<?php
/**
 * Crawler Controls: per-AI-bot allow/block matrix enforced via robots.txt.
 *
 * @package AiRankSuite
 */

declare(strict_types=1);

namespace AiRankSuite\Modules\CrawlerControls;

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

use AiRankSuite\Container;
use AiRankSuite\Contracts\ModuleInterface;
use AiRankSuite\Support\Options;
use AiRankSuite\Support\Guard;

final class CrawlerControlsModule implements ModuleInterface {

	public function id(): string {
		return 'crawler_controls';
	}

	/**
	 * Known AI crawler registry (updated with plugin releases; filterable).
	 *
	 * @return array<string, array<string, string>>
	 */
	public static function bots(): array {
		$bots = array(
			// ---- United States ------------------------------------------
			'gptbot'            => array( 'ua' => 'GPTBot', 'vendor' => 'OpenAI', 'purpose' => 'training', 'region' => 'us' ),
			'oai-searchbot'     => array( 'ua' => 'OAI-SearchBot', 'vendor' => 'OpenAI', 'purpose' => 'search', 'region' => 'us' ),
			'chatgpt-user'      => array( 'ua' => 'ChatGPT-User', 'vendor' => 'OpenAI', 'purpose' => 'on-demand', 'region' => 'us' ),
			'claudebot'         => array( 'ua' => 'ClaudeBot', 'vendor' => 'Anthropic', 'purpose' => 'training', 'region' => 'us' ),
			'claude-searchbot'  => array( 'ua' => 'Claude-SearchBot', 'vendor' => 'Anthropic', 'purpose' => 'search', 'region' => 'us' ),
			'claude-user'       => array( 'ua' => 'Claude-User', 'vendor' => 'Anthropic', 'purpose' => 'on-demand', 'region' => 'us' ),
			'perplexitybot'     => array( 'ua' => 'PerplexityBot', 'vendor' => 'Perplexity', 'purpose' => 'search', 'region' => 'us' ),
			'perplexity-user'   => array( 'ua' => 'Perplexity-User', 'vendor' => 'Perplexity', 'purpose' => 'on-demand', 'region' => 'us' ),
			'google-extended'   => array( 'ua' => 'Google-Extended', 'vendor' => 'Google', 'purpose' => 'training', 'region' => 'us' ),
			'google-notebooklm' => array( 'ua' => 'Google-NotebookLM', 'vendor' => 'Google', 'purpose' => 'on-demand', 'region' => 'us' ),
			'google-vertex'     => array( 'ua' => 'Google-CloudVertexBot', 'vendor' => 'Google', 'purpose' => 'training', 'region' => 'us' ),
			'applebot-extended' => array( 'ua' => 'Applebot-Extended', 'vendor' => 'Apple', 'purpose' => 'training', 'region' => 'us' ),
			'meta-external'     => array( 'ua' => 'meta-externalagent', 'vendor' => 'Meta', 'purpose' => 'training', 'region' => 'us' ),
			'meta-fetcher'      => array( 'ua' => 'Meta-ExternalFetcher', 'vendor' => 'Meta', 'purpose' => 'on-demand', 'region' => 'us' ),
			'amazonbot'         => array( 'ua' => 'Amazonbot', 'vendor' => 'Amazon', 'purpose' => 'search', 'region' => 'us' ),
			'bedrockbot'        => array( 'ua' => 'bedrockbot', 'vendor' => 'Amazon', 'purpose' => 'training', 'region' => 'us' ),
			'cohere-ai'         => array( 'ua' => 'cohere-ai', 'vendor' => 'Cohere', 'purpose' => 'training', 'region' => 'us' ),
			'duckassistbot'     => array( 'ua' => 'DuckAssistBot', 'vendor' => 'DuckDuckGo', 'purpose' => 'search', 'region' => 'us' ),
			'ai2bot'            => array( 'ua' => 'AI2Bot', 'vendor' => 'Allen Institute', 'purpose' => 'training', 'region' => 'us' ),
			'ccbot'             => array( 'ua' => 'CCBot', 'vendor' => 'Common Crawl', 'purpose' => 'training', 'region' => 'us' ),

			// ---- Europe -------------------------------------------------
			'mistralai-user'    => array( 'ua' => 'MistralAI-User', 'vendor' => 'Mistral AI', 'purpose' => 'on-demand', 'region' => 'eu' ),
			'yandexadditional'  => array( 'ua' => 'YandexAdditional', 'vendor' => 'Yandex', 'purpose' => 'training', 'region' => 'eu' ),

