| @@ -457,8 +457,100 @@ | ||
| 457 | 457 | return $options['embedding_model'] ?? ''; |
| 458 | 458 | } |
| 459 | 459 | |
| 460 | 460 | /** |
| 461 | + * Dimension of a STANDARD embedding model, from the dropdown and its two | |
| 462 | + * adjustable-size settings. Says nothing about a Custom Provider model — ask | |
| 463 | + * expected_embedding_dimension() for the model the site actually embeds with. | |
| 464 | + */ | |
| 465 | +public static function standard_embedding_dimension($options = null) { | |
| 466 | + if (!is_array($options)) { | |
| 467 | + $options = get_option('mxchat_options', array()); | |
| 468 | + $options = is_array($options) ? $options : array(); | |
| 469 | + } | |
| 470 | + $selected_model = $options['embedding_model'] ?? 'text-embedding-ada-002'; | |
| 471 | + $model_dimensions = array( | |
| 472 | + 'text-embedding-ada-002' => 1536, | |
| 473 | + 'text-embedding-3-small' => 1536, | |
| 474 | + 'text-embedding-3-large' => 3072, | |
| 475 | + 'voyage-2' => 1024, | |
| 476 | + 'voyage-large-2' => 1536, | |
| 477 | + 'voyage-3-large' => 2048, | |
| 478 | + 'gemini-embedding-001' => 1536, | |
| 479 | + ); | |
| 480 | + if (strpos($selected_model, 'voyage-3-large') === 0) { | |
| 481 | + return intval($options['voyage_output_dimension'] ?? 2048); | |
| 482 | + } | |
| 483 | + if (strpos($selected_model, 'gemini-embedding') === 0) { | |
| 484 | + return intval($options['gemini_output_dimension'] ?? 1536); | |
| 485 | + } | |
| 486 | + return $model_dimensions[$selected_model] ?? 1536; | |
| 487 | +} | |
| 488 | + | |
| 489 | +/** | |
| 490 | + * Dimension of the vectors the NEXT embed will produce — the one place that | |
| 491 | + * answers it (plan 418afd). With Custom Provider embeddings on, the standard | |
| 492 | + * dropdown is disabled and says nothing about the custom model, so the number | |
| 493 | + * is read from the model itself: one short probe embed, remembered per base | |
| 494 | + * URL + model. Three copies of the dropdown table used to answer instead, and | |
| 495 | + * a custom model of another size got a wrong-sized index and wrong-sized | |
| 496 | + * listing vectors. | |
| 497 | + * | |
| 498 | + * @param array|null $options mxchat_options (or a bot's merged options). | |
| 499 | + * @param bool $force Probe again even when a recent probe failed. | |
| 500 | + * @return int|WP_Error WP_Error only when a custom model's size could not be read. | |
| 501 | + */ | |
| 502 | +public static function expected_embedding_dimension($options = null, $force = false) { | |
| 503 | + if (!is_array($options)) { | |
| 504 | + $options = get_option('mxchat_options', array()); | |
| 505 | + $options = is_array($options) ? $options : array(); | |
| 506 | + } | |
| 507 | + if (isset($options['custom_provider_for_embeddings']) && $options['custom_provider_for_embeddings'] === 'on') { | |
| 508 | + return self::custom_embedding_dimension($options, $force); | |
| 509 | + } | |
| 510 | + return self::standard_embedding_dimension($options); | |
| 511 | +} | |
| 512 | + | |
| 513 | +/** | |
| 514 | + * A Custom Provider embedding model's dimension, by asking it once. The | |
| 515 | + * answer is stored against the base URL + model, so changing either asks | |
| 516 | + * again. A failure is remembered for five minutes so a dead endpoint does not | |
| 517 | + * hold up every admin page that prints the number; $force skips that memory. | |
| 518 | + * | |
| 519 | + * @return int|WP_Error | |
| 520 | + */ | |
| 521 | +public static function custom_embedding_dimension($options, $force = false) { | |
| 522 | + $model = self::resolve_custom_embedding_model($options); | |
| 523 | + $base_url = isset($options['custom_provider_base_url']) ? rtrim(trim((string) $options['custom_provider_base_url']), '/') : ''; | |
| 524 | + $key = md5(strtolower($base_url) . '|' . $model); | |
| 525 | + | |
| 526 | + $known = get_option('mxchat_custom_embedding_dimension', array()); | |
| 527 | + if (is_array($known) && ($known['key'] ?? '') === $key && (int) ($known['dimension'] ?? 0) > 0) { | |
| 528 | + return (int) $known['dimension']; | |
| 529 | + } | |
| 530 | + | |
| 531 | + $failed = get_transient('mxchat_custom_embedding_dimension_failed'); | |
| 532 | + if (!$force && is_array($failed) && ($failed['key'] ?? '') === $key) { | |
| 533 | + return new WP_Error('embedding_dimension_unknown', (string) ($failed['reason'] ?? '')); | |
| 534 | + } | |
| 535 | + | |
