| @@ -5,8 +5,17 @@ | ||
| 5 | 5 | |
| 6 | 6 | class MxChat_Utils { |
| 7 | 7 | |
| 8 | 8 | /** |
| 9 | + * True while a storage routine is re-writing an entry it is about to re-add | |
| 10 | + * (submit_chunked_content's clean-slate delete). delete_chunks_for_url() skips | |
| 11 | + * the Vector Store mirror-delete while set — an internal re-store is not an | |
| 12 | + * entry removal, and mirroring it would delete-then-reupload the entry's file | |
| 13 | + * on every chunked save (plan 15b5c6). | |
| 14 | + */ | |
| 15 | +private static $vectorstore_mirror_suspended = false; | |
| 16 | + | |
| 17 | +/** | |
| 9 | 18 | * Validate a client-supplied session id (plan-mxchat-20260731-d42bec). |
| 10 | 19 | * |
| 11 | 20 | * sanitize_text_field() — which every session_id read site used before this — |
| 12 | 21 | * preserves '/' and '..'. Harmless where the value is only an option or |
| @@ -38,8 +47,336 @@ | ||
| 38 | 47 | return preg_match('/\A[A-Za-z0-9_-]{1,128}\z/', $val) ? $val : ''; |
| 39 | 48 | } |
| 40 | 49 | |
| 41 | 50 | /** |
| 51 | + * Sanitize the lead-capture consent-checkbox label (plan b062c4). | |
| 52 | + * | |
| 53 | + * The label is owner-supplied and renders inside the widget's email form, so | |
| 54 | + * this is a security boundary, not a formatting nicety. One explicit | |
| 55 | + * allowlist, used at BOTH save time (options.php sanitize + autosave AJAX) | |
| 56 | + * and render time (widget form, admin surfaces) so the two can never drift: | |
| 57 | + * an anchor — the whole point is "I agree to the <a>Privacy Policy</a>" — | |
| 58 | + * plus inline emphasis. No block tags, no images, no style attributes. | |
| 59 | + * | |
| 60 | + * The stored consent record keeps this exact sanitized string as "the text | |
| 61 | + * the visitor saw", so it must be deterministic: same input, same output, | |
| 62 | + * whichever path ran it. | |
| 63 | + * | |
| 64 | + * @param mixed $raw Owner-entered label. | |
| 65 | + * @return string Sanitized label, capped at 1000 chars. | |
| 66 | + */ | |
| 67 | +public static function sanitize_consent_label($raw) { | |
| 68 | + if (!is_scalar($raw)) { | |
| 69 | + return ''; | |
| 70 | + } | |
| 71 | + | |
| 72 | + $allowed = array( | |
| 73 | + 'a' => array( | |
| 74 | + 'href' => true, | |
| 75 | + 'title' => true, | |
| 76 | + 'target' => true, | |
| 77 | + 'rel' => true, | |
| 78 | + ), | |
| 79 | + 'strong' => array(), | |
| 80 | + 'em' => array(), | |
| 81 | + 'br' => array(), | |
| 82 | + ); | |
| 83 | + | |
| 84 | + $label = wp_kses(trim((string) $raw), $allowed); | |
| 85 | + | |
| 86 | + return mb_substr($label, 0, 1000); | |
| 87 | +} | |
| 88 | + | |
| 89 | +/** | |
| 90 | + * Per-request cache for get_session_history(). Mirrors get_option()'s | |
| 91 | + * request-scoped caching, which the mxchat_history_ option reads got for | |
| 92 | + * free before plan 839c4c moved history reads onto the transcripts table. | |
| 93 | + */ | |
| 94 | +private static $history_cache = array(); | |
| 95 | + | |
| 96 | +/** | |
| 97 | + * Session chat history read from the transcripts table, in the exact array | |
| 98 | + * shape the legacy mxchat_history_<sid> option stored (plan 839c4c). The | |
| 99 | + * option was a second copy of state the table already held — measured | |
| 100 | + * byte-identical in role/content/order on 174 of 177 real sessions, with | |
| 101 | + * the table a superset on the rest — at up to 64 KB per option row. The | |
| 102 | + * table is now the single store; nothing writes the option any more. | |
| 103 | + * | |
| 104 | + * Shape notes, load-bearing for the consumers: | |
| 105 | + * - id: the transcripts row id (int). Integer ids make the pollers' | |
| 106 | + * ">" comparisons correct where the old uniqid() strings only worked by | |
| 107 | + * accident of hex ordering. | |
| 108 | + * - timestamp: milliseconds, derived from the table's second-resolution GMT | |
| 109 | + * column (x1000). Consumers comparing against a real-millisecond client | |
| 110 | + * cutoff MUST floor the cutoff to the second and err inclusive — see the | |
| 111 | + * persistence-off filters in class-mxchat-integrator.php. | |
| 112 | + * - agent_name: the row's user_identifier, which the writer sets to the | |
| 113 | + * same displayed_name value the option carried (agent name when present, | |
| 114 | + * else email, else identifier). | |
| 115 | + * | |
| 116 | + * Public and static so mxchat-woo / mxchat-forms can call the same accessor | |
| 117 | + * as core, guarded with method_exists against an older mxchat-basic. | |
| 118 | + * | |
| 119 | + * @param string $session_id | |
| 120 | + * @return array[] Chronological entries: id, role, content, timestamp, agent_name. | |
| 121 | + */ | |
| 122 | +public static function get_session_history($session_id) { | |
| 123 | + global $wpdb; | |
| 124 | + | |
| 125 | + $session_id = self::sanitize_session_id($session_id); | |
| 126 | + if ($session_id === '') { | |
| 127 | + return array(); | |
| 128 | + } | |
| 129 | + | |
| 130 | + if (array_key_exists($session_id, self::$history_cache)) { | |
| 131 | + return self::$history_cache[$session_id]; | |
| 132 | + } | |
| 133 | + | |
| 134 | + $table = $wpdb->prefix . 'mxchat_chat_transcripts'; | |
| 135 | + | |
| 136 | + // No SHOW TABLES guard: this is the chat hot path and the table is | |
| 137 | + // created on activation (with an admin-load safety net). A genuinely | |
| 138 | + // missing table fails the query and yields the same empty history the | |
| 139 | + // old option read produced on a fresh session. | |
| 140 | + $rows = $wpdb->get_results( | |
| 141 | + $wpdb->prepare( | |
| 142 | + "SELECT id, role, message, user_identifier, timestamp | |
| 143 | + FROM `$table` WHERE session_id = %s ORDER BY id ASC", | |
| 144 | + $session_id | |
| 145 | + ), | |
| 146 | + ARRAY_A | |
| 147 | + ); | |
| 148 | + | |
| 149 | + $history = array(); | |
| 150 | + if (is_array($rows)) { | |
| 151 | + foreach ($rows as $row) { | |
| 152 | + // The column stores GMT (current_time('mysql', 1) at the writer), | |
| 153 | + // so pin the parse to UTC rather than the site timezone. | |
| 154 | + $ts = strtotime($row['timestamp'] . ' +0000'); | |
| 155 | + $history[] = array( | |
| 156 | + 'id' => (int) $row['id'], | |
| 157 | + 'role' => (string) $row['role'], | |
| 158 | + 'content' => (string) $row['message'], | |
| 159 | + 'timestamp' => ($ts ? $ts : 0) * 1000, | |
| 160 | + 'agent_name' => (string) $row['user_identifier'], | |
| 161 | + ); | |
| 162 | + } | |
| 163 | + } | |
| 164 | + | |
| 165 | + self::$history_cache[$session_id] = $history; | |
| 166 | + | |
| 167 | + return $history; | |
| 168 | +} | |
| 169 | + | |
| 170 | +/** | |
| 171 | + * Drop the cached history for one session (or all). The writer calls this | |
| 172 | + * after every insert so a later read in the same request — e.g. the AI | |
| 173 | + * context build that follows saving the user's message — sees the new row, | |
| 174 | + * matching the read-your-own-write behavior update_option() gave the old | |
| 175 | + * option copy. | |
| 176 | + * | |
| 177 | + * @param string|null $session_id Null flushes everything (test seam). | |
| 178 | + */ | |
| 179 | +public static function flush_session_history_cache($session_id = null) { | |
| 180 | + if ($session_id === null) { | |
| 181 | + self::$history_cache = array(); | |
| 182 | + return; | |
| 183 | + } | |
| 184 | + | |
| 185 | + unset(self::$history_cache[(string) $session_id]); | |
| 186 | +} | |
| 187 | + | |
| 188 | +/** | |
| 189 | + * Append one message to a session's transcript on behalf of an add-on | |
| 190 | + * (plan 4a664e). The Forms add-on's wizard steps travel over the add-on's | |
| 191 | + * own AJAX action, never through the chat request, so nothing wrote them | |
| 192 | + * to the transcript: Transcripts, the live-agent handoff card, Download | |
| 193 | + * Transcript and every transcript email showed "[WIZARD_COMPLETE]" and no | |
| 194 | + * answers. This is the write half of get_session_history() — same row | |
| 195 | + * shape as the chat writer (visitor identity, captured name/email and | |
| 196 | + * originating page from the session store, GMT timestamp), without the | |
| 197 | + * new-session side effects (ownership, notification email, delayed | |
| 198 | + * transcript) that only the visitor's first message should trigger. | |
| 199 | + * | |
| 200 | + * Public and static so add-ons can call it, guarded with method_exists | |
| 201 | + * against an older mxchat-basic. | |
| 202 | + * | |
| 203 | + * @param string $session_id Chat session id. | |
| 204 | + * @param string $role user|bot|system|agent. | |
| 205 | + * @param string $message Stored as-is: bot rows may carry HTML, user rows | |
| 206 | + * are rendered as text by every reader. | |
| 207 | + * @return int The new transcripts row id, 0 when nothing was written. | |
| 208 | + */ | |
| 209 | +public static function save_session_message($session_id, $role, $message) { | |
| 210 | + global $wpdb; | |
| 211 | + | |
| 212 | + $session_id = self::sanitize_session_id($session_id); | |
| 213 | + $role = is_string($role) ? strtolower(trim($role)) : ''; | |
| 214 | + $message = is_scalar($message) ? trim((string) $message) : ''; | |
| 215 | + | |
| 216 | + if ($session_id === '' || $message === '' | |
| 217 | + || !in_array($role, array('user', 'bot', 'system', 'agent'), true)) { | |
| 218 | + return 0; | |
| 219 | + } | |
| 220 | + | |
| 221 | + $table = $wpdb->prefix . 'mxchat_chat_transcripts'; | |
| 222 | + | |
| 223 | + $saved_email = MxChat_Session_Store::get($session_id, 'email'); | |
| 224 | + $saved_name = MxChat_Session_Store::get($session_id, 'name'); | |
| 225 | + | |
| 226 | + $row = array( | |
| 227 | + 'user_id' => is_user_logged_in() ? get_current_user_id() : 0, | |
| 228 | + 'user_identifier' => MxChat_User::mxchat_get_user_identifier(), | |
| 229 | + 'user_email' => $saved_email ?: MxChat_User::mxchat_get_user_email(), | |
| 230 | + 'user_name' => $saved_name ?: '', | |
| 231 | + 'session_id' => $session_id, | |
| 232 | + 'role' => $role, | |
| 233 | + 'message' => $message, | |
| 234 | + 'timestamp' => current_time('mysql', 1), | |
| 235 | + ); | |
| 236 | + | |
| 237 | + // Same page attribution as the rest of the session; the column pair was | |
| 238 | + // added by an upgrade routine, so tolerate its absence like the chat | |
| 239 | + // writer does. | |
