PluginProbe
MxChat – AI Chatbot & Content Generation for WordPress / 3.2.2
MxChat – AI Chatbot & Content Generation for WordPress v3.2.2
3.2.21 3.2.20 3.2.19 3.2.18 3.2.17 3.2.16 3.2.15 3.2.14 3.2.12 3.2.13 3.2.11 3.2.10 3.2.9 3.2.8 3.2.7 3.2.6 3.2.5 3.2.4 3.2.3 3.2.2 3.2.1 2.0.3 2.0.4 2.0.5 2.0.6 All 152 releases
← All changes | admin/class-knowledge-manager.php +15 -529 3.2.113.2.2 View file →
@@ -58,9 +58,8 @@
58 58 add_action('wp_ajax_mxchat_refresh_pinecone_entries', array($this, 'ajax_mxchat_refresh_pinecone_entries'));
59 59 add_action('wp_ajax_mxchat_paginate_entries', array($this, 'ajax_mxchat_paginate_entries'));
60 60 add_action('wp_ajax_mxchat_get_entry_content', array($this, 'ajax_mxchat_get_entry_content'));
61 61 add_action('wp_ajax_mxchat_save_entry_content', array($this, 'ajax_mxchat_save_entry_content'));
62 - add_action('wp_ajax_mxchat_inspect_entry', array($this, 'ajax_mxchat_inspect_entry'));
63 62
64 63 // WordPress post management hooks
65 64 add_action('pre_post_update', array($this, 'mxchat_store_pre_update_status'), 10, 2);
66 65 add_action('post_updated', array($this, 'mxchat_handle_post_update'), 10, 3);
@@ -815,230 +814,8 @@
815 814 );
816 815 }
817 816
818 817 /**
819 - * AJAX: Inspect a knowledge entry — returns the per-chunk stored text + metadata
820 - * WITHOUT collapsing it, so a site owner can see exactly what was indexed for an
821 - * entry (plan-mxchat-20260628-d8cb4b). READ-ONLY: never re-embeds or mutates.
822 - */
823 -public function ajax_mxchat_inspect_entry() {
824 - check_ajax_referer('mxchat_inspect_entry_nonce', 'nonce');
825 -
826 - if ( ! current_user_can('manage_options') ) {
827 - wp_send_json_error( array( 'message' => esc_html__('Permission denied.', 'mxchat') ) );
828 - }
829 -
830 - $source_url = isset($_POST['source_url']) ? sanitize_text_field( wp_unslash($_POST['source_url']) ) : '';
831 - $entry_id = isset($_POST['entry_id']) ? absint($_POST['entry_id']) : 0;
832 - $data_source = isset($_POST['data_source']) ? sanitize_key($_POST['data_source']) : 'wordpress';
833 - $bot_id = isset($_POST['bot_id']) ? sanitize_key($_POST['bot_id']) : 'default';
834 -
835 - if ( $data_source === 'pinecone' ) {
836 - $result = $this->inspect_pinecone_entry( $source_url, $entry_id, $bot_id );
837 - } else {
838 - $result = $this->inspect_wordpress_entry( $source_url, $entry_id );
839 - }
840 -
841 - if ( is_wp_error( $result ) ) {
842 - wp_send_json_error( array( 'message' => $result->get_error_message() ) );
843 - }
844 -
845 - wp_send_json_success( $result );
846 -}
847 -
848 -/**
849 - * Read-only inspector for WordPress-DB entries. Mirrors get_wordpress_entry_content()
850 - * but returns each STORED chunk's exact text + length (no implode), plus the assembled
851 - * embedded text. This shows what is actually in the index, not a re-derivation from the post.
852 - */
853 -private function inspect_wordpress_entry( $source_url, $entry_id ) {
854 - global $wpdb;
855 - $table = $wpdb->prefix . 'mxchat_system_prompt_content';
856 -
857 - $rows = array();
858 -
859 - // Group by the real stored source_url — this INCLUDES "mxchat://" manual
860 - // Direct Content entries (the spec's manual-entry case), which share one
861 - // source_url across their chunk rows. Only the synthetic "_ungrouped_<id>"
862 - // display key (invented by the table view for rows with no source_url) is
863 - // excluded; those fall through to the entry_id lookup below.