| 536 | + $vector = self::generate_embedding_custom('MxChat embedding dimension check', $options); | |
| 537 | + if (!is_array($vector) || count($vector) === 0) { | |
| 538 | + $reason = (is_string($vector) && $vector !== '') ? $vector : __('The custom provider returned no embedding.', 'mxchat'); | |
| 539 | + set_transient('mxchat_custom_embedding_dimension_failed', array('key' => $key, 'reason' => $reason), 5 * MINUTE_IN_SECONDS); | |
| 540 | + return new WP_Error('embedding_dimension_unknown', $reason); | |
| 541 | + } | |
| 542 | + | |
| 543 | + update_option('mxchat_custom_embedding_dimension', array( | |
| 544 | + 'key' => $key, | |
| 545 | + 'model' => 'custom:' . $model, | |
| 546 | + 'dimension' => count($vector), | |
| 547 | + ), false); | |
| 548 | + delete_transient('mxchat_custom_embedding_dimension_failed'); | |
| 549 | + return count($vector); | |
| 550 | +} | |
| 551 | + | |
| 552 | +/** | |
| 461 | 553 | * Extract the 11-character YouTube video ID from a URL, or '' if the URL is |
| 462 | 554 | * not a single-video YouTube link. Single source of truth for both the KB |
| 463 | 555 | * ingestion side and the chat render side — do not duplicate this parsing. |
| 464 | 556 | * Channel, playlist, and search URLs deliberately return '' (only a URL that |
| @@ -535,9 +627,9 @@ | ||
| 535 | 627 | * @param string $bot_id The bot ID for multi-bot support |
| 536 | 628 | * @param string $content_type The type of content (post, page, pdf, url, manual, product, etc.) |
| 537 | 629 | * @return bool|WP_Error True on success, WP_Error on failure |
| 538 | 630 | */ |
| 539 | -public static function submit_content_to_db($content, $source_url, $api_key, $vector_id = null, $bot_id = 'default', $content_type = 'content') { | |
| 631 | +public static function submit_content_to_db($content, $source_url, $api_key, $vector_id = null, $bot_id = 'default', $content_type = 'content', $title = '') { | |
| 540 | 632 | global $wpdb; |
| 541 | 633 | $table_name = $wpdb->prefix . 'mxchat_system_prompt_content'; |
| 542 | 634 | |
| 543 | 635 | //error_log('[MXCHAT-DB] Starting database submission for URL: ' . $source_url . ' (Bot: ' . $bot_id . ', Type: ' . $content_type . ')'); |
| @@ -559,8 +651,14 @@ | ||
| 559 | 651 | if (empty($content_type)) { |
| 560 | 652 | $content_type = 'content'; // Fallback for backwards compatibility |
| 561 | 653 | } |
| 562 | 654 | |
| 655 | + // Page title captured by the importer (plan d9ee66), registered under the | |
| 656 | + // SANITIZED url so the storage writers below find it by the same key. | |
| 657 | + if (is_string($title) && trim($title) !== '') { | |
| 658 | + self::set_source_title($source_url, $title); | |
| 659 | + } | |
| 660 | + | |
| 563 | 661 | // Just ensure UTF-8 validity without aggressive escaping |
| 564 | 662 | $safe_content = wp_check_invalid_utf8($content); |
| 565 | 663 | // Remove only null bytes and other control characters, but preserve newlines (\n = \x0A) and carriage returns (\r = \x0D) |
| 566 | 664 | $safe_content = preg_replace('/[\x00-\x08\x0B\x0C\x0E-\x1F\x7F]/u', '', $safe_content); |
| @@ -620,9 +718,150 @@ | ||
| 620 | 718 | return $result; |
| 621 | 719 | } |
| 622 | 720 | } |
| 623 | 721 | |
| 722 | +/* ------------------------------------------------------------------------- | |
| 723 | + * Source titles (3.2.23, plan d9ee66) | |
| 724 | + * | |
| 725 | + * A knowledge entry imported from a URL or sitemap remembers the page's own | |
| 726 | + * title, so an add-on that names a source (AI Search's source cards) can show | |
| 727 | + * "About Us" instead of a guess from the slug ("Aboutus"). Stored in the | |
| 728 | + * `title` column of the WP knowledge table and as `title` metadata on Pinecone | |
| 729 | + * vectors / documents. The importer hands the title to submit_content_to_db(), | |
| 730 | + * which registers it under the source URL; the writers look it up by that key. | |
| 731 | + * ---------------------------------------------------------------------- */ | |
| 732 | + | |
| 733 | +/** @var array<string,string> source_url => title, for the current request */ | |
| 734 | +private static $source_titles = array(); | |
| 735 | + | |
| 736 | +public static function set_source_title($source_url, $title) { | |
| 737 | + $key = trim((string) $source_url); | |
| 738 | + $title = self::clean_title($title); | |
| 739 | + if ($key === '' || $title === '') { | |
| 740 | + return; | |
| 741 | + } | |
| 742 | + self::$source_titles[$key] = $title; | |