| 240 | + $originating = MxChat_Session_Store::get($session_id, 'originating_page'); | |
| 241 | + if (is_array($originating) && !empty($originating['url']) | |
| 242 | + && $wpdb->get_var("SHOW COLUMNS FROM `$table` LIKE 'originating_page_url'")) { | |
| 243 | + $row['originating_page_url'] = $originating['url']; | |
| 244 | + $row['originating_page_title'] = isset($originating['title']) ? $originating['title'] : ''; | |
| 245 | + } | |
| 246 | + | |
| 247 | + $inserted = $wpdb->insert($table, $row); | |
| 248 | + if ($inserted === false) { | |
| 249 | + return 0; | |
| 250 | + } | |
| 251 | + | |
| 252 | + $row_id = (int) $wpdb->insert_id; | |
| 253 | + self::flush_session_history_cache($session_id); | |
| 254 | + | |
| 255 | + return $row_id; | |
| 256 | +} | |
| 257 | + | |
| 258 | +/** | |
| 259 | + * Most recipients the Notification Email field will accept (plan 2f131a). | |
| 260 | + * A settings field is not a mailing list. | |
| 261 | + */ | |
| 262 | +const NOTIFICATION_EMAIL_MAX = 5; | |
| 263 | + | |
| 264 | +/** | |
| 265 | + * Parse the Notification Email field into a list of recipients (plan 2f131a). | |
| 266 | + * | |
| 267 | + * THE TRAP THIS EXISTS TO CLOSE: sanitize_email() cannot be the validator for | |
| 268 | + * this field, because its output for the failing input is VALID. WordPress | |
| 269 | + * strips the separator and the surplus '@' and concatenates the remains: | |
| 270 | + * | |
| 271 | + * [email protected], [email protected] -> [email protected] | |
| 272 | + * | |
| 273 | + * and is_email() then returns true on that. So every guard in the plugin passed, | |
| 274 | + * the address was stored, the autosave ticked green, and both the new-session | |
| 275 | + * notification and the auto-emailed transcript went to a domain that does not | |
| 276 | + * exist — with no error anywhere. Validating the RAW part BEFORE sanitizing is | |
| 277 | + * the whole point; reversing those two lines silently restores the bug. | |
| 278 | + * | |
| 279 | + * All-or-nothing by design: if any entry is bad the caller must store NOTHING. | |
| 280 | + * A partial accept — keeping the good addresses and dropping the bad one — is | |
| 281 | + * the same defect in a new costume, because the owner still believes everyone | |
| 282 | + * on their list is being notified. | |
| 283 | + * | |
| 284 | + * @param mixed $raw Raw field value, exactly as submitted. | |
| 285 | + * @return array{emails: string[], error: string} Empty emails + empty error | |
| 286 | + * means the field was empty. | |
| 287 | + */ | |
| 288 | +public static function parse_notification_emails($raw) { | |
| 289 | + $out = array('emails' => array(), 'error' => ''); | |
| 290 | + | |
| 291 | + if (!is_scalar($raw)) { | |
| 292 | + $out['error'] = __('The notification email could not be read.', 'mxchat'); | |
| 293 | + return $out; | |
| 294 | + } | |
| 295 | + | |
| 296 | + $raw = trim((string) $raw); | |
| 297 | + if ($raw === '') { | |
| 298 | + return $out; // genuinely empty — the caller falls back to admin_email | |
| 299 | + } | |
| 300 | + | |
| 301 | + $seen = array(); | |
| 302 | + foreach (preg_split('/[,;]/', $raw) as $part) { | |
| 303 | + $part = trim($part); | |
| 304 | + if ($part === '') { | |
| 305 | + // A trailing or doubled separator carries no address, so skipping it | |
| 306 | + // cannot silently drop a recipient. This is the ONLY thing tolerated. | |
| 307 | + continue; | |
| 308 | + } | |
| 309 | + | |
| 310 | + // RAW first. See the note above — order is load-bearing. | |
| 311 | + $clean = is_email($part) ? sanitize_email($part) : ''; | |
| 312 | + if ($clean === '' || !is_email($clean)) { | |
| 313 | + return array( | |
| 314 | + 'emails' => array(), | |
| 315 | + 'error' => sprintf( | |
| 316 | + /* translators: %s: the email address the owner typed. */ | |
| 317 | + __('"%s" is not a valid email address, so nothing was saved. Separate multiple addresses with a comma.', 'mxchat'), | |
| 318 | + esc_html($part) | |
| 319 | + ), | |
| 320 | + ); | |
| 321 | + } | |
| 322 | + | |
| 323 | + $key = strtolower($clean); | |
| 324 | + if (isset($seen[$key])) { | |
| 325 | + continue; // same address twice would simply mail them twice | |
| 326 | + } | |
| 327 | + $seen[$key] = true; | |
| 328 | + $out['emails'][] = $clean; | |
| 329 | + } | |
| 330 | + | |
| 331 | + if (count($out['emails']) > self::NOTIFICATION_EMAIL_MAX) { | |
| 332 | + return array( | |
| 333 | + 'emails' => array(), | |
| 334 | + 'error' => sprintf( | |
| 335 | + /* translators: %d: maximum number of notification recipients. */ | |
| 336 | + __('Enter at most %d email addresses, separated by commas.', 'mxchat'), | |
| 337 | + self::NOTIFICATION_EMAIL_MAX | |
| 338 | + ), | |
| 339 | + ); | |
| 340 | + } | |
| 341 | + | |
| 342 | + return $out; | |
| 343 | +} | |
| 344 | + | |
| 345 | +/** | |
| 346 | + * The stored recipient list, ready to hand to wp_mail() (plan 2f131a). | |
| 347 | + * | |
| 348 | + * Fallback rule, and it is narrow on purpose: an EMPTY field falls back to the | |
| 349 | + * site admin address, because that is the documented behaviour and an owner who | |
| 350 | + * never filled the field in still wants their notifications. A field holding | |
| 351 | + * something unusable does NOT fall back — it sends nowhere, exactly as before | |
| 352 | + * this plan. Falling back on bad input would mean a typo silently redirects a | |
| 353 | + * store's transcripts to a different mailbox than the one on screen. | |
| 354 | + * | |
| 355 | + * @param array|null $options mxchat_transcripts_options, or null to read it. | |
| 356 | + * @return string[] Recipients; empty means do not send. | |
| 357 | + */ | |
| 358 | +public static function notification_recipients($options = null) { | |
| 359 | + if (!is_array($options)) { | |
| 360 | + $options = get_option('mxchat_transcripts_options', array()); | |
| 361 | + if (!is_array($options)) { | |
| 362 | + $options = array(); | |
| 363 | + } | |
| 364 | + } | |
| 365 | + | |
| 366 | + $raw = isset($options['mxchat_notification_email']) ? $options['mxchat_notification_email'] : ''; | |
| 367 | + $raw = is_scalar($raw) ? trim((string) $raw) : ''; | |
| 368 | + | |
| 369 | + if ($raw === '') { | |
| 370 | + $admin = get_option('admin_email'); | |
| 371 | + return is_email($admin) ? array($admin) : array(); | |
| 372 | + } | |
| 373 | + | |
| 374 | + $parsed = self::parse_notification_emails($raw); | |
| 375 | + return $parsed['error'] === '' ? $parsed['emails'] : array(); | |
| 376 | +} | |
| 377 | + | |
| 378 | +/** | |
| 42 | 379 | * Centralized embedding model registry. Single source of truth for dimensions |
| 43 | 380 | * and provider, so model-switch protection logic doesn't drift across files. |
| 44 | 381 | */ |
| 45 | 382 | public static function embedding_model_registry() { |
| @@ -150,8 +487,45 @@ | ||
| 150 | 487 | return (strlen($id) === 11) ? $id : ''; |
| 151 | 488 | } |
| 152 | 489 | |
| 153 | 490 | /** |
| 491 | + * plan-mxchat-20260813-f52492 — video-card gating. | |
| 492 | + * | |
| 493 | + * Two standalone options (NOT mxchat_options — they skip the sanitize/autosave | |
| 494 | + * traps entirely), read here so the gate in the integrator and the fields on | |
| 495 | + * Knowledge -> Chunking & Retrieval can never disagree about a default. | |
| 496 | + * | |
| 497 | + * Master switch. Default ON: the card is existing behavior, and this is an | |
| 498 | + * opt-out for owners who never want one, not a new feature to opt into. | |
| 499 | + */ | |
| 500 | +public static function video_embed_enabled() { | |
| 501 | + return get_option('mxchat_video_embed_enabled', 'on') === 'on'; | |
| 502 | +} | |
| 503 | + | |
| 504 | +/** | |
| 505 | + * The video card's OWN confidence floor, as a 0-1 cosine — deliberately not | |
| 506 | + * the site-wide Similarity Threshold (default 35). "Good enough to quote in | |
| 507 | + * the answer" and "good enough to put a video on screen" are different | |
| 508 | + * questions: retrieval is allowed to be generous because the model still | |
| 509 | + * decides what to say, whereas the card is asserted to the visitor with no | |
| 510 | + * such filter. Stored as an int percentage to match the site-wide slider's | |
| 511 | + * convention; the default (55) sits above it on purpose. | |
| 512 | + * | |
| 513 | + * MXCHAT_VIDEO_EMBED_THRESHOLD_DEFAULT is the single source of that number. | |
| 514 | + */ | |
| 515 | +public static function video_embed_threshold() { | |
| 516 | + $stored = get_option('mxchat_video_embed_threshold', null); | |
| 517 | + $percent = ($stored === null || $stored === '') | |
| 518 | + ? MXCHAT_VIDEO_EMBED_THRESHOLD_DEFAULT | |
| 519 | + : (int) $stored; | |
| 520 | + if ($percent < 0) { $percent = 0; } | |
| 521 | + if ($percent > 100) { $percent = 100; } | |
| 522 | + // Cast: PHP evaluates 100/100 to int(1), so an unclamped return type would | |
| 523 | + // vary with the stored value. Callers compare against a cosine — keep it float. | |
| 524 | + return (float) $percent / 100; | |
| 525 | +} | |
| 526 | + | |
| 527 | +/** | |
| 154 | 528 | * UPDATED: Submit or update content (and its embedding) in the database. |
| 155 | 529 | * Stores in Pinecone if enabled, otherwise stores in WordPress DB. |
| 156 | 530 | * |
| 157 | 531 | * @param string $content The content to be embedded. |
| @@ -161,9 +535,9 @@ | ||
| 161 | 535 | * @param string $bot_id The bot ID for multi-bot support |
| 162 | 536 | * @param string $content_type The type of content (post, page, pdf, url, manual, product, etc.) |
| 163 | 537 | * @return bool|WP_Error True on success, WP_Error on failure |
| 164 | 538 | */ |
| 165 | -public static function submit_content_to_db($content, $source_url, $api_key, $vector_id = null, $bot_id = 'default', $content_type = 'content') { | |
| 539 | +public static function submit_content_to_db($content, $source_url, $api_key, $vector_id = null, $bot_id = 'default', $content_type = 'content', $title = '') { | |
| 166 | 540 | global $wpdb; |
| 167 | 541 | $table_name = $wpdb->prefix . 'mxchat_system_prompt_content'; |
| 168 | 542 | |
| 169 | 543 | //error_log('[MXCHAT-DB] Starting database submission for URL: ' . $source_url . ' (Bot: ' . $bot_id . ', Type: ' . $content_type . ')'); |
| @@ -168,10 +542,18 @@ | ||
| 168 | 542 | |
| 169 | 543 | //error_log('[MXCHAT-DB] Starting database submission for URL: ' . $source_url . ' (Bot: ' . $bot_id . ', Type: ' . $content_type . ')'); |