864 - if ( ! empty( $source_url ) && strpos( $source_url, '_ungrouped_' ) !== 0 ) {
865 - $rows = $wpdb->get_results( $wpdb->prepare(
866 - "SELECT id, article_content, source_url, content_type FROM {$table} WHERE source_url = %s ORDER BY id ASC",
867 - $source_url
868 - ) );
869 - }
870 -
871 - // Fallback / manual "Direct Content" entries: fetch the single row by id.
872 - if ( empty( $rows ) && $entry_id > 0 ) {
873 - $row = $wpdb->get_row( $wpdb->prepare(
874 - "SELECT id, article_content, source_url, content_type FROM {$table} WHERE id = %d",
875 - $entry_id
876 - ) );
877 - if ( $row ) {
878 - $rows = array( $row );
879 - }
880 - }
881 -
882 - if ( empty( $rows ) ) {
883 - return new WP_Error( 'not_found', esc_html__('Entry not found in the local knowledge database.', 'mxchat') );
884 - }
885 -
886 - $chunks = array();
887 - $content_type = '';
888 - foreach ( $rows as $row ) {
889 - $parsed = MxChat_Chunker::parse_stored_chunk( $row->article_content );
890 - $text = isset( $parsed['text'] ) ? $parsed['text'] : '';
891 - $index = isset( $parsed['metadata']['chunk_index'] ) ? intval( $parsed['metadata']['chunk_index'] ) : count( $chunks );
892 - $content_type = $row->content_type;
893 - $chunks[] = array(
894 - 'index' => $index,
895 - 'text' => $text,
896 - 'length' => function_exists('mb_strlen') ? mb_strlen( $text ) : strlen( $text ),
897 - 'row_id' => intval( $row->id ),
898 - );
899 - }
900 -
901 - usort( $chunks, function( $a, $b ) { return $a['index'] - $b['index']; } );
902 -
903 - $assembled = implode( "\n\n", wp_list_pluck( $chunks, 'text' ) );
904 -
905 - return array(
906 - 'store' => 'wordpress',
907 - 'source_url' => $source_url,
908 - 'content_type' => $content_type,
909 - 'is_chunked' => count( $chunks ) > 1,
910 - 'chunk_count' => count( $chunks ),
911 - 'assembled' => $assembled,
912 - 'assembled_length' => function_exists('mb_strlen') ? mb_strlen( $assembled ) : strlen( $assembled ),
913 - 'chunks' => array_values( $chunks ),
914 - // WP-DB storage carries no separate vector metadata; surface that fact
915 - // rather than letting the owner guess (the spec's taxonomy question).
916 - 'metadata' => array(),
917 - 'metadata_note' => esc_html__('Stored in the local WordPress database. Only the assembled text shown here is embedded — there are no separate vector metadata fields (e.g. taxonomy terms are not stored unless they were injected into the text itself).', 'mxchat'),
918 - );
919 -}
920 -
921 -/**
922 - * Read-only inspector for Pinecone entries. Mirrors get_pinecone_entry_content()
923 - * but keeps each vector's text + metadata instead of imploding, so the owner can
924 - * confirm exactly which metadata fields (text/source_url/type/last_updated/created_at/bot_id)
925 - * are present per chunk. READ-ONLY.