| 743 | +} | |
| 744 | + | |
| 745 | +public static function source_title_for($source_url) { | |
| 746 | + $key = trim((string) $source_url); | |
| 747 | + return ($key !== '' && isset(self::$source_titles[$key])) ? self::$source_titles[$key] : ''; | |
| 748 | +} | |
| 749 | + | |
| 750 | +/** Entities decoded, tags gone, whitespace collapsed, capped at the column width. */ | |
| 751 | +public static function clean_title($title) { | |
| 752 | + $title = html_entity_decode((string) $title, ENT_QUOTES | ENT_HTML5, 'UTF-8'); | |
| 753 | + $title = wp_strip_all_tags($title); | |
| 754 | + $title = trim((string) preg_replace('/\s+/u', ' ', $title)); | |
| 755 | + if (mb_strlen($title, 'UTF-8') > 255) { | |
| 756 | + $title = rtrim(mb_substr($title, 0, 255, 'UTF-8')); | |
| 757 | + } | |
| 758 | + return $title; | |
| 759 | +} | |
| 760 | + | |
| 624 | 761 | /** |
| 762 | + * The page's own title from fetched HTML: og:title, else <title> with a | |
| 763 | + * trailing "| Site Name" style suffix removed, else the first <h1>. Returns '' | |
| 764 | + * when nothing usable is there, so the caller keeps its own fallback. | |
| 765 | + */ | |
| 766 | +public static function extract_page_title($html) { | |
| 767 | + $html = (string) $html; | |
| 768 | + if ($html === '') { | |
| 769 | + return ''; | |
| 770 | + } | |
| 771 | + $head = substr($html, 0, 200000); | |
| 772 | + $site = ''; | |
| 773 | + if (preg_match('/<meta\s+[^>]*property=["\']og:site_name["\'][^>]*content=["\']([^"\']{1,120})["\']/i', $head, $m) | |
| 774 | + || preg_match('/<meta\s+[^>]*content=["\']([^"\']{1,120})["\'][^>]*property=["\']og:site_name["\']/i', $head, $m)) { | |
| 775 | + $site = self::clean_title($m[1]); | |
| 776 | + } | |
| 777 | + $candidate = ''; | |
| 778 | + if (preg_match('/<meta\s+[^>]*property=["\']og:title["\'][^>]*content=["\']([^"\']{1,500})["\']/i', $head, $m) | |
| 779 | + || preg_match('/<meta\s+[^>]*content=["\']([^"\']{1,500})["\'][^>]*property=["\']og:title["\']/i', $head, $m)) { | |
| 780 | + $candidate = self::clean_title($m[1]); | |
| 781 | + } | |
| 782 | + if ($candidate === '' && preg_match('/<title[^>]*>(.*?)<\/title>/is', $head, $m)) { | |
| 783 | + $candidate = self::clean_title($m[1]); | |
| 784 | + } | |
| 785 | + if ($candidate !== '') { | |
| 786 | + $candidate = self::strip_site_suffix($candidate, $site); | |
| 787 | + } | |
| 788 | + if ($candidate === '' && preg_match('/<h1[^>]*>(.*?)<\/h1>/is', $html, $m)) { | |
| 789 | + $candidate = self::clean_title($m[1]); | |
| 790 | + } | |
| 791 | + return $candidate; | |
| 792 | +} | |
| 793 | + | |
| 794 | +/** | |
| 795 | + * "About Us | Just Facts" -> "About Us". Only a TRAILING segment after a spaced | |
| 796 | + * separator goes, only when what is left is a real title (3+ characters) and the | |
| 797 | + * removed part is short enough to be a site name; a known og:site_name wins. | |
| 798 | + * Byte offsets throughout: every separator starts with an ASCII space, so a cut | |
| 799 | + * there never lands inside a multibyte character. | |
| 800 | + */ | |
| 801 | +private static function strip_site_suffix($title, $site_name = '') { | |
| 802 | + $seps = array(' | ', ' – ', ' — ', ' - ', ' :: ', ' · ', ' » ', ' • '); | |
| 803 | + if ($site_name !== '') { | |
| 804 | + foreach ($seps as $sep) { | |
| 805 | + $tail = $sep . $site_name; | |
| 806 | + if (strlen($title) > strlen($tail) && substr($title, -strlen($tail)) === $tail) { | |
| 807 | + return trim(substr($title, 0, strlen($title) - strlen($tail))); | |
| 808 | + } | |
| 809 | + } | |
| 810 | + } | |
| 811 | + $best_pos = -1; | |
| 812 | + $best_sep = ''; | |
| 813 | + foreach ($seps as $sep) { | |
| 814 | + $pos = strrpos($title, $sep); | |
| 815 | + if ($pos !== false && $pos > $best_pos) { | |
| 816 | + $best_pos = $pos; | |
| 817 | + $best_sep = $sep; | |
| 818 | + } | |
| 819 | + } | |
| 820 | + if ($best_pos > 0) { | |
| 821 | + $left = trim(substr($title, 0, $best_pos)); | |
| 822 | + $right = trim(substr($title, $best_pos + strlen($best_sep))); | |
| 823 | + if (mb_strlen($left, 'UTF-8') >= 3 && $right !== '' && mb_strlen($right, 'UTF-8') <= 60) { | |
| 824 | + return $left; | |
| 825 | + } | |
| 826 | + } | |
| 827 | + return $title; | |
| 828 | +} | |
| 829 | + | |
| 830 | +/** | |
| 831 | + * Make sure the knowledge table has the `title` column (3.2.23, plan d9ee66). | |
| 832 | + * Activation adds it; this is the safety net for a copy updated in place, and | |