| 170 | 544 | //error_log('[MXCHAT-DB] Content length: ' . strlen($content) . ' bytes'); |
| 171 | 545 | |
| 172 | - // Sanitize the source URL | |
| 173 | - $source_url = esc_url_raw($source_url); | |
| 546 | + // Sanitize the source URL. Internal identities (mxchat:// manual docs, | |
| 547 | + // upload:// file uploads) are NOT web URLs — esc_url_raw EMPTIES them | |
| 548 | + // because its protocol list is statically cached and effectively | |
| 549 | + // unfilterable (measured, plan 945406 / 0485e5) — sanitize those as text | |
| 550 | + // so upserts stay keyed to a stable identity across re-imports. | |
| 551 | + if (preg_match('#^(mxchat|upload)://#i', $source_url)) { | |
| 552 | + $source_url = sanitize_text_field($source_url); | |
| 553 | + } else { | |
| 554 | + $source_url = esc_url_raw($source_url); | |
| 555 | + } | |
| 174 | 556 | |
| 175 | 557 | // Sanitize content_type |
| 176 | 558 | $content_type = sanitize_key($content_type); |
| 177 | 559 | if (empty($content_type)) { |
| @@ -177,8 +559,14 @@ | ||
| 177 | 559 | if (empty($content_type)) { |
| 178 | 560 | $content_type = 'content'; // Fallback for backwards compatibility |
| 179 | 561 | } |
| 180 | 562 | |
| 563 | + // Page title captured by the importer (plan d9ee66), registered under the | |
| 564 | + // SANITIZED url so the storage writers below find it by the same key. | |
| 565 | + if (is_string($title) && trim($title) !== '') { | |
| 566 | + self::set_source_title($source_url, $title); | |
| 567 | + } | |
| 568 | + | |
| 181 | 569 | // Just ensure UTF-8 validity without aggressive escaping |
| 182 | 570 | $safe_content = wp_check_invalid_utf8($content); |
| 183 | 571 | // Remove only null bytes and other control characters, but preserve newlines (\n = \x0A) and carriage returns (\r = \x0D) |
| 184 | 572 | $safe_content = preg_replace('/[\x00-\x08\x0B\x0C\x0E-\x1F\x7F]/u', '', $safe_content); |
| @@ -186,9 +574,16 @@ | ||
| 186 | 574 | // Check if chunking should be applied |
| 187 | 575 | $chunker = MxChat_Chunker::from_settings(); |
| 188 | 576 | if ($chunker->should_chunk($safe_content)) { |
| 189 | 577 | //error_log('[MXCHAT-DB] Content exceeds chunk threshold, using chunked submission'); |
| 190 | - return self::submit_chunked_content($safe_content, $source_url, $api_key, $bot_id, $content_type, $chunker); | |
| 578 | + $chunk_result = self::submit_chunked_content($safe_content, $source_url, $api_key, $bot_id, $content_type, $chunker); | |
| 579 | + // Vector Store mirror gets the entry WHOLE — one file per KB entry, | |
| 580 | + // OpenAI chunks server-side; local chunking is our own embedding | |
| 581 | + // concern and never reaches the store (plan 15b5c6). | |
| 582 | + if ($chunk_result === true && class_exists('MxChat_Vectorstore_Manager')) { | |
| 583 | + MxChat_Vectorstore_Manager::sync_upsert_entry($source_url, $safe_content, $bot_id, $content_type); | |
| 584 | + } | |
| 585 | + return $chunk_result; | |
| 191 | 586 | } |
| 192 | 587 | |
| 193 | 588 | // UPDATED: Generate the embedding using bot-specific configuration |
| 194 | 589 | $embedding_vector = self::generate_embedding($content, $api_key, $bot_id); |
| @@ -206,18 +601,175 @@ | ||
| 206 | 601 | // UPDATED: Check if Pinecone is enabled for this specific bot |
| 207 | 602 | if (self::is_pinecone_enabled_for_bot($bot_id)) { |
| 208 | 603 | //error_log('[MXCHAT-DB] Pinecone is enabled for bot ' . $bot_id . ' - using Pinecone storage'); |
| 209 | 604 | // Store in Pinecone only |
| 210 | - return self::store_in_pinecone_only($embedding_vector, $content, $source_url, $vector_id, $bot_id, $content_type); | |
| 605 | + $result = self::store_in_pinecone_only($embedding_vector, $content, $source_url, $vector_id, $bot_id, $content_type); | |
| 606 | + // If this URL previously stored as CHUNKED content and the new content fits in a | |
| 607 | + // single vector, the upsert above only overwrote the base id — the old | |
| 608 | + // md5(url)_chunk_N vectors would keep serving the stale text. Sweep them. | |
| 609 | + // Only for a real source_url: chunk ids derive from it, and md5('') is shared | |
| 610 | + // by legacy URL-less entries so a blind sweep there could hit other entries. | |
| 611 | + if ($result === true && !empty($source_url)) { | |
| 612 | + self::cleanup_pinecone_chunk_stragglers($source_url, $bot_id); | |
| 613 | + } | |
| 614 | + if ($result === true && class_exists('MxChat_Vectorstore_Manager')) { | |
| 615 | + MxChat_Vectorstore_Manager::sync_upsert_entry($source_url, $safe_content, $bot_id, $content_type); | |
| 616 | + } | |
| 617 | + return $result; | |
| 211 | 618 | } else { |
| 212 | 619 | //error_log('[MXCHAT-DB] Pinecone not enabled for bot ' . $bot_id . ' - using WordPress storage'); |
| 213 | 620 | // Store in WordPress database only |
| 214 | 621 | $embedding_vector_serialized = maybe_serialize($embedding_vector); |
| 215 | - return self::store_in_wordpress_db($safe_content, $source_url, $embedding_vector_serialized, $table_name, $content_type); | |
| 622 | + $result = self::store_in_wordpress_db($safe_content, $source_url, $embedding_vector_serialized, $table_name, $content_type); | |
| 623 | + if ($result === true && class_exists('MxChat_Vectorstore_Manager')) { | |
| 624 | + MxChat_Vectorstore_Manager::sync_upsert_entry($source_url, $safe_content, $bot_id, $content_type); | |
| 625 | + } | |
| 626 | + return $result; | |
| 216 | 627 | } |
| 217 | 628 | } |
| 218 | 629 | |
| 630 | +/* ------------------------------------------------------------------------- | |
| 631 | + * Source titles (3.2.23, plan d9ee66) | |
| 632 | + * | |
| 633 | + * A knowledge entry imported from a URL or sitemap remembers the page's own | |
| 634 | + * title, so an add-on that names a source (AI Search's source cards) can show | |
| 635 | + * "About Us" instead of a guess from the slug ("Aboutus"). Stored in the | |
| 636 | + * `title` column of the WP knowledge table and as `title` metadata on Pinecone | |
| 637 | + * vectors / documents. The importer hands the title to submit_content_to_db(), | |
| 638 | + * which registers it under the source URL; the writers look it up by that key. | |
| 639 | + * ---------------------------------------------------------------------- */ | |
| 640 | + | |
| 641 | +/** @var array<string,string> source_url => title, for the current request */ | |
| 642 | +private static $source_titles = array(); | |
| 643 | + | |
| 644 | +public static function set_source_title($source_url, $title) { | |
| 645 | + $key = trim((string) $source_url); | |
| 646 | + $title = self::clean_title($title); | |
| 647 | + if ($key === '' || $title === '') { | |
| 648 | + return; | |
| 649 | + } | |
| 650 | + self::$source_titles[$key] = $title; | |
| 651 | +} | |
| 652 | + | |
| 653 | +public static function source_title_for($source_url) { | |
| 654 | + $key = trim((string) $source_url); | |
| 655 | + return ($key !== '' && isset(self::$source_titles[$key])) ? self::$source_titles[$key] : ''; | |
| 656 | +} | |
| 657 | + | |
| 658 | +/** Entities decoded, tags gone, whitespace collapsed, capped at the column width. */ | |
| 659 | +public static function clean_title($title) { | |
| 660 | + $title = html_entity_decode((string) $title, ENT_QUOTES | ENT_HTML5, 'UTF-8'); | |
| 661 | + $title = wp_strip_all_tags($title); | |
| 662 | + $title = trim((string) preg_replace('/\s+/u', ' ', $title)); | |
| 663 | + if (mb_strlen($title, 'UTF-8') > 255) { | |
| 664 | + $title = rtrim(mb_substr($title, 0, 255, 'UTF-8')); | |
| 665 | + } | |
| 666 | + return $title; | |
| 667 | +} | |
| 668 | + | |
| 219 | 669 | /** |
| 670 | + * The page's own title from fetched HTML: og:title, else <title> with a | |
| 671 | + * trailing "| Site Name" style suffix removed, else the first <h1>. Returns '' | |
| 672 | + * when nothing usable is there, so the caller keeps its own fallback. | |
| 673 | + */ | |
| 674 | +public static function extract_page_title($html) { | |
| 675 | + $html = (string) $html; | |
| 676 | + if ($html === '') { | |
| 677 | + return ''; | |
| 678 | + } | |
| 679 | + $head = substr($html, 0, 200000); | |
| 680 | + $site = ''; | |
| 681 | + if (preg_match('/<meta\s+[^>]*property=["\']og:site_name["\'][^>]*content=["\']([^"\']{1,120})["\']/i', $head, $m) | |
| 682 | + || preg_match('/<meta\s+[^>]*content=["\']([^"\']{1,120})["\'][^>]*property=["\']og:site_name["\']/i', $head, $m)) { | |
| 683 | + $site = self::clean_title($m[1]); | |
| 684 | + } | |
| 685 | + $candidate = ''; | |
| 686 | + if (preg_match('/<meta\s+[^>]*property=["\']og:title["\'][^>]*content=["\']([^"\']{1,500})["\']/i', $head, $m) | |
| 687 | + || preg_match('/<meta\s+[^>]*content=["\']([^"\']{1,500})["\'][^>]*property=["\']og:title["\']/i', $head, $m)) { | |
| 688 | + $candidate = self::clean_title($m[1]); | |
| 689 | + } | |
| 690 | + if ($candidate === '' && preg_match('/<title[^>]*>(.*?)<\/title>/is', $head, $m)) { | |
| 691 | + $candidate = self::clean_title($m[1]); | |
| 692 | + } | |
| 693 | + if ($candidate !== '') { | |
| 694 | + $candidate = self::strip_site_suffix($candidate, $site); | |
| 695 | + } | |
| 696 | + if ($candidate === '' && preg_match('/<h1[^>]*>(.*?)<\/h1>/is', $html, $m)) { | |
| 697 | + $candidate = self::clean_title($m[1]); | |
| 698 | + } | |
| 699 | + return $candidate; | |
| 700 | +} | |
| 701 | + | |
| 702 | +/** | |
| 703 | + * "About Us | Just Facts" -> "About Us". Only a TRAILING segment after a spaced | |
| 704 | + * separator goes, only when what is left is a real title (3+ characters) and the | |
| 705 | + * removed part is short enough to be a site name; a known og:site_name wins. | |
| 706 | + * Byte offsets throughout: every separator starts with an ASCII space, so a cut | |
| 707 | + * there never lands inside a multibyte character. | |
| 708 | + */ | |
| 709 | +private static function strip_site_suffix($title, $site_name = '') { | |
| 710 | + $seps = array(' | ', ' – ', ' — ', ' - ', ' :: ', ' · ', ' » ', ' • '); | |
| 711 | + if ($site_name !== '') { | |
| 712 | + foreach ($seps as $sep) { | |
| 713 | + $tail = $sep . $site_name; | |
| 714 | + if (strlen($title) > strlen($tail) && substr($title, -strlen($tail)) === $tail) { | |
| 715 | + return trim(substr($title, 0, strlen($title) - strlen($tail))); | |
| 716 | + } | |
| 717 | + } | |
| 718 | + } | |
| 719 | + $best_pos = -1; | |
| 720 | + $best_sep = ''; | |
| 721 | + foreach ($seps as $sep) { | |
