| @@ -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() { |
| @@ -1213,21 +990,16 @@ | ||
| 1213 | 990 | wp_safe_redirect(esc_url(admin_url('admin.php?page=mxchat-prompts'))); |
| 1214 | 991 | exit; |
| 1215 | 992 | } |
| 1216 | 993 | |
| 1217 | - // Fetch URL — send an honest, versioned MXChat crawler UA (not a spoofed | |
| 1218 | - // browser). Stale browser UAs are exactly what WAFs like SiteGround's | |
| 1219 | - // ModSecurity flag as scrapers, 403-ing the fetch (including PDFs served | |
| 1220 | - // from the site's own media library, which route through this same call). | |
| 1221 | - // See mxchat_ingest_user_agent(). Accept is kept for content negotiation; | |
| 1222 | - // the browser-only Accept-Language fingerprint is dropped so it stays | |
| 1223 | - // coherent with a bot identity. | |
| 994 | + // Fetch URL — use browser-like headers so servers with bot protection don't block us | |
| 1224 | 995 | $response = wp_remote_get($submitted_url, array( |
| 1225 | 996 | 'timeout' => 30, |
| 1226 | 997 | 'sslverify' => false, |
| 1227 | - 'user-agent' => mxchat_ingest_user_agent(), | |
| 998 | + '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', | |
| 1228 | 999 | 'headers' => array( |
| 1229 | 1000 | 'Accept' => 'text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8', |
| 1001 | + 'Accept-Language' => 'en-US,en;q=0.9', | |
| 1230 | 1002 | ), |
| 1231 | 1003 | )); |
| 1232 | 1004 | |
| 1233 | 1005 | if (is_wp_error($response) || wp_remote_retrieve_response_code($response) !== 200) { |
| @@ -3066,9 +2838,9 @@ | ||
| 3066 | 2838 | $response = wp_remote_head($url, array( |
| 3067 | 2839 | 'timeout' => 10, |
| 3068 | 2840 | 'sslverify' => false, |
| 3069 | 2841 | 'redirection' => 1, |
| 3070 | - 'user-agent' => mxchat_ingest_user_agent(), | |
| 2842 | + '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', | |
| 3071 | 2843 | )); |
| 3072 | 2844 | |
| 3073 | 2845 | if (!is_wp_error($response) && wp_remote_retrieve_response_code($response) === 200) { |
| 3074 | 2846 | // Found a sitemap index - parse it to get sub-sitemaps |
| @@ -3128,11 +2900,12 @@ | ||
| 3128 | 2900 | |
| 3129 | 2901 | $response = wp_remote_get($url, array( |
| 3130 | 2902 | 'timeout' => 30, |
| 3131 | 2903 | 'sslverify' => false, |
| 3132 | - 'user-agent' => mxchat_ingest_user_agent(), | |
| 2904 | + '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', | |
| 3133 | 2905 | 'headers' => array( |
| 3134 | 2906 | 'Accept' => 'text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8', |
| 2907 | + 'Accept-Language' => 'en-US,en;q=0.9', | |
| 3135 | 2908 | ), |
| 3136 | 2909 | )); |
| 3137 | 2910 | |
| 3138 | 2911 | if (is_wp_error($response)) { |
| @@ -3186,11 +2959,12 @@ | ||
| 3186 | 2959 | private function get_sitemap_url_count($url) { |
| 3187 | 2960 | $response = wp_remote_get($url, array( |
| 3188 | 2961 | 'timeout' => 30, |
| 3189 | 2962 | 'sslverify' => false, |
| 3190 | - 'user-agent' => mxchat_ingest_user_agent(), | |
| 2963 | + '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', | |
| 3191 | 2964 | 'headers' => array( |