| 833 | + * the one place add-ons ask "may I read/write title?". A positive answer is | |
| 834 | + * remembered in an option, so after the first time this costs one option read. | |
| 835 | + */ | |
| 836 | +public static function ensure_kb_title_column() { | |
| 837 | + global $wpdb; | |
| 838 | + static $ready = null; | |
| 839 | + if ($ready !== null) { | |
| 840 | + return $ready; | |
| 841 | + } | |
| 842 | + if (get_option('mxchat_kb_title_column') === '1') { | |
| 843 | + $ready = true; | |
| 844 | + return true; | |
| 845 | + } | |
| 846 | + $table = $wpdb->prefix . 'mxchat_system_prompt_content'; | |
| 847 | + $cols = $wpdb->get_col("SHOW COLUMNS FROM {$table}", 0); | |
| 848 | + if (empty($cols)) { | |
| 849 | + $ready = false; | |
| 850 | + return false; | |
| 851 | + } | |
| 852 | + if (!in_array('title', $cols, true)) { | |
| 853 | + $wpdb->query("ALTER TABLE {$table} ADD COLUMN title VARCHAR(255) DEFAULT NULL AFTER source_url"); | |
| 854 | + $cols = $wpdb->get_col("SHOW COLUMNS FROM {$table}", 0); | |
| 855 | + } | |
| 856 | + $ready = in_array('title', $cols, true); | |
| 857 | + if ($ready) { | |
| 858 | + update_option('mxchat_kb_title_column', '1', false); | |
| 859 | + } | |
| 860 | + return $ready; | |
| 861 | +} | |
| 862 | + | |
| 863 | +/** | |
| 625 | 864 | * UPDATED: Check if Pinecone is enabled and properly configured for a specific bot |
| 626 | 865 | */ |
| 627 | 866 | private static function is_pinecone_enabled_for_bot($bot_id = 'default') { |
| 628 | 867 | // For default bot or when multi-bot is not active, use original method |
| @@ -763,8 +1002,13 @@ | ||
| 763 | 1002 | //error_log('[MXCHAT-DB] Manual content - will create new entry (no duplicate check)'); |
| 764 | 1003 | } |
| 765 | 1004 | // ===== END FIX ===== |
| 766 | 1005 | |
| 1006 | + // Page title captured at import (plan d9ee66): written only when the | |
| 1007 | + // importer supplied one, so a re-index without a title keeps the old value. | |
| 1008 | + $title_value = $is_manual_content ? '' : self::source_title_for($source_url); | |
| 1009 | + $title_ready = ($title_value !== '') && self::ensure_kb_title_column(); | |
| 1010 | + | |
| 767 | 1011 | // Progressive fallback mechanism for problematic content |
| 768 | 1012 | $attempt = 1; |
| 769 | 1013 | $max_attempts = 3; |
| 770 | 1014 | $current_content = $safe_content; |
| @@ -775,39 +1019,41 @@ | ||
| 775 | 1019 | if ($existing_id) { |
| 776 | 1020 | //error_log('[MXCHAT-DB] Found existing entry (ID: ' . $existing_id . '). Updating... (Attempt ' . $attempt . ')'); |
| 777 | 1021 | |
| 778 | 1022 | // Update the existing row - UPDATED 2.5.6: Added content_type |
| 779 | - $result = $wpdb->update( | |
| 780 | - $table_name, | |
| 781 | - array( | |
| 782 | - 'url' => $source_url, | |
| 783 | - 'article_content' => $current_content, | |
| 784 | - 'embedding_vector' => $embedding_vector_serialized, | |
| 785 | - 'source_url' => $source_url, | |
| 786 | - 'content_type' => $content_type, | |
| 787 | - 'timestamp' => current_time('mysql'), | |
| 788 | - ), | |
| 789 | - array('id' => $existing_id), | |
| 790 | - array('%s','%s','%s','%s','%s','%s'), | |
| 791 | - array('%d') | |
| 1023 | + $row = array( | |
| 1024 | + 'url' => $source_url, | |
| 1025 | + 'article_content' => $current_content, | |
| 1026 | + 'embedding_vector' => $embedding_vector_serialized, | |
| 1027 | + 'source_url' => $source_url, | |
| 1028 | + 'content_type' => $content_type, | |
| 1029 | + 'timestamp' => current_time('mysql'), | |
| 792 | 1030 | ); |
| 1031 | + $formats = array('%s','%s','%s','%s','%s','%s'); | |
| 1032 | + if ($title_ready) { | |
| 1033 | + $row['title'] = $title_value; | |
| 1034 | + $formats[] = '%s'; | |
| 1035 | + } | |
| 1036 | + $result = $wpdb->update($table_name, $row, array('id' => $existing_id), $formats, array('%d')); | |
| 793 | 1037 | } else { |
| 794 | 1038 | //error_log('[MXCHAT-DB] No existing entry found. Inserting new row... (Attempt ' . $attempt . ')'); |
| 795 | 1039 | //error_log('[MXCHAT-DB] Content sample: ' . substr($current_content, 0, 1000)); |
| 796 | 1040 | |
| 797 | 1041 | // Insert a new row - UPDATED 2.5.6: Added content_type |
| 798 | - $result = $wpdb->insert( | |
| 799 | - $table_name, | |
| 800 | - array( | |
| 801 | - 'url' => $source_url, // Now unique for manual content | |
| 802 | - 'article_content' => $current_content, | |
| 803 | - 'embedding_vector' => $embedding_vector_serialized, | |