| 722 | + $pos = strrpos($title, $sep); | |
| 723 | + if ($pos !== false && $pos > $best_pos) { | |
| 724 | + $best_pos = $pos; | |
| 725 | + $best_sep = $sep; | |
| 726 | + } | |
| 727 | + } | |
| 728 | + if ($best_pos > 0) { | |
| 729 | + $left = trim(substr($title, 0, $best_pos)); | |
| 730 | + $right = trim(substr($title, $best_pos + strlen($best_sep))); | |
| 731 | + if (mb_strlen($left, 'UTF-8') >= 3 && $right !== '' && mb_strlen($right, 'UTF-8') <= 60) { | |
| 732 | + return $left; | |
| 733 | + } | |
| 734 | + } | |
| 735 | + return $title; | |
| 736 | +} | |
| 737 | + | |
| 738 | +/** | |
| 739 | + * Make sure the knowledge table has the `title` column (3.2.23, plan d9ee66). | |
| 740 | + * Activation adds it; this is the safety net for a copy updated in place, and | |
| 741 | + * the one place add-ons ask "may I read/write title?". A positive answer is | |
| 742 | + * remembered in an option, so after the first time this costs one option read. | |
| 743 | + */ | |
| 744 | +public static function ensure_kb_title_column() { | |
| 745 | + global $wpdb; | |
| 746 | + static $ready = null; | |
| 747 | + if ($ready !== null) { | |
| 748 | + return $ready; | |
| 749 | + } | |
| 750 | + if (get_option('mxchat_kb_title_column') === '1') { | |
| 751 | + $ready = true; | |
| 752 | + return true; | |
| 753 | + } | |
| 754 | + $table = $wpdb->prefix . 'mxchat_system_prompt_content'; | |
| 755 | + $cols = $wpdb->get_col("SHOW COLUMNS FROM {$table}", 0); | |
| 756 | + if (empty($cols)) { | |
| 757 | + $ready = false; | |
| 758 | + return false; | |
| 759 | + } | |
| 760 | + if (!in_array('title', $cols, true)) { | |
| 761 | + $wpdb->query("ALTER TABLE {$table} ADD COLUMN title VARCHAR(255) DEFAULT NULL AFTER source_url"); | |
| 762 | + $cols = $wpdb->get_col("SHOW COLUMNS FROM {$table}", 0); | |
| 763 | + } | |
| 764 | + $ready = in_array('title', $cols, true); | |
| 765 | + if ($ready) { | |
| 766 | + update_option('mxchat_kb_title_column', '1', false); | |
| 767 | + } | |
| 768 | + return $ready; | |
| 769 | +} | |
| 770 | + | |
| 771 | +/** | |
| 220 | 772 | * UPDATED: Check if Pinecone is enabled and properly configured for a specific bot |
| 221 | 773 | */ |
| 222 | 774 | private static function is_pinecone_enabled_for_bot($bot_id = 'default') { |
| 223 | 775 | // For default bot or when multi-bot is not active, use original method |
| @@ -329,11 +881,15 @@ | ||
| 329 | 881 | // Check if this is truly manual content (no URL at all) vs a real URL that filter_var rejects |
| 330 | 882 | // filter_var(FILTER_VALIDATE_URL) rejects valid URLs with encoded chars, non-ASCII, fragments, etc. |
| 331 | 883 | // Use a looser check: if it starts with http(s):// or has a scheme, it's a URL |
| 332 | 884 | $has_url_scheme = !empty($source_url) && preg_match('#^https?://#i', $source_url); |
| 885 | + // upload:// identities (admin document/PDF uploads, plan 0485e5) are stable | |
| 886 | + // and deduplicable — treat them like URLs so a re-upload UPDATES the row | |
| 887 | + // instead of minting a fresh manual identity (which would duplicate). | |
| 888 | + $has_stable_identity = $has_url_scheme || (!empty($source_url) && preg_match('#^upload://#i', $source_url)); | |
| 333 | 889 | // Treat legacy mxchat.ai source URLs as manual — old bug assigned the site URL to manual entries |
| 334 | 890 | $is_legacy_mxchat_url = $has_url_scheme && strpos($source_url, 'mxchat.ai') !== false; |
| 335 | - $is_manual_content = empty($source_url) || $source_url === '' || !$has_url_scheme || $is_legacy_mxchat_url; | |
| 891 | + $is_manual_content = empty($source_url) || $source_url === '' || !$has_stable_identity || $is_legacy_mxchat_url; | |
| 336 | 892 | |
| 337 | 893 | if ($is_manual_content) { |
| 338 | 894 | // Generate unique identifier for manual content to prevent overwrites |
| 339 | 895 | $source_url = 'mxchat://manual-content/' . time() . '-' . wp_generate_password(8, false); |
| @@ -354,8 +910,13 @@ | ||
| 354 | 910 | //error_log('[MXCHAT-DB] Manual content - will create new entry (no duplicate check)'); |
| 355 | 911 | } |
| 356 | 912 | // ===== END FIX ===== |
| 357 | 913 | |
| 914 | + // Page title captured at import (plan d9ee66): written only when the | |
| 915 | + // importer supplied one, so a re-index without a title keeps the old value. | |
| 916 | + $title_value = $is_manual_content ? '' : self::source_title_for($source_url); | |
| 917 | + $title_ready = ($title_value !== '') && self::ensure_kb_title_column(); | |
| 918 | + | |
| 358 | 919 | // Progressive fallback mechanism for problematic content |
| 359 | 920 | $attempt = 1; |
| 360 | 921 | $max_attempts = 3; |
| 361 | 922 | $current_content = $safe_content; |
| @@ -366,39 +927,41 @@ | ||
| 366 | 927 | if ($existing_id) { |
| 367 | 928 | //error_log('[MXCHAT-DB] Found existing entry (ID: ' . $existing_id . '). Updating... (Attempt ' . $attempt . ')'); |
| 368 | 929 | |
| 369 | 930 | // Update the existing row - UPDATED 2.5.6: Added content_type |
| 370 | - $result = $wpdb->update( | |
| 371 | - $table_name, | |
| 372 | - array( | |
| 373 | - 'url' => $source_url, | |
| 374 | - 'article_content' => $current_content, | |
| 375 | - 'embedding_vector' => $embedding_vector_serialized, | |
| 376 | - 'source_url' => $source_url, | |
| 377 | - 'content_type' => $content_type, | |
| 378 | - 'timestamp' => current_time('mysql'), | |
| 379 | - ), | |
| 380 | - array('id' => $existing_id), | |
| 381 | - array('%s','%s','%s','%s','%s','%s'), | |
| 382 | - array('%d') | |
| 931 | + $row = array( | |
| 932 | + 'url' => $source_url, | |
| 933 | + 'article_content' => $current_content, | |
| 934 | + 'embedding_vector' => $embedding_vector_serialized, | |
| 935 | + 'source_url' => $source_url, | |
| 936 | + 'content_type' => $content_type, | |
| 937 | + 'timestamp' => current_time('mysql'), | |
| 383 | 938 | ); |
| 939 | + $formats = array('%s','%s','%s','%s','%s','%s'); | |
| 940 | + if ($title_ready) { | |
| 941 | + $row['title'] = $title_value; | |
| 942 | + $formats[] = '%s'; | |
| 943 | + } | |
| 944 | + $result = $wpdb->update($table_name, $row, array('id' => $existing_id), $formats, array('%d')); | |
| 384 | 945 | } else { |
| 385 | 946 | //error_log('[MXCHAT-DB] No existing entry found. Inserting new row... (Attempt ' . $attempt . ')'); |
| 386 | 947 | //error_log('[MXCHAT-DB] Content sample: ' . substr($current_content, 0, 1000)); |
| 387 | 948 | |
| 388 | 949 | // Insert a new row - UPDATED 2.5.6: Added content_type |
| 389 | - $result = $wpdb->insert( | |
| 390 | - $table_name, | |
| 391 | - array( | |
| 392 | - 'url' => $source_url, // Now unique for manual content | |
| 393 | - 'article_content' => $current_content, | |
| 394 | - 'embedding_vector' => $embedding_vector_serialized, | |
| 395 | - 'source_url' => $source_url, // Now unique for manual content | |
| 396 | - 'content_type' => $content_type, | |
| 397 | - 'timestamp' => current_time('mysql'), | |
| 398 | - ), | |
| 399 | - array('%s','%s','%s','%s','%s','%s') | |
| 950 | + $row = array( | |
| 951 | + 'url' => $source_url, // Now unique for manual content | |
| 952 | + 'article_content' => $current_content, | |
| 953 | + 'embedding_vector' => $embedding_vector_serialized, | |
| 954 | + 'source_url' => $source_url, // Now unique for manual content | |
| 955 | + 'content_type' => $content_type, | |
| 956 | + 'timestamp' => current_time('mysql'), | |
| 400 | 957 | ); |
| 958 | + $formats = array('%s','%s','%s','%s','%s','%s'); | |
| 959 | + if ($title_ready) { | |
| 960 | + $row['title'] = $title_value; | |
| 961 | + $formats[] = '%s'; | |
| 962 | + } | |
| 963 | + $result = $wpdb->insert($table_name, $row, $formats); | |
| 401 | 964 | } |
| 402 | 965 | |
| 403 | 966 | if ($result === false) { |
| 404 | 967 | //error_log('[MXCHAT-DB] Database operation failed (Attempt ' . $attempt . ')'); |
| @@ -457,10 +1020,11 @@ | ||
| 457 | 1020 | // ===== UPDATED: Handle manual content with unique vector IDs ===== |
| 458 | 1021 | if ($vector_id) { |
| 459 | 1022 | // Use provided vector ID |
| 460 | 1023 | //error_log('[MXCHAT-PINECONE-MAIN] Using provided vector ID: ' . $vector_id); |
| 461 | - } elseif (!empty($url) && preg_match('#^https?://#i', $url)) { | |
| 462 | - // For URLs, use URL-based ID (existing behavior) | |
| 1024 | + } elseif (!empty($url) && preg_match('#^(https?|upload)://#i', $url)) { | |
| 1025 | + // For URLs — and stable upload:// identities (plan 0485e5) — use an | |
| 1026 | + // identity-derived ID so a re-import upserts the same vector. | |
| 463 | 1027 | $vector_id = md5($url); |
| 464 | 1028 | //error_log('[MXCHAT-PINECONE-MAIN] Generated vector ID from URL: ' . $vector_id); |
| 465 | 1029 | } else { |
| 466 | 1030 | // For manual content (empty/no URL scheme), generate unique ID |
| @@ -521,9 +1085,23 @@ | ||
| 521 | 1085 | 'last_updated' => time(), |
| 522 | 1086 | 'created_at' => time(), // Add creation timestamp |
| 523 | 1087 | 'bot_id' => $bot_id, // Add bot identification |
| 524 | 1088 | ); |
| 525 | - | |
| 1089 | + $page_title = self::source_title_for($url); | |
| 1090 | + if ($page_title !== '') { | |
| 1091 | + $metadata['title'] = $page_title; // page title captured at import (plan d9ee66) | |
| 1092 | + } | |
| 1093 | + | |
| 1094 | + // Document index (plan 362c31): same id, same metadata, through the | |
| 1095 | + // documents API. A site on the classic vector index never enters here. | |
| 1096 | + if (class_exists('MxChat_Pinecone_Documents') && MxChat_Pinecone_Documents::is_document_index($bot_id)) { | |
| 1097 | + return MxChat_Pinecone_Documents::upsert_one($vector_id, $content, $embedding_vector, $metadata, array( | |
| 1098 | + 'api_key' => $api_key, | |
| 1099 | + 'host' => $host, | |
| 1100 | + 'namespace' => $namespace, | |
| 1101 | + )); | |
| 1102 | + } | |
| 1103 | + | |
| 526 | 1104 | $vector_data = array( |
| 527 | 1105 | 'id' => $vector_id, |
| 528 | 1106 | 'values' => $embedding_vector, |
| 529 | 1107 | 'metadata' => $metadata |
| @@ -657,8 +1235,25 @@ | ||
| 657 | 1235 | return array('ok' => true, 'api_key' => $api_key, 'reason' => '', 'provider' => $provider); |
| 658 | 1236 | } |
| 659 | 1237 | |
| 660 | 1238 | /** |
| 1239 | + * Public QUERY-side entry point (plan 876edb). The chat pipeline's | |
| 1240 | + * MxChat_Integrator::mxchat_generate_embedding() adapter routes through here | |
| 1241 | + * so the query and index sides share ONE provider-routing implementation — | |
