
			}
			$in_ua = false;
			if ( null === $current ) {
				continue;
			}
			if ( 'disallow' === $field && '/' === $value ) {
				$groups[ $current ]['disallow_root'] = true;
			} elseif ( 'allow' === $field && '/' === $value ) {
				$groups[ $current ]['allow_root'] = true;
			}
		}
		return $groups;
	}

	private static function is_blocked( array $groups, $agent ) {
		$agent    = strtolower( $agent );
		$matching = array_filter(
			$groups,
			function ( $g ) use ( $agent ) {
				return in_array( $agent, $g['agents'], true );
			}
		);
		if ( empty( $matching ) ) {
			$matching = array_filter(
				$groups,
				function ( $g ) {
					return in_array( '*', $g['agents'], true );
				}
			);
		}
		$disallow = false;
		$allow    = false;
		foreach ( $matching as $g ) {
			$disallow = $disallow || $g['disallow_root'];
			$allow    = $allow || $g['allow_root'];
		}
		return $disallow && ! $allow;
	}

	// ── 3. "Discourage search engines" warning ──────────────────────────────

	public function hidden_site_notice() {
		if ( '0' !== (string) get_option( 'blog_public' ) || ! current_user_can( 'manage_options' ) || get_option( self::HIDDEN_DISMISSED ) ) {
			return;
		}
		$screen = function_exists( 'get_current_screen' ) ? get_current_screen() : null;
		$id     = $screen ? (string) $screen->id : '';
		if ( 'dashboard' !== $id && false === strpos( $id, 'ai-schema-gen' ) && 'options-reading' !== $id ) {
			return;
		}
		$fix_url     = wp_nonce_url( admin_url( 'admin-post.php?action=schema_ai_make_visible' ), 'schema_ai_make_visible' );
		$dismiss_url = wp_nonce_url( admin_url( 'admin-post.php?action=schema_ai_dismiss_hidden' ), 'schema_ai_dismiss_hidden' );
		?>
		<div class="notice notice-warning">
			<p>
				<strong><?php esc_html_e( 'AI Schema Gen: your site is hidden from search engines and AI assistants.', 'ai-schema-gen' ); ?></strong>
				<?php esc_html_e( '"Discourage search engines from indexing this site" is switched on in Settings > Reading, so Google, ChatGPT, Perplexity and others are told not to read it.', 'ai-schema-gen' ); ?>
			</p>
			<p>
				<a href="<?php echo esc_url( $fix_url ); ?>" class="button button-primary"><?php esc_html_e( 'Make my site visible', 'ai-schema-gen' ); ?></a>
				<a href="<?php echo esc_url( $dismiss_url ); ?>" class="button"><?php esc_html_e( 'This is on purpose (staging site)', 'ai-schema-gen' ); ?></a>
			</p>
		</div>
		<?php
	}

	public function handle_make_visible() {
		if ( ! current_user_can( 'manage_options' ) ) {
			wp_die( esc_html__( 'You are not allowed to do that.', 'ai-schema-gen' ) );
		}
		check_admin_referer( 'schema_ai_make_visible' );
		update_option( 'blog_public', '1' );
		self::clear_caches();
		wp_safe_redirect( wp_get_referer() ? wp_get_referer() : admin_url() );
		exit;
	}

	public function handle_dismiss_hidden() {
		if ( ! current_user_can( 'manage_options' ) ) {
			wp_die( esc_html__( 'You are not allowed to do that.', 'ai-schema-gen' ) );
		}
		check_admin_referer( 'schema_ai_dismiss_hidden' );
		update_option( self::HIDDEN_DISMISSED, 1, false );
		wp_safe_redirect( wp_get_referer() ? wp_get_referer() : admin_url() );
		exit;
	}
}