| 804 | - 'source_url' => $source_url, // Now unique for manual content | |
| 805 | - 'content_type' => $content_type, | |
| 806 | - 'timestamp' => current_time('mysql'), | |
| 807 | - ), | |
| 808 | - array('%s','%s','%s','%s','%s','%s') | |
| 1042 | + $row = array( | |
| 1043 | + 'url' => $source_url, // Now unique for manual content | |
| 1044 | + 'article_content' => $current_content, | |
| 1045 | + 'embedding_vector' => $embedding_vector_serialized, | |
| 1046 | + 'source_url' => $source_url, // Now unique for manual content | |
| 1047 | + 'content_type' => $content_type, | |
| 1048 | + 'timestamp' => current_time('mysql'), | |
| 809 | 1049 | ); |
| 1050 | + $formats = array('%s','%s','%s','%s','%s','%s'); | |
| 1051 | + if ($title_ready) { | |
| 1052 | + $row['title'] = $title_value; | |
| 1053 | + $formats[] = '%s'; | |
| 1054 | + } | |
| 1055 | + $result = $wpdb->insert($table_name, $row, $formats); | |
| 810 | 1056 | } |
| 811 | 1057 | |
| 812 | 1058 | if ($result === false) { |
| 813 | 1059 | //error_log('[MXCHAT-DB] Database operation failed (Attempt ' . $attempt . ')'); |
| @@ -931,8 +1177,12 @@ | ||
| 931 | 1177 | 'last_updated' => time(), |
| 932 | 1178 | 'created_at' => time(), // Add creation timestamp |
| 933 | 1179 | 'bot_id' => $bot_id, // Add bot identification |
| 934 | 1180 | ); |
| 1181 | + $page_title = self::source_title_for($url); | |
| 1182 | + if ($page_title !== '') { | |
| 1183 | + $metadata['title'] = $page_title; // page title captured at import (plan d9ee66) | |
| 1184 | + } | |
| 935 | 1185 | |
| 936 | 1186 | // Document index (plan 362c31): same id, same metadata, through the |
| 937 | 1187 | // documents API. A site on the classic vector index never enters here. |
| 938 | 1188 | if (class_exists('MxChat_Pinecone_Documents') && MxChat_Pinecone_Documents::is_document_index($bot_id)) { |
| @@ -1619,8 +1869,12 @@ | ||
| 1619 | 1869 | 'last_updated' => time(), |
| 1620 | 1870 | 'created_at' => time(), |
| 1621 | 1871 | 'bot_id' => $bot_id, |
| 1622 | 1872 | ); |
| 1873 | + $page_title = self::source_title_for($source_url); | |
| 1874 | + if ($page_title !== '') { | |
| 1875 | + $metadata['title'] = $page_title; // page title captured at import (plan d9ee66) | |
| 1876 | + } | |
| 1623 | 1877 | |
| 1624 | 1878 | // Document index (plan 362c31): same chunk id and metadata, documents API. |
| 1625 | 1879 | if (class_exists('MxChat_Pinecone_Documents') && MxChat_Pinecone_Documents::is_document_index($bot_id)) { |
| 1626 | 1880 | return MxChat_Pinecone_Documents::upsert_one($vector_id, $chunk_text, $embedding_vector, $metadata, array( |
| @@ -1673,20 +1927,25 @@ | ||
| 1673 | 1927 | global $wpdb; |
| 1674 | 1928 | |
| 1675 | 1929 | // For chunks, we always insert new rows (no duplicate checking) |
| 1676 | 1930 | // The URL includes chunk info in the metadata, but source_url stays the same for grouping |
| 1677 | - $result = $wpdb->insert( | |
| 1678 | - $table_name, | |
| 1679 | - array( | |
| 1680 | - 'url' => $source_url, | |
| 1681 | - 'article_content' => $content_with_metadata, | |
| 1682 | - 'embedding_vector' => $embedding_vector_serialized, | |
| 1683 | - 'source_url' => $source_url, | |
| 1684 | - 'content_type' => $content_type, | |
| 1685 | - 'timestamp' => current_time('mysql') | |
| 1686 | - ), | |
| 1687 | - array('%s', '%s', '%s', '%s', '%s', '%s') | |
| 1931 | + $row = array( | |
| 1932 | + 'url' => $source_url, | |
| 1933 | + 'article_content' => $content_with_metadata, | |
| 1934 | + 'embedding_vector' => $embedding_vector_serialized, | |
| 1935 | + 'source_url' => $source_url, | |
| 1936 | + 'content_type' => $content_type, | |
| 1937 | + 'timestamp' => current_time('mysql') | |
| 1688 | 1938 | ); |
| 1939 | + $formats = array('%s', '%s', '%s', '%s', '%s', '%s'); | |
| 1940 | + // Every chunk row carries the page title too (plan d9ee66) - readers group by | |
| 1941 | + // source_url and may pick any chunk as the representative. | |
| 1942 | + $page_title = self::source_title_for($source_url); | |
| 1943 | + if ($page_title !== '' && self::ensure_kb_title_column()) { | |
| 1944 | + $row['title'] = $page_title; | |
| 1945 | + $formats[] = '%s'; | |
| 1946 | + } | |
| 1947 | + $result = $wpdb->insert($table_name, $row, $formats); | |
| 1689 | 1948 | |
| 1690 | 1949 | if ($result === false) { |
| 1691 | 1950 | return new WP_Error('database_failed', 'Failed to insert chunk: ' . $wpdb->last_error); |
| 1692 | 1951 | } |
| @@ -1694,8 +1953,142 @@ | ||