| 1242 | + * the same endpoints, request bodies, and stamping semantics. The Integrator | |
| 1243 | + * keeps its own error vocabulary by translating the WP_Error this returns | |
| 1244 | + * (see the structured error data on every failure path below). | |
| 1245 | + * | |
| 1246 | + * @param string $text The text to be embedded. | |
| 1247 | + * @param string $api_key Caller-resolved API key (per-bot on the query side). | |
| 1248 | + * @param string $bot_id The bot ID for multi-bot support. | |
| 1249 | + * @return array|WP_Error The embedding vector, or WP_Error carrying the reason. | |
| 1250 | + */ | |
| 1251 | +public static function generate_query_embedding($text, $api_key, $bot_id = 'default') { | |
| 1252 | + return self::generate_embedding($text, $api_key, $bot_id); | |
| 1253 | +} | |
| 1254 | + | |
| 1255 | +/** | |
| 661 | 1256 | * UPDATED: Generate an embedding for the given text using bot-specific configuration. |
| 662 | 1257 | * |
| 663 | 1258 | * @param string $text The text to be embedded. |
| 664 | 1259 | * @param string $api_key The API key used for generating embeddings. |
| @@ -680,10 +1275,12 @@ | ||
| 680 | 1275 | // below is untouched. |
| 681 | 1276 | if (isset($options['custom_provider_for_embeddings']) && $options['custom_provider_for_embeddings'] === 'on') { |
| 682 | 1277 | $custom = self::generate_embedding_custom($text, $options); |
| 683 | 1278 | // The custom path already returns a human-readable error string — |
| 684 | - // carry it instead of collapsing to null (plan 4a7c0a). | |
| 685 | - return is_array($custom) ? $custom : new WP_Error('embedding_failed', (string) $custom); | |
| 1279 | + // carry it instead of collapsing to null (plan 4a7c0a). The 'custom' | |
| 1280 | + // branch marker lets the Integrator adapter map the string back onto | |
| 1281 | + // its own error codes (876edb). | |
| 1282 | + return is_array($custom) ? $custom : new WP_Error('embedding_failed', (string) $custom, array('branch' => 'custom')); | |
| 686 | 1283 | } |
| 687 | 1284 | |
| 688 | 1285 | $selected_model = $options['embedding_model'] ?? 'text-embedding-ada-002'; |
| 689 | 1286 | |
| @@ -695,10 +1292,13 @@ | ||
| 695 | 1292 | $endpoint = 'https://generativelanguage.googleapis.com/v1beta/models/' . $selected_model . ':embedContent'; |
| 696 | 1293 | $api_key = $options['gemini_api_key'] ?? ''; |
| 697 | 1294 | } else { |
| 698 | 1295 | $endpoint = 'https://api.openai.com/v1/embeddings'; |
| 699 | - // Use the bot-specific API key or fallback to passed API key | |
| 700 | - $api_key = $options['api_key'] ?? $api_key; | |
| 1296 | + // Prefer the caller-resolved key when one was passed — the query side | |
| 1297 | + // resolves per-bot keys at its call sites (integrator adapter, 876edb). | |
| 1298 | + // Index callers pass the preflight key, which equals this options read, | |
| 1299 | + // so nothing changes for them. | |
| 1300 | + $api_key = !empty($api_key) ? $api_key : ($options['api_key'] ?? ''); | |
| 701 | 1301 | } |
| 702 | 1302 | |
| 703 | 1303 | // Prepare request body based on provider |
| 704 | 1304 | if (strpos($selected_model, 'gemini-embedding') === 0) { |
| @@ -753,9 +1353,14 @@ | ||
| 753 | 1353 | $message = 'Embedding request failed (connection): ' . $response->get_error_message(); |
| 754 | 1354 | if (class_exists('MxChat_Admin')) { |
| 755 | 1355 | MxChat_Admin::mxchat_log_debug('embedding_error', $message, array('model' => $selected_model, 'bot_id' => $bot_id)); |
| 756 | 1356 | } |
| 757 | - return new WP_Error('embedding_failed', $message); | |
| 1357 | + return new WP_Error('embedding_failed', $message, array( | |
| 1358 | + 'branch' => 'cloud', | |
| 1359 | + 'kind' => 'connection', | |
| 1360 | + 'reason' => $response->get_error_message(), | |
| 1361 | + 'model' => $selected_model, | |
| 1362 | + )); | |
| 758 | 1363 | } |
| 759 | 1364 | |
| 760 | 1365 | $response_body = json_decode(wp_remote_retrieve_body($response), true); |
| 761 | 1366 | |
| @@ -820,9 +1425,18 @@ | ||
| 820 | 1425 | 'bot_id' => $bot_id, |
| 821 | 1426 | )); |
| 822 | 1427 | } |
| 823 | 1428 | |
| 824 | - return new WP_Error('embedding_failed', $message); | |
| 1429 | + // Structured data so the Integrator's query-side adapter can rebuild its | |
| 1430 | + // typed error contract (auth/rate-limit/quota/invalid-response) without a | |
| 1431 | + // second transport implementation (876edb). Additive — message unchanged. | |
| 1432 | + return new WP_Error('embedding_failed', $message, array( | |
| 1433 | + 'branch' => 'cloud', | |
| 1434 | + 'status' => $status, | |
| 1435 | + 'error_type' => (is_array($decoded) && isset($decoded['error']['type']) && is_string($decoded['error']['type'])) ? $decoded['error']['type'] : '', | |
| 1436 | + 'reason' => $reason, | |
| 1437 | + 'model' => $selected_model, | |
| 1438 | + )); | |
| 825 | 1439 | } |
| 826 | 1440 | |
| 827 | 1441 | /** |
| 828 | 1442 | * Generate an embedding via a Custom (OpenAI-compatible) provider's /embeddings route. |
| @@ -877,9 +1491,13 @@ | ||
| 877 | 1491 | 'body' => wp_json_encode(['input' => $text, 'model' => $model]), |
| 878 | 1492 | 'timeout' => 60, |
| 879 | 1493 | ]); |
| 880 | 1494 | if (is_wp_error($response)) { |
| 881 | - return 'Connection error when generating embeddings (custom provider): ' . $response->get_error_message(); | |
| 1495 | + return self::log_custom_embedding_failure( | |
| 1496 | + 'Connection error when generating embeddings (custom provider): ' . $response->get_error_message(), | |
| 1497 | + $model, | |
| 1498 | + $api_key | |
| 1499 | + ); | |
| 882 | 1500 | } |
| 883 | 1501 | |
| 884 | 1502 | $status = wp_remote_retrieve_response_code($response); |
| 885 | 1503 | $body = json_decode(wp_remote_retrieve_body($response), true); |
| @@ -884,9 +1502,14 @@ | ||
| 884 | 1502 | $status = wp_remote_retrieve_response_code($response); |
| 885 | 1503 | $body = json_decode(wp_remote_retrieve_body($response), true); |
| 886 | 1504 | if ($status !== 200) { |
| 887 | 1505 | $msg = isset($body['error']['message']) ? $body['error']['message'] : 'HTTP ' . $status; |
| 888 | - return 'Custom embedding endpoint error: ' . $msg; | |
| 1506 | + return self::log_custom_embedding_failure( | |
| 1507 | + 'Custom embedding endpoint error: ' . $msg, | |
| 1508 | + $model, | |
| 1509 | + $api_key, | |
| 1510 | + (int) $status | |
| 1511 | + ); | |
| 889 | 1512 | } |
| 890 | 1513 | if (isset($body['data'][0]['embedding']) && is_array($body['data'][0]['embedding'])) { |
| 891 | 1514 | // Stamp the custom model identity so the active-embedding-model mismatch |
| 892 | 1515 | // warning reflects the real (custom) model rather than the built-in setting. |
| @@ -892,12 +1515,42 @@ | ||
| 892 | 1515 | // warning reflects the real (custom) model rather than the built-in setting. |
| 893 | 1516 | self::stamp_active_embedding_model('custom:' . $model); |
| 894 | 1517 | return $body['data'][0]['embedding']; |
| 895 | 1518 | } |
| 896 | - return 'Invalid embedding response from custom provider.'; | |
| 1519 | + return self::log_custom_embedding_failure('Invalid embedding response from custom provider.', $model, $api_key); | |
| 897 | 1520 | } |
| 898 | 1521 | |
| 899 | 1522 | /** |
| 1523 | + * Record a custom-provider embedding failure in the Debug Mode log, then | |
| 1524 | + * return the message unchanged so callers keep their string-error contract. | |
| 1525 | + * The cloud branch has logged its failures since 4a7c0a; the custom branch | |
| 1526 | + * never did, so chat-side failures on Custom-provider installs were | |
| 1527 | + * invisible to Debug Mode despite the 3.2.18 readme saying otherwise | |
| 1528 | + * (plan 71e4b6). Same scrub-then-log shape as embedding_failure_error(). | |
| 1529 | + * | |
| 1530 | + * @param string $message Human-readable failure (the caller's return value). | |
| 1531 | + * @param string $model Resolved custom embedding model. | |
| 1532 | + * @param string $api_key Scrubbed out of the logged message if it ever appears. | |
| 1533 | + * @param int $status HTTP status when one was received, 0 otherwise. | |
| 1534 | + * @return string The (scrubbed) message. | |
| 1535 | + */ | |
| 1536 | +private static function log_custom_embedding_failure($message, $model, $api_key, $status = 0) { | |
| 1537 | + if (is_string($api_key) && $api_key !== '') { | |
| 1538 | + $message = str_replace($api_key, '[redacted]', $message); | |
| 1539 | + } | |
| 1540 | + | |
| 1541 | + if (class_exists('MxChat_Admin')) { | |
| 1542 | + $context = array('model' => 'custom:' . $model); | |
| 1543 | + if ($status > 0) { | |
| 1544 | + $context['status'] = $status; | |
| 1545 | + } | |
| 1546 | + MxChat_Admin::mxchat_log_debug('embedding_error', $message, $context); | |
| 1547 | + } | |
| 1548 | + | |
| 1549 | + return $message; | |
| 1550 | +} | |
| 1551 | + | |
| 1552 | +/** | |
| 900 | 1553 | * Submit content as multiple chunks |
| 901 | 1554 | * |
| 902 | 1555 | * Splits large content into chunks, generates embeddings for each, |
| 903 | 1556 | * and stores them with chunk metadata for later reassembly. |
| @@ -916,10 +1569,26 @@ | ||
| 916 | 1569 | |
| 917 | 1570 | //error_log('[MXCHAT-CHUNK-DEBUG] Starting chunked submission for: ' . $source_url); |
| 918 | 1571 | //error_log('[MXCHAT-CHUNK-DEBUG] Content length: ' . strlen($content) . ' chars'); |
| 919 | 1572 | |
| 920 | - // First, delete any existing chunks for this URL (clean slate) | |
| 1573 | + // URL-less (manual) content needs a minted identity BEFORE chunk ids are derived: | |
| 1574 | + // every chunk id is md5(source_url)_chunk_N, so with source_url = '' EVERY long | |
| 1575 | + // manual document shared the md5('') prefix — and the clean-slate delete below | |
| 1576 | + // wiped the PREVIOUS manual entry's chunks each time a new one was added. The | |
| 1577 | + // single-vector paths already mint (mxchat:// in WP, manual_* in Pinecone); this | |
| 1578 | + // was the one storage path that didn't. Keep an mxchat:// identity if the caller | |
| 1579 | + // already carries one. | |
| 1580 | + if (strpos($source_url, 'mxchat://') !== 0 && strpos($source_url, 'upload://') !== 0 && !preg_match('#^https?://#i', $source_url)) { | |
| 1581 | + $source_url = 'mxchat://manual-content/' . time() . '-' . wp_generate_password(8, false); | |
| 1582 | + } | |
| 1583 | + | |
| 1584 | + // First, delete any existing chunks for this URL (clean slate). | |
| 1585 | + // Mirror-suspended: this is a re-store, not an entry removal — the Vector | |