| 3192 | 2965 | 'Accept' => 'text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8', |
| 2966 | + 'Accept-Language' => 'en-US,en;q=0.9', | |
| 3193 | 2967 | ), |
| 3194 | 2968 | )); |
| 3195 | 2969 | |
| 3196 | 2970 | if (is_wp_error($response)) { |
| @@ -3216,9 +2990,9 @@ | ||
| 3216 | 2990 | |
| 3217 | 2991 | $response = wp_remote_get($robots_url, array( |
| 3218 | 2992 | 'timeout' => 15, |
| 3219 | 2993 | 'sslverify' => false, |
| 3220 | - 'user-agent' => mxchat_ingest_user_agent(), | |
| 2994 | + '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', | |
| 3221 | 2995 | )); |
| 3222 | 2996 | |
| 3223 | 2997 | if (is_wp_error($response)) { |
| 3224 | 2998 | return $sitemaps; |
| @@ -3708,22 +3482,9 @@ | ||
| 3708 | 3482 | } |
| 3709 | 3483 | |
| 3710 | 3484 | // Get bot_id from request |
| 3711 | 3485 | $bot_id = isset($_POST['bot_id']) ? sanitize_key($_POST['bot_id']) : 'default'; |
| 3712 | - | |
| 3713 | - // ACF→PDF extraction is opt-in per import batch. Persist the last-used value so users | |
| 3714 | - // don't re-check on every batch; the default is OFF for installs that haven't set it. | |
| 3715 | - $extract_acf_pdfs = !empty($_POST['extract_acf_pdfs']) && $_POST['extract_acf_pdfs'] !== 'false'; | |
| 3716 | - $mxchat_options = get_option('mxchat_options', array()); | |
| 3717 | - if (!is_array($mxchat_options)) { | |
| 3718 | - $mxchat_options = array(); | |
| 3719 | - } | |
| 3720 | - $prior_default = !empty($mxchat_options['acf_pdf_extract_default']); | |
| 3721 | - if ($prior_default !== $extract_acf_pdfs) { | |
| 3722 | - $mxchat_options['acf_pdf_extract_default'] = $extract_acf_pdfs ? 1 : 0; | |
| 3723 | - update_option('mxchat_options', $mxchat_options); | |
| 3724 | - } | |
| 3725 | - | |
| 3486 | + | |
| 3726 | 3487 | // Process only ONE post at a time to avoid request size issues |
| 3727 | 3488 | $post_id = reset($post_ids); |
| 3728 | 3489 | $post = get_post($post_id); |
| 3729 | 3490 | |
| @@ -3829,57 +3590,23 @@ | ||
| 3829 | 3590 | } |
| 3830 | 3591 | |
| 3831 | 3592 | // ADD ACF FIELDS SUPPORT |
| 3832 | 3593 | $acf_fields = $this->mxchat_get_acf_fields_for_post($post_id); |
| 3833 | - $pdf_extracted_count = 0; | |
| 3834 | 3594 | if (!empty($acf_fields)) { |
| 3835 | 3595 | $acf_content_parts = array(); |
| 3836 | - $pdf_attachment_ids = array(); | |
| 3837 | - | |
| 3596 | + | |
| 3838 | 3597 | foreach ($acf_fields as $field_name => $field_value) { |
| 3839 | 3598 | $formatted_value = $this->mxchat_format_acf_field_value($field_value, $field_name, $post_id); |
| 3840 | - | |
| 3599 | + | |
| 3841 | 3600 | if (!empty($formatted_value)) { |
| 3842 | 3601 | $field_label = ucwords(str_replace('_', ' ', $field_name)); |
| 3843 | 3602 | $acf_content_parts[] = $field_label . ": " . $formatted_value; |
| 3844 | 3603 | } |
| 3845 | - | |
| 3846 | - // Walk this field's value tree for any PDF attachment references and queue them for extraction. | |
| 3847 | - // Only when the user opted into ACF→PDF extraction for this batch; otherwise the ACF text | |
| 3848 | - // still lands in the KB but the heavier PDF parsing is skipped. | |
| 3849 | - if ($extract_acf_pdfs) { | |