| 1694 | 1953 | return true; |
| 1695 | 1954 | } |
| 1696 | 1955 | |
| 1697 | 1956 | /** |
| 1957 | + * Option holding the page an uploaded file should be cited as (plan 025ea2). | |
| 1958 | + * Keyed "<bot_id>|upload://<filename>" => https URL. The entry itself keeps | |
| 1959 | + * its upload:// identity, so a re-upload of the same file still replaces the | |
| 1960 | + * same entry and never collides with the cited page's own knowledge entry. | |
| 1961 | + */ | |
| 1962 | +const UPLOAD_SOURCE_URLS_OPTION = 'mxchat_upload_source_urls'; | |
| 1963 | + | |
| 1964 | +/** | |
| 1965 | + * Whether a stored source is an uploaded file (Document Upload / PDF Upload). | |
| 1966 | + * | |
| 1967 | + * @param string $source_url Stored source identity. | |
| 1968 | + * @return bool | |
| 1969 | + */ | |
| 1970 | +public static function is_upload_source($source_url) { | |
| 1971 | + return is_string($source_url) && stripos($source_url, 'upload://') === 0; | |
| 1972 | +} | |
| 1973 | + | |
| 1974 | +/** | |
| 1975 | + * File name of an uploaded source, for a plain-text label. | |
| 1976 | + * | |
| 1977 | + * @param string $source_url Stored source identity (upload://<filename>). | |
| 1978 | + * @return string | |
| 1979 | + */ | |
| 1980 | +public static function upload_source_name($source_url) { | |
| 1981 | + return self::is_upload_source($source_url) ? substr(self::upload_source_base($source_url), strlen('upload://')) : ''; | |
| 1982 | +} | |
| 1983 | + | |
| 1984 | +/** | |
| 1985 | + * An uploaded source without its page fragment (plan dcc19d). A page of an | |
| 1986 | + * uploaded PDF is stored as upload://<filename>#page=<n>; the file it belongs | |
| 1987 | + * to, and the key its Source URL is kept under, is upload://<filename>. | |
| 1988 | + * | |
| 1989 | + * @param string $source_url Stored source identity. | |
| 1990 | + * @return string | |
| 1991 | + */ | |
| 1992 | +public static function upload_source_base($source_url) { | |
| 1993 | + return is_string($source_url) ? preg_replace('/#page=\d+$/', '', $source_url) : ''; | |
| 1994 | +} | |
| 1995 | + | |
| 1996 | +/** | |
| 1997 | + * Page number carried by an uploaded PDF page's identity, 0 when it has none. | |
| 1998 | + * | |
| 1999 | + * @param string $source_url Stored source identity. | |
| 2000 | + * @return int | |
| 2001 | + */ | |
| 2002 | +public static function upload_source_page($source_url) { | |
| 2003 | + return (self::is_upload_source($source_url) && preg_match('/#page=(\d+)$/', $source_url, $m)) ? (int) $m[1] : 0; | |
| 2004 | +} | |
| 2005 | + | |
| 2006 | +/** | |
| 2007 | + * Identity one page of a PDF is stored under: <pdf>#page=<n>. A web PDF goes | |
| 2008 | + * through esc_url() as before. An uploaded PDF (upload://<filename>) must not: | |
| 2009 | + * esc_url() empties that scheme, which stored every page of an uploaded PDF | |
| 2010 | + * as anonymous manual content, so the file could not be named, cited, given | |
| 2011 | + * a Source URL, or replaced by a re-upload (plan dcc19d). | |
| 2012 | + * | |
| 2013 | + * @param string $pdf_url The PDF's own source (URL or upload://<filename>). | |
| 2014 | + * @param int $page_number 1-based page number. | |
| 2015 | + * @return string | |
| 2016 | + */ | |
| 2017 | +public static function pdf_page_identity($pdf_url, $page_number) { | |
| 2018 | + $identity = $pdf_url . '#page=' . (int) $page_number; | |
| 2019 | + return self::is_upload_source($pdf_url) ? sanitize_text_field($identity) : esc_url($identity); | |
| 2020 | +} | |
| 2021 | + | |
| 2022 | +/** | |
| 2023 | + * Remember (or clear) the page an uploaded file should be cited as. | |
| 2024 | + * Only http(s) URLs are kept; anything else clears the entry. | |
| 2025 | + * | |
| 2026 | + * @param string $source_label upload://<filename> | |
| 2027 | + * @param string $bot_id Bot the upload belongs to. | |
| 2028 | + * @param string $url Page URL, or '' to clear. | |
| 2029 | + * @return string The URL stored, '' when cleared. | |
| 2030 | + */ | |
| 2031 | +public static function set_upload_source_url($source_label, $bot_id, $url) { | |
| 2032 | + if (!self::is_upload_source($source_label)) { | |
| 2033 | + return ''; | |
| 2034 | + } | |
| 2035 | + $bot_id = ($bot_id === '' || $bot_id === null) ? 'default' : (string) $bot_id; | |
| 2036 | + $url = is_string($url) ? esc_url_raw(trim($url), array('http', 'https')) : ''; | |
| 2037 | + if ($url !== '' && !preg_match('#^https?://#i', $url)) { | |