926 - */
927 -private function inspect_pinecone_entry( $source_url, $entry_id, $bot_id ) {
928 - if ( ! class_exists('MxChat_Pinecone_Manager') ) {
929 - return new WP_Error( 'pinecone_unavailable', esc_html__('Pinecone manager not available.', 'mxchat') );
930 - }
931 -
932 - if ( $bot_id === 'default' || ! class_exists('MxChat_Multi_Bot_Manager') ) {
933 - $pinecone_options = get_option('mxchat_pinecone_addon_options');
934 - $api_key = $pinecone_options['mxchat_pinecone_api_key'] ?? '';
935 - $host = $pinecone_options['mxchat_pinecone_host'] ?? '';
936 - $namespace = $pinecone_options['mxchat_pinecone_namespace'] ?? '';
937 - } else {
938 - $bot_config = apply_filters('mxchat_get_bot_pinecone_config', array(), $bot_id);
939 - $api_key = $bot_config['api_key'] ?? '';
940 - $host = $bot_config['host'] ?? '';
941 - $namespace = $bot_config['namespace'] ?? '';
942 - }
943 -
944 - if ( empty($host) || empty($api_key) ) {
945 - return new WP_Error( 'pinecone_config', esc_html__('Pinecone not configured.', 'mxchat') );
946 - }
947 -
948 - $base_id = md5( $source_url );
949 - $vector_ids = array( $base_id );
950 -
951 - $list_url = "https://{$host}/vectors/list";
952 - $list_body = array( 'prefix' => $base_id . '_chunk_', 'limit' => 100 );
953 - if ( ! empty($namespace) ) {
954 - $list_body['namespace'] = $namespace;
955 - }
956 -
957 - $list_resp = wp_remote_post( $list_url, array(
958 - 'headers' => array( 'Api-Key' => $api_key, 'Content-Type' => 'application/json' ),
959 - 'body' => wp_json_encode( $list_body ),
960 - 'timeout' => 15,
961 - ) );
962 -
963 - if ( ! is_wp_error($list_resp) ) {
964 - $list_data = json_decode( wp_remote_retrieve_body($list_resp), true );
965 - if ( ! empty($list_data['vectors']) ) {
966 - foreach ( $list_data['vectors'] as $v ) {
967 - $vector_ids[] = $v['id'];
968 - }
969 - }
970 - }
971 -
972 - $fetch_url = "https://{$host}/vectors/fetch";
973 - $fetch_body = array( 'ids' => $vector_ids );
974 - if ( ! empty($namespace) ) {
975 - $fetch_body['namespace'] = $namespace;
976 - }
977 -
978 - $fetch_resp = wp_remote_post( $fetch_url, array(
979 - 'headers' => array( 'Api-Key' => $api_key, 'Content-Type' => 'application/json' ),
980 - 'body' => wp_json_encode( $fetch_body ),
981 - 'timeout' => 15,
982 - ) );
983 -
984 - if ( is_wp_error($fetch_resp) ) {
985 - return new WP_Error( 'pinecone_fetch', esc_html__('Failed to fetch from Pinecone.', 'mxchat') );
986 - }
987 -
988 - $fetch_data = json_decode( wp_remote_retrieve_body($fetch_resp), true );
989 - $vectors = $fetch_data['vectors'] ?? array();
990 -
991 - if ( empty($vectors) ) {
992 - return new WP_Error( 'not_found', esc_html__('Entry not found in Pinecone.', 'mxchat') );
993 - }
994 -
995 - // Whitelisted metadata fields the spec calls out — shown so devs can confirm
996 - // what is (and is NOT) stored per vector.
997 - $meta_fields = array( 'text', 'source_url', 'type', 'last_updated', 'created_at', 'bot_id', 'chunk_index', 'total_chunks' );
998 - $chunks = array();
999 - $content_type = '';
1000 - foreach ( $vectors as $vid => $vector ) {
1001 - $meta = isset($vector['metadata']) && is_array($vector['metadata']) ? $vector['metadata'] : array();
1002 - $text = $meta['text'] ?? '';
1003 - $index = isset($meta['chunk_index']) ? intval($meta['chunk_index']) : count($chunks);
1004 - $content_type = $meta['type'] ?? $content_type;
1005 -
1006 - $clean_meta = array();
1007 - foreach ( $meta_fields as $field ) {
1008 - if ( array_key_exists( $field, $meta ) && $field !== 'text' ) {
1009 - $clean_meta[ $field ] = is_scalar( $meta[ $field ] ) ? (string) $meta[ $field ] : wp_json_encode( $meta[ $field ] );
1010 - }
1011 - }
1012 -
1013 - $chunks[] = array(
1014 - 'index' => $index,
1015 - 'text' => $text,
1016 - 'length' => function_exists('mb_strlen') ? mb_strlen( $text ) : strlen( $text ),
1017 - 'vector_id' => (string) $vid,
1018 - 'metadata' => $clean_meta,
1019 - );
1020 - }
1021 -
1022 - usort( $chunks, function( $a, $b ) { return $a['index'] - $b['index']; } );
1023 -
1024 - $assembled = implode( "\n\n", wp_list_pluck( $chunks, 'text' ) );
1025 -
1026 - return array(
1027 - 'store' => 'pinecone',
1028 - 'source_url' => $source_url,
1029 - 'content_type' => $content_type,
1030 - 'is_chunked' => count( $chunks ) > 1,
1031 - 'chunk_count' => count( $chunks ),
1032 - 'assembled' => $assembled,
1033 - 'assembled_length' => function_exists('mb_strlen') ? mb_strlen( $assembled ) : strlen( $assembled ),
1034 - 'chunks' => array_values( $chunks ),
1035 - 'metadata' => array(),
1036 - 'metadata_note' => esc_html__('Stored in Pinecone. Each chunk above lists the vector metadata fields actually present — if a field you expect (such as taxonomy terms) is missing here, it was not stored as metadata and is only searchable if it appears in the embedded text.', 'mxchat'),
1037 - );
1038 -}
1039 -
1040 -/**
1041 818 * AJAX: Save edited content — re-chunks and re-embeds as needed.