| 3850 | - $this->mxchat_collect_pdf_attachment_ids_from_acf_value($field_value, $pdf_attachment_ids); | |
| 3851 | - } | |
| 3852 | 3604 | } |
| 3853 | - | |
| 3605 | + | |
| 3854 | 3606 | if (!empty($acf_content_parts)) { |
| 3855 | 3607 | $content .= "\n\n" . implode("\n", $acf_content_parts); |
| 3856 | 3608 | } |
| 3857 | - | |
| 3858 | - // Extract text from each unique PDF found in ACF fields and append as a labeled section | |
| 3859 | - if ($extract_acf_pdfs && !empty($pdf_attachment_ids)) { | |
| 3860 | - $pdf_attachment_ids = array_unique(array_filter(array_map('intval', $pdf_attachment_ids))); | |
| 3861 | - $pdf_sections = array(); | |
| 3862 | - foreach ($pdf_attachment_ids as $att_id) { | |
| 3863 | - $pdf_text = $this->mxchat_extract_pdf_text_by_attachment_id($att_id); | |
| 3864 | - if (!empty($pdf_text)) { | |
| 3865 | - $pdf_title = get_the_title($att_id); | |
| 3866 | - $pdf_url = wp_get_attachment_url($att_id); | |
| 3867 | - $header = 'PDF Attachment'; | |
| 3868 | - if (!empty($pdf_title)) { | |
| 3869 | - $header .= ': ' . $pdf_title; | |
| 3870 | - } | |
| 3871 | - if (!empty($pdf_url)) { | |
| 3872 | - $header .= ' (' . $pdf_url . ')'; | |
| 3873 | - } | |
| 3874 | - $pdf_sections[] = $header . "\n" . $pdf_text; | |
| 3875 | - $pdf_extracted_count++; | |
| 3876 | - } | |
| 3877 | - } | |
| 3878 | - if (!empty($pdf_sections)) { | |
| 3879 | - $content .= "\n\nAttached PDFs:\n" . implode("\n\n", $pdf_sections); | |
| 3880 | - } | |
| 3881 | - } | |
| 3882 | 3609 | } |
| 3883 | 3610 | |
| 3884 | 3611 | // ADD CUSTOM POST META SUPPORT (whitelisted non-ACF meta fields) |
| 3885 | 3612 | $custom_meta = $this->mxchat_get_whitelisted_post_meta($post_id); |
| @@ -4009,9 +3736,8 @@ | ||
| 4009 | 3736 | 'title' => $post->post_title, |
| 4010 | 3737 | 'operation_type' => $operation_type, |
| 4011 | 3738 | 'vector_id' => $vector_id, |
| 4012 | 3739 | 'acf_fields_found' => $acf_field_count, |
| 4013 | - 'pdf_extracted_count' => (int) $pdf_extracted_count, | |
| 4014 | 3740 | 'content_preview' => substr($content, 0, 100) . '...', |
| 4015 | 3741 | 'bot_id' => $bot_id |
| 4016 | 3742 | )); |
| 4017 | 3743 | exit; |
| @@ -4426,20 +4152,9 @@ | ||
| 4426 | 4152 | |
| 4427 | 4153 | // Get bot-specific options |
| 4428 | 4154 | $bot_options = $this->get_bot_options($bot_id); |
| 4429 | 4155 | $options = !empty($bot_options) ? $bot_options : get_option('mxchat_options'); |
| 4430 | - | |
| 4431 | - // Opt-in: when the custom provider is selected for embeddings, index through | |
| 4432 | - // the same custom endpoint the query path uses so stored vectors and query | |
| 4433 | - // vectors share a model. Returns the vector array on success, or an error | |
| 4434 | - // string on failure (this function's existing failure contract). | |
| 4435 | - if (isset($options['custom_provider_for_embeddings']) && $options['custom_provider_for_embeddings'] === 'on') { | |
| 4436 | - if (!class_exists('MxChat_Utils')) { | |
| 4437 | - require_once dirname(__FILE__) . '/../includes/class-mxchat-utils.php'; | |
| 4438 | - } | |
| 4439 | - return MxChat_Utils::generate_embedding_custom($text, $options); | |