| 2038 | + $url = ''; | |
| 2039 | + } | |
| 2040 | + | |
| 2041 | + $map = get_option(self::UPLOAD_SOURCE_URLS_OPTION, array()); | |
| 2042 | + if (!is_array($map)) { | |
| 2043 | + $map = array(); | |
| 2044 | + } | |
| 2045 | + $key = $bot_id . '|' . $source_label; | |
| 2046 | + if ($url === '') { | |
| 2047 | + if (!isset($map[$key])) { | |
| 2048 | + return ''; | |
| 2049 | + } | |
| 2050 | + unset($map[$key]); | |
| 2051 | + } else { | |
| 2052 | + $map[$key] = $url; | |
| 2053 | + } | |
| 2054 | + if (empty($map)) { | |
| 2055 | + delete_option(self::UPLOAD_SOURCE_URLS_OPTION); | |
| 2056 | + } else { | |
| 2057 | + update_option(self::UPLOAD_SOURCE_URLS_OPTION, $map, false); | |
| 2058 | + } | |
| 2059 | + return $url; | |
| 2060 | +} | |
| 2061 | + | |
| 2062 | +/** | |
| 2063 | + * The page an uploaded file is cited as, '' when it has none. | |
| 2064 | + * Looks under the asking bot first, then under the default bot (on the | |
| 2065 | + * WordPress database store every bot reads the same entries). | |
| 2066 | + * | |
| 2067 | + * @param string $source_label upload://<filename> | |
| 2068 | + * @param string $bot_id Bot answering the turn. | |
| 2069 | + * @return string | |
| 2070 | + */ | |
| 2071 | +public static function get_upload_source_url($source_label, $bot_id = 'default') { | |
| 2072 | + if (!self::is_upload_source($source_label)) { | |
| 2073 | + return ''; | |
| 2074 | + } | |
| 2075 | + $map = get_option(self::UPLOAD_SOURCE_URLS_OPTION, array()); | |
| 2076 | + if (!is_array($map) || empty($map)) { | |
| 2077 | + return ''; | |
| 2078 | + } | |
| 2079 | + // A PDF page (upload://<file>#page=<n>) is cited as its file's page. | |
| 2080 | + $source_label = self::upload_source_base($source_label); | |
| 2081 | + $bot_id = ($bot_id === '' || $bot_id === null) ? 'default' : (string) $bot_id; | |
| 2082 | + foreach (array($bot_id . '|' . $source_label, 'default|' . $source_label) as $key) { | |
| 2083 | + if (!empty($map[$key]) && preg_match('#^https?://#i', $map[$key])) { | |
| 2084 | + return $map[$key]; | |
| 2085 | + } | |
| 2086 | + } | |
| 2087 | + return ''; | |
| 2088 | +} | |
| 2089 | + | |
| 2090 | +/** | |
| 1698 | 2091 | * Delete all chunks for a given URL |
| 1699 | 2092 | * |
| 1700 | 2093 | * @param string $source_url The source URL |
| 1701 | 2094 | * @param string $bot_id The bot ID |
| @@ -1710,12 +2103,124 @@ | ||
| 1710 | 2103 | MxChat_Vectorstore_Manager::sync_delete_entry($source_url, $bot_id); |
| 1711 | 2104 | } |
| 1712 | 2105 | |
| 1713 | 2106 | if (self::is_pinecone_enabled_for_bot($bot_id)) { |
| 1714 | - return self::delete_pinecone_chunks_by_url($source_url, $bot_id); | |
| 2107 | + $result = self::delete_pinecone_chunks_by_url($source_url, $bot_id); | |
| 1715 | 2108 | } else { |
| 1716 | - return self::delete_wordpress_chunks_by_url($source_url); | |
| 2109 | + $result = self::delete_wordpress_chunks_by_url($source_url); | |
| 1717 | 2110 | } |
| 2111 | + | |
| 2112 | + // The entry is gone for good (not the clean-slate half of a re-store): | |
| 2113 | + // a media file's "already indexed, nothing changed" signature must go | |
| 2114 | + // with it, or auto-sync keeps skipping the file on every later save. | |
| 2115 | + if (!self::$vectorstore_mirror_suspended && !is_wp_error($result)) { | |
| 2116 | + self::forget_attachment_signature($source_url, $bot_id); | |
| 2117 | + } | |
| 2118 | + | |
| 2119 | + return $result; | |
| 2120 | +} | |
| 2121 | + | |
| 2122 | +/** | |
| 2123 | + * Clear the media auto-sync signature (_mxchat_kb_attachment_sig) of the | |
| 2124 | + * attachment a just-deleted knowledge entry was indexed from. | |
| 2125 | + * | |
| 2126 | + * Call it wherever a WHOLE entry is removed (never for the clean-slate half | |
| 2127 | + * of a re-store, and never when sibling chunk rows survive). Posts and pages | |
| 2128 | + * carry no signature; for them this is two indexed lookups that find nothing. | |
| 2129 | + * | |
| 2130 | + * @param string $source_url The entry's source URL. | |
| 2131 | + * @param string $bot_id The bot whose entry was removed. | |
| 2132 | + */ | |
| 2133 | +public static function forget_attachment_signature($source_url, $bot_id = 'default') { | |
| 2134 | + global $wpdb; | |
| 2135 | + | |
| 2136 | + $source_url = (string) $source_url; | |
| 2137 | + if (!preg_match('#^https?://#i', $source_url) || !self::attachment_signature_follows_bot($bot_id)) { | |
| 2138 | + return; | |
| 2139 | + } | |