1042 819 * Works for both WordPress DB and Pinecone entries.
1043 820 */
1044 821 public function ajax_mxchat_save_entry_content() {
@@ -3705,22 +3482,9 @@
3705 3482 }
3706 3483
3707 3484 // Get bot_id from request
3708 3485 $bot_id = isset($_POST['bot_id']) ? sanitize_key($_POST['bot_id']) : 'default';
3709 -
3710 - // ACF→PDF extraction is opt-in per import batch. Persist the last-used value so users
3711 - // don't re-check on every batch; the default is OFF for installs that haven't set it.
3712 - $extract_acf_pdfs = !empty($_POST['extract_acf_pdfs']) && $_POST['extract_acf_pdfs'] !== 'false';
3713 - $mxchat_options = get_option('mxchat_options', array());
3714 - if (!is_array($mxchat_options)) {
3715 - $mxchat_options = array();
3716 - }
3717 - $prior_default = !empty($mxchat_options['acf_pdf_extract_default']);
3718 - if ($prior_default !== $extract_acf_pdfs) {
3719 - $mxchat_options['acf_pdf_extract_default'] = $extract_acf_pdfs ? 1 : 0;
3720 - update_option('mxchat_options', $mxchat_options);
3721 - }
3722 -
3486 +
3723 3487 // Process only ONE post at a time to avoid request size issues
3724 3488 $post_id = reset($post_ids);
3725 3489 $post = get_post($post_id);
3726 3490
@@ -3826,57 +3590,23 @@
3826 3590 }
3827 3591
3828 3592 // ADD ACF FIELDS SUPPORT
3829 3593 $acf_fields = $this->mxchat_get_acf_fields_for_post($post_id);
3830 - $pdf_extracted_count = 0;
3831 3594 if (!empty($acf_fields)) {
3832 3595 $acf_content_parts = array();
3833 - $pdf_attachment_ids = array();
3834 -
3596 +
3835 3597 foreach ($acf_fields as $field_name => $field_value) {
3836 3598 $formatted_value = $this->mxchat_format_acf_field_value($field_value, $field_name, $post_id);
3837 -
3599 +
3838 3600 if (!empty($formatted_value)) {
3839 3601 $field_label = ucwords(str_replace('_', ' ', $field_name));
3840 3602 $acf_content_parts[] = $field_label . ": " . $formatted_value;
3841 3603 }
3842 -
3843 - // Walk this field's value tree for any PDF attachment references and queue them for extraction.
3844 - // Only when the user opted into ACF→PDF extraction for this batch; otherwise the ACF text
3845 - // still lands in the KB but the heavier PDF parsing is skipped.
3846 - if ($extract_acf_pdfs) {
3847 - $this->mxchat_collect_pdf_attachment_ids_from_acf_value($field_value, $pdf_attachment_ids);
3848 - }
3849 3604 }
3850 -
3605 +
3851 3606 if (!empty($acf_content_parts)) {
3852 3607 $content .= "\n\n" . implode("\n", $acf_content_parts);
3853 3608 }
3854 -
3855 - // Extract text from each unique PDF found in ACF fields and append as a labeled section
3856 - if ($extract_acf_pdfs && !empty($pdf_attachment_ids)) {
3857 - $pdf_attachment_ids = array_unique(array_filter(array_map('intval', $pdf_attachment_ids)));
3858 - $pdf_sections = array();
3859 - foreach ($pdf_attachment_ids as $att_id) {
3860 - $pdf_text = $this->mxchat_extract_pdf_text_by_attachment_id($att_id);
3861 - if (!empty($pdf_text)) {
3862 - $pdf_title = get_the_title($att_id);
3863 - $pdf_url = wp_get_attachment_url($att_id);
3864 - $header = 'PDF Attachment';
3865 - if (!empty($pdf_title)) {
3866 - $header .= ': ' . $pdf_title;
3867 - }
3868 - if (!empty($pdf_url)) {
3869 - $header .= ' (' . $pdf_url . ')';
3870 - }
3871 - $pdf_sections[] = $header . "\n" . $pdf_text;
3872 - $pdf_extracted_count++;
3873 - }
3874 - }
3875 - if (!empty($pdf_sections)) {
3876 - $content .= "\n\nAttached PDFs:\n" . implode("\n\n", $pdf_sections);
3877 - }
3878 - }
3879 3609 }
3880 3610
3881 3611 // ADD CUSTOM POST META SUPPORT (whitelisted non-ACF meta fields)
3882 3612 $custom_meta = $this->mxchat_get_whitelisted_post_meta($post_id);
@@ -4006,9 +3736,8 @@
4006 3736 'title' => $post->post_title,
4007 3737 'operation_type' => $operation_type,
4008 3738 'vector_id' => $vector_id,
4009 3739 'acf_fields_found' => $acf_field_count,