| 4440 | - } | |
| 4441 | - | |
| 4156 | + | |
| 4442 | 4157 | $selected_model = $options['embedding_model'] ?? 'text-embedding-ada-002'; |
| 4443 | 4158 | //error_log('[MXCHAT-EMBED] Selected embedding model for bot ' . $bot_id . ': ' . $selected_model); |
| 4444 | 4159 | |
| 4445 | 4160 | // Determine provider and endpoint |
| @@ -5357,197 +5072,8 @@ | ||
| 5357 | 5072 | return implode(', ', array_filter($text_parts)); |
| 5358 | 5073 | } |
| 5359 | 5074 | |
| 5360 | 5075 | /** |
| 5361 | - * Walk an ACF field value tree and collect attachment IDs for any value that | |
| 5362 | - * resolves to a PDF in the WordPress media library. Handles the three shapes | |
| 5363 | - * ACF returns for File/Image/URL fields (array with ID+url, integer attachment ID, | |
| 5364 | - * plain URL string), and recurses through repeater/group/flexible content. | |
| 5365 | - * | |
| 5366 | - * @param mixed $value The ACF field value (any depth) | |
| 5367 | - * @param array $out Accumulator (passed by reference) for attachment IDs | |
| 5368 | - * @param int $depth Recursion guard | |
| 5369 | - */ | |
| 5370 | -private function mxchat_collect_pdf_attachment_ids_from_acf_value($value, &$out, $depth = 0) { | |
| 5371 | - if ($depth > 6) { | |
| 5372 | - return; // prevent runaway recursion on circular/very-deep structures | |
| 5373 | - } | |
| 5374 | - | |
| 5375 | - if (empty($value)) { | |
| 5376 | - return; | |
| 5377 | - } | |
| 5378 | - | |
| 5379 | - // Array shapes: ACF File/Image return value=array; repeaters/groups are arrays of arrays | |
| 5380 | - if (is_array($value)) { | |
| 5381 | - // Direct File/Image-style array (has 'url' and usually 'ID' + 'mime_type') | |
| 5382 | - $looks_like_attachment = isset($value['url']) || isset($value['ID']) || isset($value['id']); | |
| 5383 | - if ($looks_like_attachment) { | |
| 5384 | - $att_id = 0; | |
| 5385 | - if (!empty($value['ID']) && is_numeric($value['ID'])) { | |
| 5386 | - $att_id = (int) $value['ID']; | |
| 5387 | - } elseif (!empty($value['id']) && is_numeric($value['id'])) { | |
| 5388 | - $att_id = (int) $value['id']; | |
| 5389 | - } elseif (!empty($value['url']) && is_string($value['url'])) { | |
| 5390 | - $att_id = (int) attachment_url_to_postid($value['url']); | |
| 5391 | - } | |
| 5392 | - | |
| 5393 | - $is_pdf = false; | |
| 5394 | - if (!empty($value['mime_type']) && $value['mime_type'] === 'application/pdf') { | |
| 5395 | - $is_pdf = true; | |
| 5396 | - } elseif (!empty($value['subtype']) && strtolower((string) $value['subtype']) === 'pdf') { | |
| 5397 | - $is_pdf = true; | |
| 5398 | - } elseif (!empty($value['url']) && is_string($value['url']) && $this->mxchat_url_looks_like_pdf($value['url'])) { | |
| 5399 | - $is_pdf = true; | |
| 5400 | - } elseif ($att_id && get_post_mime_type($att_id) === 'application/pdf') { | |
| 5401 | - $is_pdf = true; | |
| 5402 | - } | |
| 5403 | - | |
| 5404 | - if ($is_pdf && $att_id && get_post_mime_type($att_id) === 'application/pdf') { | |
| 5405 | - $out[] = $att_id; | |
| 5406 | - } | |
| 5407 | - // An array node that represents one attachment doesn't contain other | |