| 1586 | + // Store file is replaced (or kept, on hash match) by the caller's | |
| 1587 | + // sync_upsert_entry after storage succeeds. | |
| 1588 | + self::$vectorstore_mirror_suspended = true; | |
| 921 | 1589 | $delete_result = self::delete_chunks_for_url($source_url, $bot_id); |
| 1590 | + self::$vectorstore_mirror_suspended = false; | |
| 922 | 1591 | if (is_wp_error($delete_result)) { |
| 923 | 1592 | //error_log('[MXCHAT-CHUNK-DEBUG] Warning: Failed to delete existing chunks: ' . $delete_result->get_error_message()); |
| 924 | 1593 | // Continue anyway - we'll overwrite with upsert |
| 925 | 1594 | } |
| @@ -1108,9 +1777,22 @@ | ||
| 1108 | 1777 | 'last_updated' => time(), |
| 1109 | 1778 | 'created_at' => time(), |
| 1110 | 1779 | 'bot_id' => $bot_id, |
| 1111 | 1780 | ); |
| 1781 | + $page_title = self::source_title_for($source_url); | |
| 1782 | + if ($page_title !== '') { | |
| 1783 | + $metadata['title'] = $page_title; // page title captured at import (plan d9ee66) | |
| 1784 | + } | |
| 1112 | 1785 | |
| 1786 | + // Document index (plan 362c31): same chunk id and metadata, documents API. | |
| 1787 | + if (class_exists('MxChat_Pinecone_Documents') && MxChat_Pinecone_Documents::is_document_index($bot_id)) { | |
| 1788 | + return MxChat_Pinecone_Documents::upsert_one($vector_id, $chunk_text, $embedding_vector, $metadata, array( | |
| 1789 | + 'api_key' => $api_key, | |
| 1790 | + 'host' => $host, | |
| 1791 | + 'namespace' => $namespace, | |
| 1792 | + )); | |
| 1793 | + } | |
| 1794 | + | |
| 1113 | 1795 | $vector_data = array( |
| 1114 | 1796 | 'id' => $vector_id, |
| 1115 | 1797 | 'values' => $embedding_vector, |
| 1116 | 1798 | 'metadata' => $metadata |
| @@ -1153,20 +1835,25 @@ | ||
| 1153 | 1835 | global $wpdb; |
| 1154 | 1836 | |
| 1155 | 1837 | // For chunks, we always insert new rows (no duplicate checking) |
| 1156 | 1838 | // The URL includes chunk info in the metadata, but source_url stays the same for grouping |
| 1157 | - $result = $wpdb->insert( | |
| 1158 | - $table_name, | |
| 1159 | - array( | |
| 1160 | - 'url' => $source_url, | |
| 1161 | - 'article_content' => $content_with_metadata, | |
| 1162 | - 'embedding_vector' => $embedding_vector_serialized, | |
| 1163 | - 'source_url' => $source_url, | |
| 1164 | - 'content_type' => $content_type, | |
| 1165 | - 'timestamp' => current_time('mysql') | |
| 1166 | - ), | |
| 1167 | - array('%s', '%s', '%s', '%s', '%s', '%s') | |
| 1839 | + $row = array( | |
| 1840 | + 'url' => $source_url, | |
| 1841 | + 'article_content' => $content_with_metadata, | |
| 1842 | + 'embedding_vector' => $embedding_vector_serialized, | |
| 1843 | + 'source_url' => $source_url, | |
| 1844 | + 'content_type' => $content_type, | |
| 1845 | + 'timestamp' => current_time('mysql') | |
| 1168 | 1846 | ); |
| 1847 | + $formats = array('%s', '%s', '%s', '%s', '%s', '%s'); | |
| 1848 | + // Every chunk row carries the page title too (plan d9ee66) - readers group by | |
| 1849 | + // source_url and may pick any chunk as the representative. | |
| 1850 | + $page_title = self::source_title_for($source_url); | |
| 1851 | + if ($page_title !== '' && self::ensure_kb_title_column()) { | |
| 1852 | + $row['title'] = $page_title; | |
| 1853 | + $formats[] = '%s'; | |
| 1854 | + } | |
| 1855 | + $result = $wpdb->insert($table_name, $row, $formats); | |
| 1169 | 1856 | |
| 1170 | 1857 | if ($result === false) { |
| 1171 | 1858 | return new WP_Error('database_failed', 'Failed to insert chunk: ' . $wpdb->last_error); |
| 1172 | 1859 | } |
| @@ -1183,8 +1870,14 @@ | ||
| 1183 | 1870 | */ |
| 1184 | 1871 | public static function delete_chunks_for_url($source_url, $bot_id = 'default') { |
| 1185 | 1872 | //error_log('[MXCHAT-CHUNK-DELETE] Deleting chunks for URL: ' . $source_url); |
| 1186 | 1873 | |
| 1874 | + // Entry removal — mirror it to the Vector Store (unless a storage routine | |
| 1875 | + // is mid-re-store, see $vectorstore_mirror_suspended). | |
| 1876 | + if (!self::$vectorstore_mirror_suspended && class_exists('MxChat_Vectorstore_Manager')) { | |
| 1877 | + MxChat_Vectorstore_Manager::sync_delete_entry($source_url, $bot_id); | |
| 1878 | + } | |
| 1879 | + | |
| 1187 | 1880 | if (self::is_pinecone_enabled_for_bot($bot_id)) { |
| 1188 | 1881 | return self::delete_pinecone_chunks_by_url($source_url, $bot_id); |
| 1189 | 1882 | } else { |
| 1190 | 1883 | return self::delete_wordpress_chunks_by_url($source_url); |
| @@ -1193,10 +1886,16 @@ | ||
| 1193 | 1886 | |
| 1194 | 1887 | /** |
| 1195 | 1888 | * Delete all chunks for a URL from Pinecone |
| 1196 | 1889 | */ |
| 1197 | -private static function delete_pinecone_chunks_by_url($source_url, $bot_id) { | |
| 1198 | - // Get Pinecone configuration | |
| 1890 | +/** | |
| 1891 | + * Delete leftover md5(url)_chunk_N vectors after a URL's content was re-stored | |
| 1892 | + * as a SINGLE vector (content shrank below the chunk threshold on edit/re-import). | |
| 1893 | + * Unlike delete_pinecone_chunks_by_url this leaves the base id alone — the caller | |
| 1894 | + * just upserted the new content there. Failure is logged, not fatal: the save | |
| 1895 | + * itself succeeded, and the next save retries the sweep. | |
| 1896 | + */ | |
| 1897 | +private static function cleanup_pinecone_chunk_stragglers($source_url, $bot_id) { | |
| 1199 | 1898 | if ($bot_id === 'default' || !class_exists('MxChat_Multi_Bot_Manager')) { |
| 1200 | 1899 | $pinecone_options = get_option('mxchat_pinecone_addon_options'); |
| 1201 | 1900 | $api_key = $pinecone_options['mxchat_pinecone_api_key'] ?? ''; |
| 1202 | 1901 | $host = $pinecone_options['mxchat_pinecone_host'] ?? ''; |
| @@ -1215,21 +1914,30 @@ | ||
| 1215 | 1914 | } |
| 1216 | 1915 | } |
| 1217 | 1916 | |
| 1218 | 1917 | if (empty($host) || empty($api_key)) { |
| 1219 | - return new WP_Error('pinecone_config', 'Pinecone is not properly configured'); | |
| 1918 | + return; | |
| 1220 | 1919 | } |
| 1221 | 1920 | |
| 1222 | - $base_vector_id = md5($source_url); | |
| 1223 | - $vectors_to_delete = array(); | |
| 1921 | + // Document index (plan 362c31): list the chunk ids and delete them through the documents API. | |
| 1922 | + if (class_exists('MxChat_Pinecone_Documents') && MxChat_Pinecone_Documents::is_document_index($bot_id)) { | |
| 1923 | + $docs_cfg = array('api_key' => $api_key, 'host' => $host, 'namespace' => $namespace); | |
| 1924 | + $doc_stragglers = MxChat_Pinecone_Documents::list_ids(md5($source_url) . '_chunk_', $docs_cfg); | |
| 1925 | + if (!empty($doc_stragglers)) { | |
| 1926 | + $doc_result = MxChat_Pinecone_Documents::delete_ids($doc_stragglers, $docs_cfg); | |
| 1927 | + if (is_wp_error($doc_result) && class_exists('MxChat_Admin') && method_exists('MxChat_Admin', 'mxchat_log_debug')) { | |
| 1928 | + MxChat_Admin::mxchat_log_debug('pinecone_error', 'Failed to sweep stale chunk documents after single-vector re-store', array('source_url' => $source_url, 'bot_id' => $bot_id, 'count' => count($doc_stragglers))); | |
| 1929 | + } | |
| 1930 | + } | |
| 1931 | + return; | |
| 1932 | + } | |
| 1224 | 1933 | |
| 1225 | - // Add the original single-vector ID (for non-chunked content) | |
| 1226 | - $vectors_to_delete[] = $base_vector_id; | |
| 1934 | + $stragglers = array(); | |
| 1227 | 1935 | |
| 1228 | - // Pinecone /vectors/list is a GET endpoint with query-string parameters; a POST here returns a | |
| 1229 | - // non-200 silently and we end up only deleting the base vector, leaving chunks orphaned. | |
| 1936 | + // Pinecone /vectors/list is a GET endpoint with query-string parameters (a POST | |
| 1937 | + // answers 200-with-an-empty-body, which reads as "no stragglers"). | |
| 1230 | 1938 | $query_params = array( |
| 1231 | - 'prefix' => $base_vector_id . '_chunk_', | |
| 1939 | + 'prefix' => md5($source_url) . '_chunk_', | |
| 1232 | 1940 | 'limit' => 100, |
| 1233 | 1941 | ); |
| 1234 | 1942 | if (!empty($namespace)) { |
| 1235 | 1943 | $query_params['namespace'] = $namespace; |
| @@ -1236,9 +1944,8 @@ | ||
| 1236 | 1944 | } |
| 1237 | 1945 | |
| 1238 | 1946 | $list_url = "https://{$host}/vectors/list?" . http_build_query($query_params); |
| 1239 | 1947 | |
| 1240 | - // Paginate in case a URL has more than 100 chunks. | |
| 1241 | 1948 | do { |
| 1242 | 1949 | $list_response = wp_remote_get($list_url, array( |
| 1243 | 1950 | 'headers' => array( |
| 1244 | 1951 | 'Api-Key' => $api_key, |
| @@ -1254,9 +1961,9 @@ | ||
| 1254 | 1961 | $list_data = json_decode(wp_remote_retrieve_body($list_response), true); |
| 1255 | 1962 | if (!empty($list_data['vectors'])) { |
| 1256 | 1963 | foreach ($list_data['vectors'] as $vector) { |
| 1257 | 1964 | if (isset($vector['id'])) { |
| 1258 | - $vectors_to_delete[] = $vector['id']; | |
| 1965 | + $stragglers[] = $vector['id']; | |
| 1259 | 1966 | } |
| 1260 | 1967 | } |
| 1261 | 1968 | } |
| 1262 | 1969 | |
| @@ -1268,8 +1975,188 @@ | ||
| 1268 | 1975 | $query_params['paginationToken'] = $next_token; |
| 1269 | 1976 | $list_url = "https://{$host}/vectors/list?" . http_build_query($query_params); |
| 1270 | 1977 | } while (true); |
| 1271 | 1978 | |
| 1979 | + if (empty($stragglers)) { | |
| 1980 | + return; | |
| 1981 | + } | |
| 1982 | + | |
| 1983 | + $delete_body = array('ids' => $stragglers); | |
| 1984 | + if (!empty($namespace)) { | |
| 1985 | + $delete_body['namespace'] = $namespace; | |
| 1986 | + } | |
| 1987 | + | |
| 1988 | + $delete_response = wp_remote_post("https://{$host}/vectors/delete", array( | |
| 1989 | + 'headers' => array( | |
| 1990 | + 'Api-Key' => $api_key, | |
| 1991 | + 'accept' => 'application/json', | |
| 1992 | + 'content-type' => 'application/json' | |
| 1993 | + ), | |
| 1994 | + 'body' => wp_json_encode($delete_body), | |
| 1995 | + 'timeout' => 30 | |
| 1996 | + )); | |
| 1997 | + | |
| 1998 | + if ((is_wp_error($delete_response) || wp_remote_retrieve_response_code($delete_response) !== 200) | |
| 1999 | + && class_exists('MxChat_Admin') && method_exists('MxChat_Admin', 'mxchat_log_debug')) { | |
| 2000 | + MxChat_Admin::mxchat_log_debug('pinecone_error', 'Failed to sweep stale chunk vectors after single-vector re-store', array('source_url' => $source_url, 'bot_id' => $bot_id, 'count' => count($stragglers))); | |
| 2001 | + } | |
| 2002 | +} | |
| 2003 | + | |
| 2004 | +/** | |
| 2005 | + * Resolve the Pinecone credentials for a bot: the bot's own config when the | |