| 2140 | + | |
| 2141 | + $attachment_ids = array(); | |
| 2142 | + $by_url = attachment_url_to_postid($source_url); | |
| 2143 | + if ($by_url) { | |
| 2144 | + $attachment_ids[] = (int) $by_url; | |
| 2145 | + } | |
| 2146 | + | |
| 2147 | + // The signature stores the exact URL the indexer used, which can differ | |
| 2148 | + // from where the file lives now (replaced file, scaled variant). Bounded | |
| 2149 | + // by the number of signed attachments: the meta_key index does the work. | |
| 2150 | + $candidates = $wpdb->get_results($wpdb->prepare( | |
| 2151 | + "SELECT post_id, meta_value FROM {$wpdb->postmeta} | |
| 2152 | + WHERE meta_key = '_mxchat_kb_attachment_sig' AND meta_value LIKE %s | |
| 2153 | + LIMIT 20", | |
| 2154 | + '%' . $wpdb->esc_like($source_url) . '%' | |
| 2155 | + )); | |
| 2156 | + foreach ((array) $candidates as $candidate) { | |
| 2157 | + $signature = maybe_unserialize($candidate->meta_value); | |
| 2158 | + if (is_array($signature) && isset($signature['url']) && $signature['url'] === $source_url) { | |
| 2159 | + $attachment_ids[] = (int) $candidate->post_id; | |
| 2160 | + } | |
| 2161 | + } | |
| 2162 | + | |
| 2163 | + foreach (array_unique($attachment_ids) as $attachment_id) { | |
| 2164 | + delete_post_meta($attachment_id, '_mxchat_kb_attachment_sig'); | |
| 2165 | + } | |
| 2166 | +} | |
| 2167 | + | |
| 2168 | +/** | |
| 2169 | + * The same, for callers that hold a Pinecone vector id instead of a URL. | |
| 2170 | + * Only a base id (md5 of the source URL) names a whole entry; a single | |
| 2171 | + * chunk id leaves the rest of the entry in place, so it is ignored. | |
| 2172 | + */ | |
| 2173 | +public static function forget_attachment_signature_by_key($entry_key, $bot_id = 'default') { | |
| 2174 | + global $wpdb; | |
| 2175 | + | |
| 2176 | + $entry_key = (string) $entry_key; | |
| 2177 | + if (!preg_match('/^[a-f0-9]{32}$/', $entry_key) || !self::attachment_signature_follows_bot($bot_id)) { | |
| 2178 | + return; | |
| 2179 | + } | |
| 2180 | + | |
| 2181 | + $signed = $wpdb->get_results( | |
| 2182 | + "SELECT post_id, meta_value FROM {$wpdb->postmeta} | |
| 2183 | + WHERE meta_key = '_mxchat_kb_attachment_sig' LIMIT 5000" | |
| 2184 | + ); | |
| 2185 | + foreach ((array) $signed as $row) { | |
| 2186 | + $signature = maybe_unserialize($row->meta_value); | |
| 2187 | + if (is_array($signature) && isset($signature['url']) && md5((string) $signature['url']) === $entry_key) { | |
| 2188 | + delete_post_meta((int) $row->post_id, '_mxchat_kb_attachment_sig'); | |
| 2189 | + } | |
| 2190 | + } | |
| 2191 | +} | |
| 2192 | + | |
| 2193 | +/** | |
| 2194 | + * Every media entry of a bot was removed at once ("Delete all"). | |
| 2195 | + */ | |
| 2196 | +public static function forget_all_attachment_signatures($bot_id = 'default') { | |
| 2197 | + if ($bot_id !== 'default' && $bot_id !== 'testing' && !self::is_pinecone_enabled_for_bot($bot_id)) { | |
| 2198 | + // WordPress-database "Delete all" for another bot removes that bot's | |
| 2199 | + // rows only; the default bot's entries are still there. | |
| 2200 | + return; | |
| 2201 | + } | |
| 2202 | + if (self::attachment_signature_follows_bot($bot_id)) { | |
| 2203 | + delete_post_meta_by_key('_mxchat_kb_attachment_sig'); | |
| 2204 | + } | |
| 2205 | +} | |
| 2206 | + | |
| 2207 | +/** | |
| 2208 | + * The signature vouches for the DEFAULT bot's entry (media auto-sync only | |
| 2209 | + * indexes for the default bot). Removing another bot's copy from a Pinecone | |
| 2210 | + * store of its own leaves that entry, and so the signature, in place. | |
| 2211 | + */ | |
| 2212 | +private static function attachment_signature_follows_bot($bot_id) { | |
| 2213 | + if ($bot_id === 'default' || $bot_id === 'testing' || !self::is_pinecone_enabled_for_bot($bot_id)) { | |
| 2214 | + return true; | |
| 2215 | + } | |
| 2216 | + if (!class_exists('MxChat_Pinecone_Documents') || !self::is_pinecone_enabled_for_bot('default')) { | |
| 2217 | + return false; | |
| 2218 | + } | |
| 2219 | + $theirs = MxChat_Pinecone_Documents::config($bot_id); | |
| 2220 | + $ours = MxChat_Pinecone_Documents::config('default'); | |
| 2221 | + return strtolower((string) $theirs['host']) === strtolower((string) $ours['host']) | |
| 2222 | + && (string) $theirs['namespace'] === (string) $ours['namespace']; | |
| 1718 | 2223 | } |
| 1719 | 2224 | |
| 1720 | 2225 | /** |
| 1721 | 2226 | * Delete all chunks for a URL from Pinecone |