| 5408 | - // attachments inside it — done with this branch. | |
| 5409 | - return; | |
| 5410 | - } | |
| 5411 | - | |
| 5412 | - // Recurse: repeater rows, flexible-content layouts, groups, etc. | |
| 5413 | - foreach ($value as $sub) { | |
| 5414 | - $this->mxchat_collect_pdf_attachment_ids_from_acf_value($sub, $out, $depth + 1); | |
| 5415 | - } | |
| 5416 | - return; | |
| 5417 | - } | |
| 5418 | - | |
| 5419 | - // Plain numeric attachment ID (ACF File field set to "Return: ID") | |
| 5420 | - if (is_numeric($value)) { | |
| 5421 | - $att_id = (int) $value; | |
| 5422 | - if ($att_id > 0 && get_post_mime_type($att_id) === 'application/pdf') { | |
| 5423 | - $out[] = $att_id; | |
| 5424 | - } | |
| 5425 | - return; | |
| 5426 | - } | |
| 5427 | - | |
| 5428 | - // Plain string — URL pointing at a PDF (ACF File field set to "Return: URL", or a custom URL/text field) | |
| 5429 | - if (is_string($value)) { | |
| 5430 | - $trimmed = trim($value); | |
| 5431 | - if ($trimmed !== '' && $this->mxchat_url_looks_like_pdf($trimmed)) { | |
| 5432 | - $att_id = (int) attachment_url_to_postid($trimmed); | |
| 5433 | - if ($att_id > 0 && get_post_mime_type($att_id) === 'application/pdf') { | |
| 5434 | - $out[] = $att_id; | |
| 5435 | - } | |
| 5436 | - } | |
| 5437 | - return; | |
| 5438 | - } | |
| 5439 | -} | |
| 5440 | - | |
| 5441 | -/** | |
| 5442 | - * Heuristic: does this URL/string look like a PDF reference? | |
| 5443 | - * Tolerates query strings and fragments (#page=2). | |
| 5444 | - */ | |
| 5445 | -private function mxchat_url_looks_like_pdf($url) { | |
| 5446 | - if (!is_string($url) || $url === '') { | |
| 5447 | - return false; | |
| 5448 | - } | |
| 5449 | - // Strip query + fragment before checking extension | |
| 5450 | - $path = preg_replace('/[?#].*$/', '', $url); | |
| 5451 | - return (bool) preg_match('/\.pdf$/i', $path); | |
| 5452 | -} | |
| 5453 | - | |
| 5454 | -/** | |
| 5455 | - * Extract text from a PDF attachment by ID using the bundled Smalot parser. | |
| 5456 | - * Reads the file directly from disk via get_attached_file (no HTTP fetch). | |
| 5457 | - * Result is cached on the attachment as post_meta keyed by file mtime so we | |
| 5458 | - * only parse the same PDF once unless the file changes on disk. | |
| 5459 | - * | |
| 5460 | - * @param int $attachment_id | |
| 5461 | - * @return string Extracted plain text, or '' on failure. | |
| 5462 | - */ | |
| 5463 | -private function mxchat_extract_pdf_text_by_attachment_id($attachment_id) { | |
| 5464 | - $attachment_id = (int) $attachment_id; | |
| 5465 | - if ($attachment_id <= 0) { | |
| 5466 | - return ''; | |
| 5467 | - } | |
| 5468 | - if (get_post_mime_type($attachment_id) !== 'application/pdf') { | |
| 5469 | - return ''; | |
| 5470 | - } | |
| 5471 | - | |
| 5472 | - $pdf_path = get_attached_file($attachment_id); | |
| 5473 | - if (empty($pdf_path) || !file_exists($pdf_path) || !is_readable($pdf_path)) { | |
| 5474 | - return ''; | |
| 5475 | - } | |
| 5476 | - | |
| 5477 | - // Raw-file size cap. Parsing very large PDFs can OOM the request; skip with a log entry | |
