← All changes | admin/class-knowledge-manager.php +26 -101 3.2.1 → 3.1.5 View file →

		@@ -163,13 +163,9 @@
163	163	public function mxchat_is_pdf_url($url, $response) {
164	164	$content_type = wp_remote_retrieve_header($response, 'content-type');
165	165	$file_extension = strtolower(pathinfo($url, PATHINFO_EXTENSION));
166	166
167		- // Check Content-Disposition header for .pdf filename (Google Drive sends this)
168		- $disposition = wp_remote_retrieve_header($response, 'content-disposition');
169		- $has_pdf_disposition = ! empty($disposition) && stripos($disposition, '.pdf') !== false;
170		-
171		- return strpos($content_type, 'pdf') !== false || $file_extension === 'pdf' || $has_pdf_disposition;
	167	+ return strpos($content_type, 'pdf') !== false || $file_extension === 'pdf';
172	168	}
173	169
174	170
175	171	public function mxchat_handle_pdf_for_knowledge_base($pdf_url, $response, $bot_id = 'default') {
		@@ -863,17 +859,10 @@
863	859	}
864	860
865	861	// For manual entries (no source_url or mxchat:// prefix), delete the old entry by ID first
866	862	// so submit_content_to_db creates a replacement instead of a duplicate
867		- // Also treat legacy mxchat.ai source URLs as manual — old bug assigned the site URL to manual entries
868		- $is_legacy_manual = !empty($source_url) && strpos($source_url, 'mxchat.ai') !== false && strpos($source_url, 'mxchat://') !== 0;
869		- if ( $entry_id > 0 && (empty($source_url) || strpos($source_url, 'mxchat://') === 0 || $is_legacy_manual) ) {
	863	+ if ( $entry_id > 0 && (empty($source_url) || strpos($source_url, 'mxchat://') === 0) ) {
870	864	$wpdb->delete( $table, array( 'id' => $entry_id ), array( '%d' ) );
871		- // Clear legacy URL so submit_content_to_db generates a unique mxchat:// identifier
872		- // instead of reusing the shared URL (which would mass-delete other entries with the same URL)
873		- if ( $is_legacy_manual ) {
874		- $source_url = '';
875		- }
876	865	}
877	866
878	867	// Use the existing submit_content_to_db which handles chunking, Pinecone, and WP DB
879	868	$vector_id = ! empty($source_url) ? md5($source_url) : md5('mxchat_manual_' . $entry_id);
		@@ -950,22 +939,9 @@
950	939	exit;
951	940	}
952	941
953	942	$submitted_url = esc_url_raw($_POST['sitemap_url']);
954		-
955		- // Convert Google Drive sharing URLs to direct download URLs
956		- if ( strpos($submitted_url, 'drive.google.com') !== false ) {
957		- $file_id = '';
958		- if ( preg_match('/[?&]id=([a-zA-Z0-9_-]+)/', $submitted_url, $m) ) {
959		- $file_id = $m[1];
960		- } elseif ( preg_match('#/file/d/([a-zA-Z0-9_-]+)#', $submitted_url, $m) ) {
961		- $file_id = $m[1];
962		- }
963		- if ( ! empty($file_id) ) {
964		- $submitted_url = 'https://drive.google.com/uc?export=download&id=' . $file_id;
965		- }
966		- }
967		-
	943	+
968	944	// Get bot_id from form submission
969	945	$bot_id = isset($_POST['bot_id']) ? sanitize_key($_POST['bot_id']) : 'default';
970	946
971	947	// Get bot-specific options and validate API key
		@@ -993,18 +969,10 @@
993	969	wp_safe_redirect(esc_url(admin_url('admin.php?page=mxchat-prompts')));
994	970	exit;
995	971	}
996	972
997		- // Fetch URL — use browser-like headers so servers with bot protection don't block us
998		- $response = wp_remote_get($submitted_url, array(
999		- 'timeout' => 30,
1000		- 'sslverify' => false,
1001		- 'user-agent' => 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36',
1002		- 'headers' => array(
1003		- 'Accept' => 'text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8',
1004		- 'Accept-Language' => 'en-US,en;q=0.9',
1005		- ),
1006		- ));
	973	+ // Fetch URL