4010 - 'pdf_extracted_count' => (int) $pdf_extracted_count,
4011 3740 'content_preview' => substr($content, 0, 100) . '...',
4012 3741 'bot_id' => $bot_id
4013 3742 ));
4014 3743 exit;
@@ -4423,20 +4152,9 @@
4423 4152
4424 4153 // Get bot-specific options
4425 4154 $bot_options = $this->get_bot_options($bot_id);
4426 4155 $options = !empty($bot_options) ? $bot_options : get_option('mxchat_options');
4427 -
4428 - // Opt-in: when the custom provider is selected for embeddings, index through
4429 - // the same custom endpoint the query path uses so stored vectors and query
4430 - // vectors share a model. Returns the vector array on success, or an error
4431 - // string on failure (this function's existing failure contract).
4432 - if (isset($options['custom_provider_for_embeddings']) && $options['custom_provider_for_embeddings'] === 'on') {
4433 - if (!class_exists('MxChat_Utils')) {
4434 - require_once dirname(__FILE__) . '/../includes/class-mxchat-utils.php';
4435 - }
4436 - return MxChat_Utils::generate_embedding_custom($text, $options);
4437 - }
4438 -
4156 +
4439 4157 $selected_model = $options['embedding_model'] ?? 'text-embedding-ada-002';
4440 4158 //error_log('[MXCHAT-EMBED] Selected embedding model for bot ' . $bot_id . ': ' . $selected_model);
4441 4159
4442 4160 // Determine provider and endpoint
@@ -5354,197 +5072,8 @@
5354 5072 return implode(', ', array_filter($text_parts));
5355 5073 }
5356 5074
5357 5075 /**
5358 - * Walk an ACF field value tree and collect attachment IDs for any value that
5359 - * resolves to a PDF in the WordPress media library. Handles the three shapes
5360 - * ACF returns for File/Image/URL fields (array with ID+url, integer attachment ID,
5361 - * plain URL string), and recurses through repeater/group/flexible content.
5362 - *
5363 - * @param mixed $value The ACF field value (any depth)
5364 - * @param array $out Accumulator (passed by reference) for attachment IDs
5365 - * @param int $depth Recursion guard
5366 - */
5367 -private function mxchat_collect_pdf_attachment_ids_from_acf_value($value, &$out, $depth = 0) {
5368 - if ($depth > 6) {
5369 - return; // prevent runaway recursion on circular/very-deep structures
5370 - }
5371 -
5372 - if (empty($value)) {
5373 - return;
5374 - }
5375 -
5376 - // Array shapes: ACF File/Image return value=array; repeaters/groups are arrays of arrays
5377 - if (is_array($value)) {
5378 - // Direct File/Image-style array (has 'url' and usually 'ID' + 'mime_type')
5379 - $looks_like_attachment = isset($value['url']) || isset($value['ID']) || isset($value['id']);
5380 - if ($looks_like_attachment) {
5381 - $att_id = 0;
5382 - if (!empty($value['ID']) && is_numeric($value['ID'])) {
5383 - $att_id = (int) $value['ID'];
5384 - } elseif (!empty($value['id']) && is_numeric($value['id'])) {
5385 - $att_id = (int) $value['id'];
5386 - } elseif (!empty($value['url']) && is_string($value['url'])) {
5387 - $att_id = (int) attachment_url_to_postid($value['url']);
5388 - }
5389 -
5390 - $is_pdf = false;
5391 - if (!empty($value['mime_type']) && $value['mime_type'] === 'application/pdf') {
5392 - $is_pdf = true;
5393 - } elseif (!empty($value['subtype']) && strtolower((string) $value['subtype']) === 'pdf') {
5394 - $is_pdf = true;
5395 - } elseif (!empty($value['url']) && is_string($value['url']) && $this->mxchat_url_looks_like_pdf($value['url'])) {
5396 - $is_pdf = true;
5397 - } elseif ($att_id && get_post_mime_type($att_id) === 'application/pdf') {
5398 - $is_pdf = true;
5399 - }
5400 -
5401 - if ($is_pdf && $att_id && get_post_mime_type($att_id) === 'application/pdf') {
5402 - $out[] = $att_id;
5403 - }
5404 - // An array node that represents one attachment doesn't contain other
5405 - // attachments inside it — done with this branch.