| 5478 | - // and let the rest of the ACF content land in the KB. Filterable for users who need it bigger. | |
| 5479 | - $default_max_bytes = 25 * 1024 * 1024; | |
| 5480 | - $max_bytes = (int) apply_filters('mxchat_acf_pdf_max_bytes', $default_max_bytes, $attachment_id, $pdf_path); | |
| 5481 | - if ($max_bytes > 0) { | |
| 5482 | - $file_size = @filesize($pdf_path); | |
| 5483 | - if ($file_size !== false && $file_size > $max_bytes) { | |
| 5484 | - error_log(sprintf( | |
| 5485 | - '[mxchat] ACF PDF skipped (over size cap): attachment %d "%s" %d bytes > cap %d', | |
| 5486 | - $attachment_id, | |
| 5487 | - basename($pdf_path), | |
| 5488 | - $file_size, | |
| 5489 | - $max_bytes | |
| 5490 | - )); | |
| 5491 | - return ''; | |
| 5492 | - } | |
| 5493 | - } | |
| 5494 | - | |
| 5495 | - $mtime = @filemtime($pdf_path); | |
| 5496 | - $cache_meta_key = '_mxchat_acf_pdf_text_v1'; | |
| 5497 | - $cached = get_post_meta($attachment_id, $cache_meta_key, true); | |
| 5498 | - if (is_array($cached) && isset($cached['mtime'], $cached['text']) && (int) $cached['mtime'] === (int) $mtime) { | |
| 5499 | - return (string) $cached['text']; | |
| 5500 | - } | |
| 5501 | - | |
| 5502 | - $text = ''; | |
| 5503 | - try { | |
| 5504 | - if (function_exists('mxchat_load_pdf_parser')) { | |
| 5505 | - mxchat_load_pdf_parser(); | |
| 5506 | - } | |
| 5507 | - if (!class_exists('\\Smalot\\PdfParser\\Parser')) { | |
| 5508 | - return ''; | |
| 5509 | - } | |
| 5510 | - $parser = new \Smalot\PdfParser\Parser(); | |
| 5511 | - $pdf = $parser->parseFile($pdf_path); | |
| 5512 | - $pages = $pdf->getPages(); | |
| 5513 | - $page_texts = array(); | |
| 5514 | - foreach ($pages as $page) { | |
| 5515 | - $page_text = ''; | |
| 5516 | - try { | |
| 5517 | - $page_text = $page->getText(); | |
| 5518 | - } catch (\Exception $e) { | |
| 5519 | - $page_text = ''; | |
| 5520 | - } | |
| 5521 | - if (!empty($page_text)) { | |
| 5522 | - $page_texts[] = $page_text; | |
| 5523 | - } | |
| 5524 | - } | |
| 5525 | - $text = trim(implode("\n\n", $page_texts)); | |
| 5526 | - } catch (\Exception $e) { | |
| 5527 | - error_log('[mxchat] ACF PDF extraction failed for attachment ' . $attachment_id . ': ' . $e->getMessage()); | |
| 5528 | - return ''; | |
| 5529 | - } catch (\Throwable $e) { | |
| 5530 | - error_log('[mxchat] ACF PDF extraction error for attachment ' . $attachment_id . ': ' . $e->getMessage()); | |
| 5531 | - return ''; | |
| 5532 | - } | |
| 5533 | - | |
| 5534 | - // Cap per-PDF text to avoid blowing up the embedding payload on enormous PDFs. | |
| 5535 | - // The chunker downstream will still split this into multiple vectors. | |
| 5536 | - $max_len = (int) apply_filters('mxchat_acf_pdf_text_max_length', 50000); | |
| 5537 | - if ($max_len > 0 && strlen($text) > $max_len) { | |
| 5538 | - $text = substr($text, 0, $max_len); | |
| 5539 | - } | |
| 5540 | - | |
| 5541 | - update_post_meta($attachment_id, $cache_meta_key, array( | |
| 5542 | - 'mtime' => (int) $mtime, | |
| 5543 | - 'text' => $text, | |
| 5544 | - )); | |
| 5545 | - | |
| 5546 | - return $text; | |
| 5547 | -} | |
| 5548 | - | |
| 5549 | -/** | |