	974	+ $response = wp_remote_get($submitted_url, array('timeout' => 30));
1007	975
1008	976	if (is_wp_error($response) || wp_remote_retrieve_response_code($response) !== 200) {
1009	977	$error_message = is_wp_error($response) ? $response->get_error_message() : 'HTTP Status: ' . wp_remote_retrieve_response_code($response);
1010	978	set_transient('mxchat_admin_notice_error',
		@@ -2838,12 +2806,11 @@
2838	2806	foreach ($primary_indexes as $path => $source) {
2839	2807	$url = trailingslashit($site_url) . $path;
2840	2808
2841	2809	$response = wp_remote_head($url, array(
2842		- 'timeout' => 10,
	2810	+ 'timeout' => 3, // Short timeout
2843	2811	'sslverify' => false,
2844		- 'redirection' => 1,
2845		- 'user-agent' => 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36',
	2812	+ 'redirection' => 1
2846	2813	));
2847	2814
2848	2815	if (!is_wp_error($response) && wp_remote_retrieve_response_code($response) === 200) {
2849	2816	// Found a sitemap index - parse it to get sub-sitemaps
		@@ -2901,15 +2868,10 @@
2901	2868	private function parse_sitemap_index($url) {
2902	2869	$sub_sitemaps = array();
2903	2870
2904	2871	$response = wp_remote_get($url, array(
2905		- 'timeout' => 30,
2906		- 'sslverify' => false,
2907		- 'user-agent' => 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36',
2908		- 'headers' => array(
2909		- 'Accept' => 'text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8',
2910		- 'Accept-Language' => 'en-US,en;q=0.9',
2911		- ),
	2872	+ 'timeout' => 5,
	2873	+ 'sslverify' => false
2912	2874	));
2913	2875
2914	2876	if (is_wp_error($response)) {
2915	2877	return $sub_sitemaps;
		@@ -2960,15 +2922,10 @@
2960	2922	* Get URL count from a sitemap
2961	2923	*/
2962	2924	private function get_sitemap_url_count($url) {
2963	2925	$response = wp_remote_get($url, array(
2964		- 'timeout' => 30,
2965		- 'sslverify' => false,
2966		- 'user-agent' => 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36',
2967		- 'headers' => array(
2968		- 'Accept' => 'text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8',
2969		- 'Accept-Language' => 'en-US,en;q=0.9',
2970		- ),
	2926	+ 'timeout' => 10,
	2927	+ 'sslverify' => false
2971	2928	));
2972	2929
2973	2930	if (is_wp_error($response)) {
2974	2931	return 0;
		@@ -2991,11 +2948,10 @@
2991	2948	$sitemaps = array();
2992	2949	$robots_url = trailingslashit($site_url) . 'robots.txt';
2993	2950
2994	2951	$response = wp_remote_get($robots_url, array(
2995		- 'timeout' => 15,
2996		- 'sslverify' => false,
2997		- 'user-agent' => 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36',
	2952	+ 'timeout' => 5,
	2953	+ 'sslverify' => false
2998	2954	));
2999	2955
3000	2956	if (is_wp_error($response)) {
3001	2957	return $sitemaps;
		@@ -7121,37 +7077,11 @@
7121	7077	throw new Exception('Unknown item type: ' . $item_type);
7122	7078	}
7123	7079
7124	7080	if (is_wp_error($result)) {
7125		- $error_code = $result->get_error_code();
7126		- // Content errors (empty page, sanitization) are permanent — retrying won't help
7127		- $permanent_codes = array('empty_page', 'empty_after_sanitization', 'no_api_key', 'page_not_found');
7128		- if (in_array($error_code, $permanent_codes)) {
7129		- // Mark as permanently failed — set attempts = max_attempts so it won't be retried
7130		- $current_item = $wpdb->get_row($wpdb->prepare(
7131		- "SELECT max_attempts FROM $table_name WHERE id = %d", $item_id
7132		- ));
7133		- $wpdb->update(
7134		- $table_name,
7135		- array(
7136		- 'status' => 'failed',
7137		- 'error_message' => $result->get_error_message(),
7138		- 'attempts' => $current_item ? $current_item->max_attempts : 3
7139		- ),
7140		- array('id' => $item_id),
7141		- array('%s', '%s', '%d'),