5406 - return;
5407 - }
5408 -
5409 - // Recurse: repeater rows, flexible-content layouts, groups, etc.
5410 - foreach ($value as $sub) {
5411 - $this->mxchat_collect_pdf_attachment_ids_from_acf_value($sub, $out, $depth + 1);
5412 - }
5413 - return;
5414 - }
5415 -
5416 - // Plain numeric attachment ID (ACF File field set to "Return: ID")
5417 - if (is_numeric($value)) {
5418 - $att_id = (int) $value;
5419 - if ($att_id > 0 && get_post_mime_type($att_id) === 'application/pdf') {
5420 - $out[] = $att_id;
5421 - }
5422 - return;
5423 - }
5424 -
5425 - // Plain string — URL pointing at a PDF (ACF File field set to "Return: URL", or a custom URL/text field)
5426 - if (is_string($value)) {
5427 - $trimmed = trim($value);
5428 - if ($trimmed !== '' && $this->mxchat_url_looks_like_pdf($trimmed)) {
5429 - $att_id = (int) attachment_url_to_postid($trimmed);
5430 - if ($att_id > 0 && get_post_mime_type($att_id) === 'application/pdf') {
5431 - $out[] = $att_id;
5432 - }
5433 - }
5434 - return;
5435 - }
5436 -}
5437 -
5438 -/**
5439 - * Heuristic: does this URL/string look like a PDF reference?
5440 - * Tolerates query strings and fragments (#page=2).
5441 - */
5442 -private function mxchat_url_looks_like_pdf($url) {
5443 - if (!is_string($url) || $url === '') {
5444 - return false;
5445 - }
5446 - // Strip query + fragment before checking extension
5447 - $path = preg_replace('/[?#].*$/', '', $url);
5448 - return (bool) preg_match('/\.pdf$/i', $path);
5449 -}
5450 -
5451 -/**
5452 - * Extract text from a PDF attachment by ID using the bundled Smalot parser.
5453 - * Reads the file directly from disk via get_attached_file (no HTTP fetch).
5454 - * Result is cached on the attachment as post_meta keyed by file mtime so we
5455 - * only parse the same PDF once unless the file changes on disk.
5456 - *
5457 - * @param int $attachment_id
5458 - * @return string Extracted plain text, or '' on failure.
5459 - */
5460 -private function mxchat_extract_pdf_text_by_attachment_id($attachment_id) {
5461 - $attachment_id = (int) $attachment_id;
5462 - if ($attachment_id <= 0) {
5463 - return '';
5464 - }
5465 - if (get_post_mime_type($attachment_id) !== 'application/pdf') {
5466 - return '';
5467 - }
5468 -
5469 - $pdf_path = get_attached_file($attachment_id);
5470 - if (empty($pdf_path) || !file_exists($pdf_path) || !is_readable($pdf_path)) {
5471 - return '';
5472 - }
5473 -
5474 - // Raw-file size cap. Parsing very large PDFs can OOM the request; skip with a log entry
5475 - // and let the rest of the ACF content land in the KB. Filterable for users who need it bigger.