| 5550 | 5076 | * Handle ACF save - fires after ACF fields are saved |
| 5551 | 5077 | * This ensures ACF field data is available when syncing to knowledge base |
| 5552 | 5078 | */ |
| 5553 | 5079 | public function mxchat_handle_acf_save($post_id) { |
| @@ -5790,9 +5316,8 @@ | ||
| 5790 | 5316 | // ADD ACF FIELDS SUPPORT (matches ajax_mxchat_process_selected_content behavior) |
| 5791 | 5317 | $acf_fields = $this->mxchat_get_acf_fields_for_post($post_id); |
| 5792 | 5318 | if (!empty($acf_fields)) { |
| 5793 | 5319 | $acf_content_parts = array(); |
| 5794 | - $pdf_attachment_ids = array(); | |
| 5795 | 5320 | |
| 5796 | 5321 | foreach ($acf_fields as $field_name => $field_value) { |
| 5797 | 5322 | $formatted_value = $this->mxchat_format_acf_field_value($field_value, $field_name, $post_id); |
| 5798 | 5323 | if (!empty($formatted_value)) { |
| @@ -5799,44 +5324,13 @@ | ||
| 5799 | 5324 | // Convert field name to readable label |
| 5800 | 5325 | $field_label = ucwords(str_replace(['_', '-'], ' ', $field_name)); |
| 5801 | 5326 | $acf_content_parts[] = $field_label . ": " . $formatted_value; |
| 5802 | 5327 | } |
| 5803 | - | |
| 5804 | - $this->mxchat_collect_pdf_attachment_ids_from_acf_value($field_value, $pdf_attachment_ids); | |
| 5805 | 5328 | } |
| 5806 | 5329 | |
| 5807 | 5330 | if (!empty($acf_content_parts)) { |
| 5808 | 5331 | $final_content .= "\n\n" . implode("\n", $acf_content_parts); |
| 5809 | 5332 | } |
| 5810 | - | |
| 5811 | - // Gate the auto-sync PDF-extraction loop behind an opt-in option. | |
| 5812 | - // Mirrors the per-batch checkbox the manual content selector has; the | |
| 5813 | - // 25 MB size cap lives in the shared extractor so it applies in both | |
| 5814 | - // paths regardless. Default OFF — re-parsing every ACF PDF on every | |
| 5815 | - // editor save is expensive and most sites don't want it. | |
| 5816 | - $autosync_extract_acf_pdfs = get_option('mxchat_auto_sync_acf_pdfs', '0') === '1'; | |
| 5817 | - if ($autosync_extract_acf_pdfs && !empty($pdf_attachment_ids)) { | |
| 5818 | - $pdf_attachment_ids = array_unique(array_filter(array_map('intval', $pdf_attachment_ids))); | |
| 5819 | - $pdf_sections = array(); | |
| 5820 | - foreach ($pdf_attachment_ids as $att_id) { | |
| 5821 | - $pdf_text = $this->mxchat_extract_pdf_text_by_attachment_id($att_id); | |
| 5822 | - if (!empty($pdf_text)) { | |
| 5823 | - $pdf_title = get_the_title($att_id); | |
| 5824 | - $pdf_url = wp_get_attachment_url($att_id); | |
| 5825 | - $header = 'PDF Attachment'; | |
| 5826 | - if (!empty($pdf_title)) { | |
| 5827 | - $header .= ': ' . $pdf_title; | |
| 5828 | - } | |
| 5829 | - if (!empty($pdf_url)) { | |
| 5830 | - $header .= ' (' . $pdf_url . ')'; | |
| 5831 | - } | |
| 5832 | - $pdf_sections[] = $header . "\n" . $pdf_text; | |
| 5833 | - } | |
| 5834 | - } | |
| 5835 | - if (!empty($pdf_sections)) { | |
| 5836 | - $final_content .= "\n\nAttached PDFs:\n" . implode("\n\n", $pdf_sections); | |
| 5837 | - } | |
| 5838 | - } | |
| 5839 | 5333 | } |
| 5840 | 5334 | |
| 5841 | 5335 | // ADD CUSTOM POST META SUPPORT (whitelisted non-ACF meta fields) |