7142		- array('%d')
7143		- );
7144		- wp_send_json_error(array(
7145		- 'message' => $result->get_error_message(),
7146		- 'permanent_failure' => true,
7147		- 'item_id' => $item_id
7148		- ));
7149		- return;
7150		- }
7151	7081	throw new Exception($result->get_error_message());
7152	7082	}
7153		-
	7083	+
7154	7084	if ($result === false) {
7155	7085	throw new Exception('Processing returned false - item may be empty or invalid');
7156	7086	}
7157	7087
		@@ -7441,24 +7371,21 @@
7441	7371	return new WP_Error('no_pages', 'PDF has no pages');
7442	7372	}
7443	7373
7444	7374	$processed = 0;
7445		- $skipped_pages = array();
7446	7375
7447	7376	for ($i = 0; $i < $total_pages; $i++) {
7448		- $page_num = $i + 1;
7449	7377	$text = $pages[$i]->getText();
7450	7378	if (empty($text)) {
7451		- $skipped_pages[] = 'Page ' . $page_num . ': No text could be extracted — page may contain only images, links, or non-standard encoding';
7452	7379	continue;
7453	7380	}
7454	7381
7455	7382	$sanitized = $this->mxchat_sanitize_content_for_api($text);
7456	7383	if (empty($sanitized)) {
7457		- $skipped_pages[] = 'Page ' . $page_num . ': Text was extracted but contained only special characters, control codes, or unsupported content';
7458	7384	continue;
7459	7385	}
7460	7386
	7387	+ $page_num = $i + 1;
7461	7388	$metadata = array(
7462	7389	'document_type' => 'pdf',
7463	7390	'total_pages' => $total_pages,
7464	7391	'current_page' => $page_num,
		@@ -7482,12 +7409,8 @@
7482	7409
7483	7410	// Clean up the temp PDF file
7484	7411	wp_delete_file($pdf_path);
7485	7412
7486		- if (!empty($skipped_pages)) {
7487		- error_log('MxChat PDF: Skipped ' . count($skipped_pages) . ' of ' . $total_pages . ' pages: ' . implode('; ', $skipped_pages));
7488		- }
7489		-
7490	7413	return $processed > 0 ? true : false;
7491	7414
7492	7415	} catch (Exception $e) {
7493	7416	if (file_exists($pdf_path)) {
		@@ -7652,17 +7575,18 @@
7652	7575	return new WP_Error('page_not_found', 'Page ' . $page_number . ' not found in PDF');
7653	7576	}
7654	7577
7655	7578	$text = $pages[$page_number - 1]->getText();
7656		-
	7579	+
7657	7580	if (empty($text)) {
7658		- return new WP_Error('empty_page', 'Page ' . $page_number . ': No text could be extracted — page may contain only images, links, or non-standard encoding');
	7581	+ // Not an error - just an empty page
	7582	+ return false;
7659	7583	}
7660		-
	7584	+
7661	7585	$sanitized = $this->mxchat_sanitize_content_for_api($text);
7662		-
	7586	+
7663	7587	if (empty($sanitized)) {
7664		- return new WP_Error('empty_after_sanitization', 'Page ' . $page_number . ': Text was extracted but contained only special characters, control codes, or unsupported content that was removed during cleanup');
	7588	+ return false;
7665	7589	}
7666	7590
7667	7591	// Create metadata
7668	7592	$metadata = array(
		@@ -7766,16 +7690,17 @@
7766	7690
7767	7691	// Calculate percentage
7768	7692	$percentage = $total > 0 ? round((($completed + $failed) / $total) * 100) : 0;
7769	7693
7770		- // Get failed items details (include all failed items, not just those that exhausted retries)
	7694	+ // Get failed items details
7771	7695	$failed_items = array();
7772	7696	if ($failed > 0) {
7773	7697	$failed_items = $wpdb->get_results($wpdb->prepare(
7774		- "SELECT item_type, item_data, error_message, attempts
7775		- FROM $table_name
7776		- WHERE queue_id = %s
	7698	+ "SELECT item_type, item_data, error_message, attempts
	7699	+ FROM $table_name
	7700	+ WHERE queue_id = %s
7777	7701	AND status = 'failed'
	7702	+ AND attempts >= max_attempts
7778	7703	ORDER BY id DESC
7779	7704	LIMIT 50",
7780	7705	$queue_id
7781	7706	));