| 2006 | + * multi-bot add-on supplies one, else the site-wide options. Extracted so the | |
| 2007 | + * delete path and the count path can never resolve different credentials. | |
| 2008 | + * | |
| 2009 | + * @param string $bot_id | |
| 2010 | + * @return array{api_key:string,host:string,namespace:string} | |
| 2011 | + */ | |
| 2012 | +private static function pinecone_config_for_bot($bot_id) { | |
| 2013 | + if ($bot_id === 'default' || !class_exists('MxChat_Multi_Bot_Manager')) { | |
| 2014 | + $bot_pinecone_config = array(); | |
| 2015 | + } else { | |
| 2016 | + $bot_pinecone_config = apply_filters('mxchat_get_bot_pinecone_config', array(), $bot_id); | |
| 2017 | + } | |
| 2018 | + | |
| 2019 | + if (empty($bot_pinecone_config)) { | |
| 2020 | + $pinecone_options = get_option('mxchat_pinecone_addon_options'); | |
| 2021 | + return array( | |
| 2022 | + 'api_key' => $pinecone_options['mxchat_pinecone_api_key'] ?? '', | |
| 2023 | + 'host' => $pinecone_options['mxchat_pinecone_host'] ?? '', | |
| 2024 | + 'namespace' => $pinecone_options['mxchat_pinecone_namespace'] ?? '', | |
| 2025 | + // 'vector' | 'document' (plan 362c31) — the delete / count paths branch on it. | |
| 2026 | + 'index_type' => (($pinecone_options['mxchat_pinecone_index_type'] ?? 'vector') === 'document') ? 'document' : 'vector', | |
| 2027 | + ); | |
| 2028 | + } | |
| 2029 | + | |
| 2030 | + return array( | |
| 2031 | + 'api_key' => $bot_pinecone_config['api_key'] ?? '', | |
| 2032 | + 'host' => $bot_pinecone_config['host'] ?? '', | |
| 2033 | + 'namespace' => $bot_pinecone_config['namespace'] ?? '', | |
| 2034 | + 'index_type' => (($bot_pinecone_config['index_type'] ?? 'vector') === 'document') ? 'document' : 'vector', | |
| 2035 | + ); | |
| 2036 | +} | |
| 2037 | + | |
| 2038 | +/** | |
| 2039 | + * List every Pinecone vector id under a prefix, following pagination. | |
| 2040 | + * | |
| 2041 | + * Pinecone /vectors/list is a GET endpoint with query-string parameters; a POST here returns a | |
| 2042 | + * non-200 silently and we end up seeing only the base vector, leaving chunks orphaned. | |
| 2043 | + * | |
| 2044 | + * @param string $prefix Vector-id prefix to match. | |
| 2045 | + * @param array $cfg From pinecone_config_for_bot(). | |
| 2046 | + * @return array List of vector ids (empty when none match or the API is unreachable). | |
| 2047 | + */ | |
| 2048 | +private static function pinecone_list_ids_by_prefix($prefix, array $cfg) { | |
| 2049 | + // Document index (plan 362c31): the documents API lists by prefix too. | |
| 2050 | + if (($cfg['index_type'] ?? 'vector') === 'document' && class_exists('MxChat_Pinecone_Documents')) { | |
| 2051 | + return MxChat_Pinecone_Documents::list_ids($prefix, $cfg); | |
| 2052 | + } | |
| 2053 | + | |
| 2054 | + $ids = array(); | |
| 2055 | + | |
| 2056 | + $query_params = array( | |
| 2057 | + 'prefix' => $prefix, | |
| 2058 | + 'limit' => 100, | |
| 2059 | + ); | |
| 2060 | + if (!empty($cfg['namespace'])) { | |
| 2061 | + $query_params['namespace'] = $cfg['namespace']; | |
| 2062 | + } | |
| 2063 | + | |
| 2064 | + $list_url = "https://{$cfg['host']}/vectors/list?" . http_build_query($query_params); | |
| 2065 | + | |
| 2066 | + // Paginate in case a URL has more than 100 chunks. | |
| 2067 | + do { | |
| 2068 | + $list_response = wp_remote_get($list_url, array( | |
| 2069 | + 'headers' => array( | |
| 2070 | + 'Api-Key' => $cfg['api_key'], | |
| 2071 | + 'accept' => 'application/json', | |
| 2072 | + ), | |
| 2073 | + 'timeout' => 30, | |
| 2074 | + )); | |
| 2075 | + | |
| 2076 | + if (is_wp_error($list_response) || wp_remote_retrieve_response_code($list_response) !== 200) { | |
| 2077 | + break; | |
| 2078 | + } | |
| 2079 | + | |
| 2080 | + $list_data = json_decode(wp_remote_retrieve_body($list_response), true); | |
| 2081 | + if (!empty($list_data['vectors'])) { | |
| 2082 | + foreach ($list_data['vectors'] as $vector) { | |
| 2083 | + if (isset($vector['id'])) { | |
| 2084 | + $ids[] = $vector['id']; | |
| 2085 | + } | |
| 2086 | + } | |
| 2087 | + } | |
| 2088 | + | |
| 2089 | + $next_token = $list_data['pagination']['next'] ?? ''; | |
| 2090 | + if (empty($next_token)) { | |
| 2091 | + break; | |
| 2092 | + } | |
| 2093 | + | |
| 2094 | + $query_params['paginationToken'] = $next_token; | |
| 2095 | + $list_url = "https://{$cfg['host']}/vectors/list?" . http_build_query($query_params); | |
| 2096 | + } while (true); | |
| 2097 | + | |
| 2098 | + return $ids; | |
| 2099 | +} | |
| 2100 | + | |
| 2101 | +/** | |
| 2102 | + * How many stored pieces a URL currently has, in whichever backend holds it. | |
| 2103 | + * | |
| 2104 | + * Branches on the SAME condition delete_chunks_for_url() branches on, so the | |
| 2105 | + * number reported is the number that call removes. Used by the REST DELETE | |
| 2106 | + * route to answer "was anything actually there?" without a second delete path. | |
| 2107 | + * | |
| 2108 | + * @param string $source_url | |
| 2109 | + * @param string $bot_id | |
| 2110 | + * @return int | |
| 2111 | + */ | |
| 2112 | +public static function count_chunks_for_url($source_url, $bot_id = 'default') { | |
| 2113 | + if (self::is_pinecone_enabled_for_bot($bot_id)) { | |
| 2114 | + $cfg = self::pinecone_config_for_bot($bot_id); | |
| 2115 | + if (empty($cfg['host']) || empty($cfg['api_key'])) { | |
| 2116 | + return 0; | |
| 2117 | + } | |
| 2118 | + // The base id and every "<base>_chunk_N" share the base as a prefix, | |
| 2119 | + // so one prefix listing covers chunked and non-chunked content alike. | |
| 2120 | + return count(self::pinecone_list_ids_by_prefix(md5($source_url), $cfg)); | |
| 2121 | + } | |
| 2122 | + | |
| 2123 | + global $wpdb; | |
| 2124 | + $table_name = $wpdb->prefix . 'mxchat_system_prompt_content'; | |
| 2125 | + | |
| 2126 | + return (int) $wpdb->get_var($wpdb->prepare( | |
| 2127 | + "SELECT COUNT(*) FROM {$table_name} WHERE source_url = %s", | |
| 2128 | + $source_url | |
| 2129 | + )); | |
| 2130 | +} | |
| 2131 | + | |
| 2132 | +private static function delete_pinecone_chunks_by_url($source_url, $bot_id) { | |
| 2133 | + $cfg = self::pinecone_config_for_bot($bot_id); | |
| 2134 | + $api_key = $cfg['api_key']; | |
| 2135 | + $host = $cfg['host']; | |
| 2136 | + $namespace = $cfg['namespace']; | |
| 2137 | + | |
| 2138 | + if (empty($host) || empty($api_key)) { | |
| 2139 | + return new WP_Error('pinecone_config', 'Pinecone is not properly configured'); | |
| 2140 | + } | |
| 2141 | + | |
| 2142 | + // Document index (plan 362c31): base id + every <md5>_chunk_N, documents API. | |
| 2143 | + if (($cfg['index_type'] ?? 'vector') === 'document' && class_exists('MxChat_Pinecone_Documents')) { | |
| 2144 | + $doc_ids = array_merge(array(md5($source_url)), MxChat_Pinecone_Documents::list_ids(md5($source_url) . '_chunk_', $cfg)); | |
| 2145 | + return MxChat_Pinecone_Documents::delete_ids($doc_ids, $cfg); | |
| 2146 | + } | |
| 2147 | + | |
| 2148 | + $base_vector_id = md5($source_url); | |
| 2149 | + $vectors_to_delete = array(); | |
| 2150 | + | |
| 2151 | + // Add the original single-vector ID (for non-chunked content) | |
| 2152 | + $vectors_to_delete[] = $base_vector_id; | |
| 2153 | + | |
| 2154 | + $vectors_to_delete = array_merge( | |
| 2155 | + $vectors_to_delete, | |
| 2156 | + self::pinecone_list_ids_by_prefix($base_vector_id . '_chunk_', $cfg) | |
| 2157 | + ); | |
| 2158 | + | |
| 1272 | 2159 | if (empty($vectors_to_delete)) { |
| 1273 | 2160 | //error_log('[MXCHAT-CHUNK-DELETE] No vectors found to delete'); |
| 1274 | 2161 | return true; |
| 1275 | 2162 | } |
| @@ -1364,6 +2251,261 @@ | ||
| 1364 | 2251 | |
| 1365 | 2252 | $capability = $index_exists ? 'fulltext' : 'like'; |
| 1366 | 2253 | update_option('mxchat_hybrid_keyword_capability', $capability); |
| 1367 | 2254 | return $capability; |
| 2255 | +} | |
| 2256 | + | |
| 2257 | +/** | |
| 2258 | + * Public entry for re-embedding already-stored content in place (wp mxchat | |
| 2259 | + * rtl-repair, plan d1e6f7). Thin wrapper so the repair CLI gets the exact | |
| 2260 | + * provider routing the import path uses — the repaired vector must come from | |
| 2261 | + * the same model family the bot indexes with, or retrieval stays broken. | |
| 2262 | + */ | |
| 2263 | +public static function regenerate_embedding($text, $api_key, $bot_id = 'default') { | |
| 2264 | + return self::generate_embedding($text, $api_key, $bot_id); | |
| 2265 | +} | |
| 2266 | + | |
| 2267 | +/** | |
| 2268 | + * Restore logical character order in PDF-extracted RTL text (plan 32bf9e). | |
| 2269 | + * | |
| 2270 | + * The bundled Smalot parser only un-reverses text runs tagged with the | |
| 2271 | + * ReversedChars marked-content operator (Word emits it; LibreOffice and most | |
| 2272 | + * other producers do not), so their Hebrew/Arabic PDFs extract in visual | |
| 2273 | + * (reversed) order and embed/search as garbage. This is OUR post-processing | |
| 2274 | + * seam over getText() — the parser itself is never patched (it gets replaced | |
| 2275 | + * wholesale on library updates). | |
| 2276 | + * | |
| 2277 | + * Heuristic and deliberately conservative, per line: | |
| 2278 | + * - lines without strong RTL codepoints are untouched (a fully-Latin line in | |
| 2279 | + * an RTL document therefore stays as extracted — accepted limitation); | |
| 2280 | + * - Arabic presentation forms are a definitive visual-order signal (they only | |
| 2281 | + * appear in shaped output): de-shape to base letters and reverse; | |
| 2282 | + * - otherwise flip only on positive evidence — Hebrew final-letter position | |
| 2283 | + * (a sofit at word START only happens in reversed text) or sentence | |
| 2284 | + * punctuation position (leading in visual order, trailing in logical); | |
| 2285 | + * - ambiguous lines are left alone: a conservative miss beats corrupting a | |
| 2286 | + * Word-produced extraction the parser already handled (the double-flip | |
| 2287 | + * guard this plan's approval named mandatory). | |
| 2288 | + * | |
| 2289 | + * @param string $text One extracted page string, straight from getText(). | |
| 2290 | + * @param string $context Caller tag for the Debug Mode entry (site + page). | |
| 2291 | + * @return string Text with RTL lines restored to logical order. | |
| 2292 | + */ | |
| 2293 | +public static function normalize_pdf_rtl($text, $context = '') { | |
| 2294 | + if (!is_string($text) || '' === $text) { | |
| 2295 | + return $text; | |