| 5842 | 5336 | $custom_meta = $this->mxchat_get_whitelisted_post_meta($post_id); |
| @@ -6854,16 +6348,16 @@ | ||
| 6854 | 6348 | wp_send_json_error('Unauthorized access'); |
| 6855 | 6349 | exit; |
| 6856 | 6350 | } |
| 6857 | 6351 | |
| 6858 | - $tag_input = isset($_POST['tag_slug']) ? sanitize_text_field($_POST['tag_slug']) : ''; | |
| 6352 | + $tag_slug = isset($_POST['tag_slug']) ? sanitize_text_field($_POST['tag_slug']) : ''; | |
| 6859 | 6353 | $role_restriction = isset($_POST['role_restriction']) ? sanitize_text_field($_POST['role_restriction']) : 'public'; |
| 6860 | - | |
| 6861 | - if (empty($tag_input)) { | |
| 6862 | - 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'); | |
| 6863 | 6357 | exit; |
| 6864 | 6358 | } |
| 6865 | - | |
| 6359 | + | |
| 6866 | 6360 | // Validate role restriction |
| 6867 | 6361 | $valid_roles = array_keys($this->mxchat_get_role_options()); |
| 6868 | 6362 | if (!in_array($role_restriction, $valid_roles)) { |
| 6869 | 6363 | wp_send_json_error('Invalid role restriction'); |
| @@ -6868,27 +6362,16 @@ | ||
| 6868 | 6362 | if (!in_array($role_restriction, $valid_roles)) { |
| 6869 | 6363 | wp_send_json_error('Invalid role restriction'); |
| 6870 | 6364 | exit; |
| 6871 | 6365 | } |
| 6872 | - | |
| 6873 | - // Resolve the tag by slug first, then fall back to its display name, so users can | |
| 6874 | - // enter either "premium-content" or "Premium Content". (plan b8bcf5 — the field is | |
| 6875 | - // labeled by name but previously validated by slug only, producing the confusing | |
| 6876 | - // "Tag does not exist in WordPress" error when a real tag's name was typed.) | |
| 6877 | - $term = get_term_by('slug', $tag_input, 'post_tag'); | |
| 6366 | + | |
| 6367 | + // Check if tag exists in WordPress | |
| 6368 | + $term = get_term_by('slug', $tag_slug, 'post_tag'); | |
| 6878 | 6369 | if (!$term) { |
| 6879 | - $term = get_term_by('name', $tag_input, 'post_tag'); | |
| 6880 | - } | |
| 6881 | - if (!$term) { | |
| 6882 | - 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'); | |
| 6883 | 6371 | exit; |
| 6884 | 6372 | } |
| 6885 | - | |
| 6886 | - // Always key the mapping by the RESOLVED slug — apply_role_restriction_to_post() | |
| 6887 | - // compares against each post's tag slugs, so the stored key must be a slug, | |
| 6888 | - // never the raw (possibly display-name) input. | |
| 6889 | - $tag_slug = $term->slug; | |
| 6890 | - | |
| 6373 | + | |
| 6891 | 6374 | // Get existing mappings |
| 6892 | 6375 | $mappings = get_option('mxchat_tag_role_mappings', array()); |
| 6893 | 6376 | |
| 6894 | 6377 | // Check if mapping already exists |
| @@ -7758,9 +7241,9 @@ | ||
| 7758 | 7241 | $is_likely_pdf = (strtolower(pathinfo(parse_url($url, PHP_URL_PATH) ?: '', PATHINFO_EXTENSION)) === 'pdf'); |
| 7759 | 7242 | $response = wp_remote_get($url, array( |
| 7760 | 7243 | 'timeout' => $is_likely_pdf ? 120 : 30, |
| 7761 | 7244 | 'redirection' => 5, |
| 7762 | - 'user-agent' => mxchat_ingest_user_agent(), | |
| 7245 | + 'user-agent' => 'MxChat/1.0' | |
| 7763 | 7246 | )); |
| 7764 | 7247 | |
| 7765 | 7248 | if (is_wp_error($response)) { |
| 7766 | 7249 | return $response; |