5476 - $default_max_bytes = 25 * 1024 * 1024;
5477 - $max_bytes = (int) apply_filters('mxchat_acf_pdf_max_bytes', $default_max_bytes, $attachment_id, $pdf_path);
5478 - if ($max_bytes > 0) {
5479 - $file_size = @filesize($pdf_path);
5480 - if ($file_size !== false && $file_size > $max_bytes) {
5481 - error_log(sprintf(
5482 - '[mxchat] ACF PDF skipped (over size cap): attachment %d "%s" %d bytes > cap %d',
5483 - $attachment_id,
5484 - basename($pdf_path),
5485 - $file_size,
5486 - $max_bytes
5487 - ));
5488 - return '';
5489 - }
5490 - }
5491 -
5492 - $mtime = @filemtime($pdf_path);
5493 - $cache_meta_key = '_mxchat_acf_pdf_text_v1';
5494 - $cached = get_post_meta($attachment_id, $cache_meta_key, true);
5495 - if (is_array($cached) && isset($cached['mtime'], $cached['text']) && (int) $cached['mtime'] === (int) $mtime) {
5496 - return (string) $cached['text'];
5497 - }
5498 -
5499 - $text = '';
5500 - try {
5501 - if (function_exists('mxchat_load_pdf_parser')) {
5502 - mxchat_load_pdf_parser();
5503 - }
5504 - if (!class_exists('\\Smalot\\PdfParser\\Parser')) {
5505 - return '';
5506 - }
5507 - $parser = new \Smalot\PdfParser\Parser();
5508 - $pdf = $parser->parseFile($pdf_path);
5509 - $pages = $pdf->getPages();
5510 - $page_texts = array();
5511 - foreach ($pages as $page) {
5512 - $page_text = '';
5513 - try {
5514 - $page_text = $page->getText();
5515 - } catch (\Exception $e) {
5516 - $page_text = '';
5517 - }
5518 - if (!empty($page_text)) {
5519 - $page_texts[] = $page_text;
5520 - }
5521 - }
5522 - $text = trim(implode("\n\n", $page_texts));
5523 - } catch (\Exception $e) {
5524 - error_log('[mxchat] ACF PDF extraction failed for attachment ' . $attachment_id . ': ' . $e->getMessage());
5525 - return '';
5526 - } catch (\Throwable $e) {
5527 - error_log('[mxchat] ACF PDF extraction error for attachment ' . $attachment_id . ': ' . $e->getMessage());
5528 - return '';
5529 - }
5530 -
5531 - // Cap per-PDF text to avoid blowing up the embedding payload on enormous PDFs.
5532 - // The chunker downstream will still split this into multiple vectors.
5533 - $max_len = (int) apply_filters('mxchat_acf_pdf_text_max_length', 50000);
5534 - if ($max_len > 0 && strlen($text) > $max_len) {
5535 - $text = substr($text, 0, $max_len);
5536 - }
5537 -
5538 - update_post_meta($attachment_id, $cache_meta_key, array(
5539 - 'mtime' => (int) $mtime,
5540 - 'text' => $text,
5541 - ));
5542 -
5543 - return $text;
5544 -}
5545 -
5546 -/**
5547 5076 * Handle ACF save - fires after ACF fields are saved
5548 5077 * This ensures ACF field data is available when syncing to knowledge base
5549 5078 */
5550 5079 public function mxchat_handle_acf_save($post_id) {
@@ -5787,9 +5316,8 @@
5787 5316 // ADD ACF FIELDS SUPPORT (matches ajax_mxchat_process_selected_content behavior)
5788 5317 $acf_fields = $this->mxchat_get_acf_fields_for_post($post_id);
5789 5318 if (!empty($acf_fields)) {
5790 5319 $acf_content_parts = array();
5791 - $pdf_attachment_ids = array();
5792 5320
5793 5321 foreach ($acf_fields as $field_name => $field_value) {
5794 5322 $formatted_value = $this->mxchat_format_acf_field_value($field_value, $field_name, $post_id);
5795 5323 if (!empty($formatted_value)) {
@@ -5796,44 +5324,13 @@
5796 5324 // Convert field name to readable label
5797 5325 $field_label = ucwords(str_replace(['_', '-'], ' ', $field_name));
5798 5326 $acf_content_parts[] = $field_label . ": " . $formatted_value;
5799 5327 }
5800 -
5801 - $this->mxchat_collect_pdf_attachment_ids_from_acf_value($field_value, $pdf_attachment_ids);
5802 5328 }
5803 5329
5804 5330 if (!empty($acf_content_parts)) {
5805 5331 $final_content .= "\n\n" . implode("\n", $acf_content_parts);
5806 5332 }
5807 -
5808 - // Gate the auto-sync PDF-extraction loop behind an opt-in option.
5809 - // Mirrors the per-batch checkbox the manual content selector has; the
5810 - // 25 MB size cap lives in the shared extractor so it applies in both
5811 - // paths regardless. Default OFF — re-parsing every ACF PDF on every
5812 - // editor save is expensive and most sites don't want it.