			// ---- China --------------------------------------------------
			// The Chinese assistants are the fastest-growing source of AI
			// answers outside the US, and several of them crawl far more
			// aggressively than their Western counterparts.
			'bytespider'        => array( 'ua' => 'Bytespider', 'vendor' => 'ByteDance (Doubao)', 'purpose' => 'training', 'region' => 'cn' ),
			'tiktokspider'      => array( 'ua' => 'TikTokSpider', 'vendor' => 'ByteDance', 'purpose' => 'training', 'region' => 'cn' ),
			'petalbot'          => array( 'ua' => 'PetalBot', 'vendor' => 'Huawei (Petal / Xiaoyi)', 'purpose' => 'search', 'region' => 'cn' ),
			'pangubot'          => array( 'ua' => 'PanguBot', 'vendor' => 'Huawei (Pangu)', 'purpose' => 'training', 'region' => 'cn' ),
			'deepseekbot'       => array(
				'ua'      => 'DeepSeekBot',
				'vendor'  => 'DeepSeek',
				'purpose' => 'training',
				'region'  => 'cn',
				// Honest caveat: DeepSeek is widely reported to fetch without a
				// named identifier, so this rule is a best effort. Telling a
				// customer they have "blocked DeepSeek" would be a promise the
				// protocol cannot keep.
				'note'    => 'unreliable',
			),
			'baiduspider'       => array( 'ua' => 'Baiduspider', 'vendor' => 'Baidu (ERNIE)', 'purpose' => 'search', 'region' => 'cn' ),
			'yisouspider'       => array( 'ua' => 'YisouSpider', 'vendor' => 'Alibaba / UC (Quark, Qwen)', 'purpose' => 'search', 'region' => 'cn' ),
			'sogouspider'       => array( 'ua' => 'Sogou web spider', 'vendor' => 'Tencent / Sogou (Yuanbao)', 'purpose' => 'search', 'region' => 'cn' ),
			'360spider'         => array( 'ua' => '360Spider', 'vendor' => '360 (Nano AI)', 'purpose' => 'search', 'region' => 'cn' ),
		);
		/**
		 * Filter the AI bot registry.
		 *
		 * @param array $bots Bot definitions keyed by id.
		 */
		return (array) apply_filters( 'airanksuite/crawler_controls/bots', $bots );
	}

	/**
	 * Default settings: everything allowed (site owner opts into blocking).
	 *
	 * @return array<string, mixed>
	 */
	public static function defaults(): array {
		return array(
			'enabled' => true,
			'blocked' => array(),
			// Record which crawlers arrive. Off by default: it writes an
			// option per crawler request, and that is a decision to make
			// knowingly rather than one a plugin makes for you.
			'log'     => false,
			// robots.txt is a request, not a fence. `enforce` refuses the
			// request outright for bots that ignore it.
			'enforce' => false,
		);
	}

	public function register( Container $container ): void {}

	public function boot( Container $container ): void {
		Guard::filter( 'robots_txt', array( $this, 'robots_rules' ), 20, 1 );

		// Front-end only, and as early as a request can be identified. Not on
		// admin, cron or REST: none of those are ever a crawler, and checking
		// on every one of them would be a tax on the admin for nothing.
		Guard::action( 'parse_request', array( $this, 'watch' ), 1 );
		add_action(
			'rest_api_init',
			static function () use ( $container ): void {
				( new Rest\CrawlersController( $container->get( Options::class ) ) )->register_routes();
			}
		);
	}

	/**
	 * Note an AI crawler's visit, and turn it away if asked to.
	 *
	 * This is the difference between a robots.txt rule and an actual block.
	 * A crawler that ignores robots.txt — and several on the list are widely
	 * reported to — is unaffected by every switch on the settings screen. A
	 * 403 is not.
	 */
	public function watch(): void {
		if ( is_admin() || wp_doing_cron() || wp_doing_ajax() ) {
			return;
		}
		if ( defined( 'REST_REQUEST' ) && REST_REQUEST ) {
			return;
		}

		$settings = \AiRankSuite\Kernel::service( Options::class )->get( 'crawler_controls', self::defaults() );
		if ( empty( $settings['log'] ) && empty( $settings['enforce'] ) ) {
			return;
		}

		$agent = isset( $_SERVER['HTTP_USER_AGENT'] )
			? sanitize_text_field( wp_unslash( (string) $_SERVER['HTTP_USER_AGENT'] ) )
			: '';
		$bot = Services\CrawlerLog::identify( $agent );
		if ( null === $bot ) {
			return;
		}

		$blocked = in_array( $bot, (array) $settings['blocked'], true );
		$refuse  = $blocked && ! empty( $settings['enforce'] ) && ! empty( $settings['enabled'] );

		$path = isset( $_SERVER['REQUEST_URI'] )
			? esc_url_raw( wp_unslash( (string) $_SERVER['REQUEST_URI'] ) )
			: '/';

		if ( ! empty( $settings['log'] ) ) {
			Services\CrawlerLog::record( $bot, $path, $refuse ? 403 : 200 );
		}

		if ( $refuse ) {
			// 403 rather than 404: a crawler told the page does not exist may
			// drop the URL permanently, which is a different and worse outcome
			// than being told it is not allowed to have it.
			status_header( 403 );
			nocache_headers();
			header( 'Content-Type: text/plain; charset=utf-8' );
			echo esc_html__( 'This crawler is not permitted on this site.', 'ai-rank-suite' );
			exit;
		}
	}

	/**
	 * Append managed per-bot rules to the virtual robots.txt.
	 *
	 * Note: a physical robots.txt file bypasses this filter — the admin UI
	 * warns when one exists.
	 *
	 * @param string $output Current robots.txt content.
	 */
	public function robots_rules( string $output ): string {
		$container = \AiRankSuite\Kernel::instance()->container();
		$settings  = $container->get( Options::class )->get( 'crawler_controls', self::defaults() );
		if ( empty( $settings['enabled'] ) ) {
			return $output;
		}

		$blocked = (array) $settings['blocked'];
		if ( empty( $blocked ) ) {
			return $output;
		}

		$lines   = array( '', '# Managed by AI Rank Suite — AI crawler controls' );
		$bots    = self::bots();
		foreach ( $blocked as $bot_id ) {
			if ( ! isset( $bots[ $bot_id ] ) ) {
				continue;
			}
			$lines[] = 'User-agent: ' . $bots[ $bot_id ]['ua'];
			$lines[] = 'Disallow: /';
			$lines[] = '';
		}

		return $output . implode( "\n", $lines );
	}
}