5813 - $autosync_extract_acf_pdfs = get_option('mxchat_auto_sync_acf_pdfs', '0') === '1';
5814 - if ($autosync_extract_acf_pdfs && !empty($pdf_attachment_ids)) {
5815 - $pdf_attachment_ids = array_unique(array_filter(array_map('intval', $pdf_attachment_ids)));
5816 - $pdf_sections = array();
5817 - foreach ($pdf_attachment_ids as $att_id) {
5818 - $pdf_text = $this->mxchat_extract_pdf_text_by_attachment_id($att_id);
5819 - if (!empty($pdf_text)) {
5820 - $pdf_title = get_the_title($att_id);
5821 - $pdf_url = wp_get_attachment_url($att_id);
5822 - $header = 'PDF Attachment';
5823 - if (!empty($pdf_title)) {
5824 - $header .= ': ' . $pdf_title;
5825 - }
5826 - if (!empty($pdf_url)) {
5827 - $header .= ' (' . $pdf_url . ')';
5828 - }
5829 - $pdf_sections[] = $header . "\n" . $pdf_text;
5830 - }
5831 - }
5832 - if (!empty($pdf_sections)) {
5833 - $final_content .= "\n\nAttached PDFs:\n" . implode("\n\n", $pdf_sections);
5834 - }
5835 - }
5836 5333 }
5837 5334
5838 5335 // ADD CUSTOM POST META SUPPORT (whitelisted non-ACF meta fields)
5839 5336 $custom_meta = $this->mxchat_get_whitelisted_post_meta($post_id);
@@ -6851,16 +6348,16 @@
6851 6348 wp_send_json_error('Unauthorized access');
6852 6349 exit;
6853 6350 }
6854 6351
6855 - $tag_input = isset($_POST['tag_slug']) ? sanitize_text_field($_POST['tag_slug']) : '';
6352 + $tag_slug = isset($_POST['tag_slug']) ? sanitize_text_field($_POST['tag_slug']) : '';
6856 6353 $role_restriction = isset($_POST['role_restriction']) ? sanitize_text_field($_POST['role_restriction']) : 'public';
6857 -
6858 - if (empty($tag_input)) {
6859 - wp_send_json_error('Please enter a tag name or slug');
6354 +
6355 + if (empty($tag_slug)) {
6356 + wp_send_json_error('Tag slug is required');
6860 6357 exit;
6861 6358 }
6862 -
6359 +
6863 6360 // Validate role restriction
6864 6361 $valid_roles = array_keys($this->mxchat_get_role_options());
6865 6362 if (!in_array($role_restriction, $valid_roles)) {
6866 6363 wp_send_json_error('Invalid role restriction');
@@ -6865,27 +6362,16 @@
6865 6362 if (!in_array($role_restriction, $valid_roles)) {
6866 6363 wp_send_json_error('Invalid role restriction');
6867 6364 exit;
6868 6365 }
6869 -
6870 - // Resolve the tag by slug first, then fall back to its display name, so users can
6871 - // enter either "premium-content" or "Premium Content". (plan b8bcf5 — the field is
6872 - // labeled by name but previously validated by slug only, producing the confusing
6873 - // "Tag does not exist in WordPress" error when a real tag's name was typed.)
6874 - $term = get_term_by('slug', $tag_input, 'post_tag');
6366 +
6367 + // Check if tag exists in WordPress
6368 + $term = get_term_by('slug', $tag_slug, 'post_tag');
6875 6369 if (!$term) {
6876 - $term = get_term_by('name', $tag_input, 'post_tag');
6877 - }
6878 - if (!$term) {
6879 - wp_send_json_error('No tag with that name or slug exists yet. Create it under Posts → Tags first, then enter its name or slug.');
6370 + wp_send_json_error('Tag does not exist in WordPress');
6880 6371 exit;
6881 6372 }
6882 -
6883 - // Always key the mapping by the RESOLVED slug — apply_role_restriction_to_post()
6884 - // compares against each post's tag slugs, so the stored key must be a slug,
6885 - // never the raw (possibly display-name) input.
6886 - $tag_slug = $term->slug;
6887 -
6373 +
6888 6374 // Get existing mappings
6889 6375 $mappings = get_option('mxchat_tag_role_mappings', array());
6890 6376
6891 6377 // Check if mapping already exists