| 2296 | + } | |
| 2297 | + // Escape hatch for sites whose PDFs already extract logically. | |
| 2298 | + if (!apply_filters('mxchat_pdf_rtl_normalize', true, $text)) { | |
| 2299 | + return $text; | |
| 2300 | + } | |
| 2301 | + // Fast bail: nothing RTL anywhere in the page. | |
| 2302 | + if (!preg_match('/[\x{0590}-\x{05FF}\x{0600}-\x{06FF}\x{0750}-\x{077F}\x{FB50}-\x{FDFF}\x{FE70}-\x{FEFF}]/u', $text)) { | |
| 2303 | + return $text; | |
| 2304 | + } | |
| 2305 | + | |
| 2306 | + $parts = preg_split('/(\R)/u', $text, -1, PREG_SPLIT_DELIM_CAPTURE); | |
| 2307 | + if (false === $parts) { | |
| 2308 | + return $text; | |
| 2309 | + } | |
| 2310 | + | |
| 2311 | + $flipped_lines = 0; | |
| 2312 | + $deshaped_lines = 0; | |
| 2313 | + foreach ($parts as $i => $part) { | |
| 2314 | + if ('' === $part || preg_match('/^\R$/u', $part)) { | |
| 2315 | + continue; | |
| 2316 | + } | |
| 2317 | + $was_flipped = false; | |
| 2318 | + $was_deshaped = false; | |
| 2319 | + $new = self::pdf_rtl_normalize_line($part, $was_flipped, $was_deshaped); | |
| 2320 | + if ($new !== $part) { | |
| 2321 | + $parts[$i] = $new; | |
| 2322 | + } | |
| 2323 | + if ($was_flipped) { | |
| 2324 | + $flipped_lines++; | |
| 2325 | + } | |
| 2326 | + if ($was_deshaped) { | |
| 2327 | + $deshaped_lines++; | |
| 2328 | + } | |
| 2329 | + } | |
| 2330 | + | |
| 2331 | + if (($flipped_lines || $deshaped_lines) && class_exists('MxChat_Admin')) { | |
| 2332 | + MxChat_Admin::mxchat_log_debug('pdf_rtl_normalized', 'RTL PDF text restored to logical order', array( | |
| 2333 | + 'context' => (string) $context, | |
| 2334 | + 'lines_flipped' => $flipped_lines, | |
| 2335 | + 'lines_deshaped' => $deshaped_lines, | |
| 2336 | + 'decision' => 'visual-order extraction detected', | |
| 2337 | + )); | |
| 2338 | + } | |
| 2339 | + | |
| 2340 | + return implode('', $parts); | |
| 2341 | +} | |
| 2342 | + | |
| 2343 | +/** | |
| 2344 | + * Normalize one line. Sets $flipped/$deshaped for the caller's debug entry. | |
| 2345 | + */ | |
| 2346 | +private static function pdf_rtl_normalize_line($line, &$flipped, &$deshaped) { | |
| 2347 | + $flipped = false; | |
| 2348 | + $deshaped = false; | |
| 2349 | + | |
| 2350 | + if (!preg_match('/[\x{0590}-\x{05FF}\x{0600}-\x{06FF}\x{0750}-\x{077F}\x{FB50}-\x{FDFF}\x{FE70}-\x{FEFF}]/u', $line)) { | |
| 2351 | + return $line; | |
| 2352 | + } | |
| 2353 | + | |
| 2354 | + $has_forms = (bool) preg_match('/[\x{FB50}-\x{FDFF}\x{FE70}-\x{FEFF}]/u', $line); | |
| 2355 | + $work = $line; | |
| 2356 | + if ($has_forms) { | |
| 2357 | + $work = strtr($work, self::pdf_rtl_deshape_map()); | |
| 2358 | + $deshaped = ($work !== $line); | |
| 2359 | + } | |
| 2360 | + | |
| 2361 | + $verdict = 'ambiguous'; | |
| 2362 | + if ($has_forms) { | |
| 2363 | + // Shaped glyph codepoints only exist in visual-order output. | |
| 2364 | + $verdict = 'visual'; | |
| 2365 | + } else { | |
| 2366 | + // Strong-direction dominance gate first: an LTR-dominant line with an | |
| 2367 | + // embedded RTL word is not flip material. | |
| 2368 | + $rtl_count = preg_match_all('/[\x{0590}-\x{05FF}\x{0600}-\x{06FF}\x{0750}-\x{077F}]/u', $work, $m_rtl); | |
| 2369 | + $ltr_count = preg_match_all('/[A-Za-z]/u', $work, $m_ltr); | |
| 2370 | + if ($rtl_count < 1 || $rtl_count <= $ltr_count) { | |
| 2371 | + return $line; | |
| 2372 | + } | |
| 2373 | + | |
| 2374 | + // Hebrew final letters (ך ם ן ף ץ) end words in logical text; one at | |
| 2375 | + // a word START (Hebrew letter follows, none precedes) is reversal | |
| 2376 | + // evidence. Positional, so it survives the line being reversed. | |
| 2377 | + $sofit_initial = preg_match_all('/(?<![\x{05D0}-\x{05EA}])[\x{05DA}\x{05DD}\x{05DF}\x{05E3}\x{05E5}](?=[\x{05D0}-\x{05EA}])/u', $work, $m_i); | |
| 2378 | + $sofit_terminal = preg_match_all('/(?<=[\x{05D0}-\x{05EA}])[\x{05DA}\x{05DD}\x{05DF}\x{05E3}\x{05E5}](?![\x{05D0}-\x{05EA}])/u', $work, $m_t); | |
| 2379 | + if ($sofit_initial > $sofit_terminal) { | |
| 2380 | + $verdict = 'visual'; | |
| 2381 | + } elseif ($sofit_terminal > $sofit_initial) { | |
| 2382 | + $verdict = 'logical'; | |
| 2383 | + } else { | |
| 2384 | + // Sentence punctuation lands at the visual LEFT edge of an RTL | |
| 2385 | + // line, i.e. the START of a visual-order extraction. | |
| 2386 | + $trimmed = trim($work); | |
| 2387 | + $starts_punct = (bool) preg_match('/^[.?!:;,]/u', $trimmed); | |
| 2388 | + $ends_punct = (bool) preg_match('/[.?!:;,]$/u', $trimmed); | |
| 2389 | + if ($starts_punct && !$ends_punct) { | |
| 2390 | + $verdict = 'visual'; | |
| 2391 | + } elseif ($ends_punct && !$starts_punct) { | |
| 2392 | + $verdict = 'logical'; | |
| 2393 | + } | |
| 2394 | + } | |
| 2395 | + } | |
| 2396 | + | |
| 2397 | + if ('visual' !== $verdict) { | |
| 2398 | + // Ambiguous or logical: hand back the original line UNLESS we | |
| 2399 | + // de-shaped (de-shaping alone is always safe — same letters, same | |
| 2400 | + // order, un-ligated). | |
| 2401 | + return $deshaped ? $work : $line; | |
| 2402 | + } | |
| 2403 | + | |
| 2404 | + $flipped = true; | |
| 2405 | + return self::pdf_rtl_flip_line($work); | |
| 2406 | +} | |
| 2407 | + | |
| 2408 | +/** | |
| 2409 | + * Reverse a visual-order line into logical order: full character reversal, | |
| 2410 | + * mirror paired punctuation, then re-reverse embedded LTR runs (Latin words | |
| 2411 | + * and digit sequences, incl. Arabic-Indic digits) so they stay readable. | |
| 2412 | + */ | |
| 2413 | +private static function pdf_rtl_flip_line($line) { | |
| 2414 | + $chars = preg_split('//u', $line, -1, PREG_SPLIT_NO_EMPTY); | |
| 2415 | + if (false === $chars) { | |
| 2416 | + return $line; | |
| 2417 | + } | |
| 2418 | + $reversed = implode('', array_reverse($chars)); | |
| 2419 | + $reversed = strtr($reversed, array( | |
| 2420 | + '(' => ')', ')' => '(', | |
| 2421 | + '[' => ']', ']' => '[', | |
| 2422 | + '{' => '}', '}' => '{', | |
| 2423 | + '<' => '>', '>' => '<', | |
| 2424 | + )); | |
| 2425 | + $restored = preg_replace_callback( | |
| 2426 | + '/[0-9A-Za-z\x{0660}-\x{0669}\x{06F0}-\x{06F9}](?:[0-9A-Za-z\x{0660}-\x{0669}\x{06F0}-\x{06F9} .,\'"%\-:\/]*[0-9A-Za-z\x{0660}-\x{0669}\x{06F0}-\x{06F9}])?/u', | |
| 2427 | + function ($m) { | |
| 2428 | + $run = preg_split('//u', $m[0], -1, PREG_SPLIT_NO_EMPTY); | |
| 2429 | + return false === $run ? $m[0] : implode('', array_reverse($run)); | |
| 2430 | + }, | |
| 2431 | + $reversed | |
| 2432 | + ); | |
| 2433 | + return null === $restored ? $reversed : $restored; | |
| 2434 | +} | |
| 2435 | + | |
| 2436 | +/** | |
| 2437 | + * Arabic presentation forms (A + B) -> base letters. Built once from range | |
| 2438 | + * specs rather than ~120 hand-written literal entries; every codepoint in a | |
| 2439 | + * range maps to the same base sequence (isolated/final/initial/medial forms | |
| 2440 | + * of one letter are contiguous in the FE70 block). | |
| 2441 | + */ | |
| 2442 | +private static function pdf_rtl_deshape_map() { | |
| 2443 | + static $map = null; | |
| 2444 | + if (null !== $map) { | |
| 2445 | + return $map; | |
| 2446 | + } | |
| 2447 | + $ranges = array( | |
| 2448 | + // Form B harakat (each pair = standalone + tatweel-joined form). | |
| 2449 | + array(0xFE70, 0xFE71, array(0x064B)), array(0xFE72, 0xFE72, array(0x064C)), | |
| 2450 | + array(0xFE74, 0xFE74, array(0x064D)), array(0xFE76, 0xFE77, array(0x064E)), | |
| 2451 | + array(0xFE78, 0xFE79, array(0x064F)), array(0xFE7A, 0xFE7B, array(0x0650)), | |
| 2452 | + array(0xFE7C, 0xFE7D, array(0x0651)), array(0xFE7E, 0xFE7F, array(0x0652)), | |
| 2453 | + // Form B letters. | |
| 2454 | + array(0xFE80, 0xFE80, array(0x0621)), array(0xFE81, 0xFE82, array(0x0622)), | |
| 2455 | + array(0xFE83, 0xFE84, array(0x0623)), array(0xFE85, 0xFE86, array(0x0624)), | |
| 2456 | + array(0xFE87, 0xFE88, array(0x0625)), array(0xFE89, 0xFE8C, array(0x0626)), | |
| 2457 | + array(0xFE8D, 0xFE8E, array(0x0627)), array(0xFE8F, 0xFE92, array(0x0628)), | |
| 2458 | + array(0xFE93, 0xFE94, array(0x0629)), array(0xFE95, 0xFE98, array(0x062A)), | |
| 2459 | + array(0xFE99, 0xFE9C, array(0x062B)), array(0xFE9D, 0xFEA0, array(0x062C)), | |
| 2460 | + array(0xFEA1, 0xFEA4, array(0x062D)), array(0xFEA5, 0xFEA8, array(0x062E)), | |
| 2461 | + array(0xFEA9, 0xFEAA, array(0x062F)), array(0xFEAB, 0xFEAC, array(0x0630)), | |
| 2462 | + array(0xFEAD, 0xFEAE, array(0x0631)), array(0xFEAF, 0xFEB0, array(0x0632)), | |
| 2463 | + array(0xFEB1, 0xFEB4, array(0x0633)), array(0xFEB5, 0xFEB8, array(0x0634)), | |
| 2464 | + array(0xFEB9, 0xFEBC, array(0x0635)), array(0xFEBD, 0xFEC0, array(0x0636)), | |
| 2465 | + array(0xFEC1, 0xFEC4, array(0x0637)), array(0xFEC5, 0xFEC8, array(0x0638)), | |
| 2466 | + array(0xFEC9, 0xFECC, array(0x0639)), array(0xFECD, 0xFED0, array(0x063A)), | |
| 2467 | + array(0xFED1, 0xFED4, array(0x0641)), array(0xFED5, 0xFED8, array(0x0642)), | |
| 2468 | + array(0xFED9, 0xFEDC, array(0x0643)), array(0xFEDD, 0xFEE0, array(0x0644)), | |
| 2469 | + array(0xFEE1, 0xFEE4, array(0x0645)), array(0xFEE5, 0xFEE8, array(0x0646)), | |
| 2470 | + array(0xFEE9, 0xFEEC, array(0x0647)), array(0xFEED, 0xFEEE, array(0x0648)), | |
| 2471 | + array(0xFEEF, 0xFEF0, array(0x0649)), array(0xFEF1, 0xFEF4, array(0x064A)), | |
| 2472 | + // Form B lam-alef ligatures decompose to two letters. | |
| 2473 | + array(0xFEF5, 0xFEF6, array(0x0644, 0x0622)), array(0xFEF7, 0xFEF8, array(0x0644, 0x0623)), | |
| 2474 | + array(0xFEF9, 0xFEFA, array(0x0644, 0x0625)), array(0xFEFB, 0xFEFC, array(0x0644, 0x0627)), | |
| 2475 | + // Form A: Persian / Urdu letters in common use. | |
| 2476 | + array(0xFB56, 0xFB59, array(0x067E)), array(0xFB66, 0xFB69, array(0x0679)), | |
| 2477 | + array(0xFB7A, 0xFB7D, array(0x0686)), array(0xFB88, 0xFB89, array(0x0688)), | |
| 2478 | + array(0xFB8A, 0xFB8B, array(0x0698)), array(0xFB8E, 0xFB91, array(0x06A9)), | |
| 2479 | + array(0xFB92, 0xFB95, array(0x06AF)), array(0xFBA6, 0xFBA9, array(0x06C1)), | |
| 2480 | + array(0xFBAA, 0xFBAD, array(0x06BE)), array(0xFBAE, 0xFBAF, array(0x06D2)), | |
| 2481 | + array(0xFBFC, 0xFBFF, array(0x06CC)), | |
| 2482 | + ); | |
| 2483 | + $map = array(); | |
| 2484 | + foreach ($ranges as $range) { | |
| 2485 | + $base = ''; | |
| 2486 | + foreach ($range[2] as $cp) { | |
| 2487 | + $base .= self::pdf_rtl_cp_to_utf8($cp); | |
| 2488 | + } | |
| 2489 | + for ($cp = $range[0]; $cp <= $range[1]; $cp++) { | |
| 2490 | + $map[self::pdf_rtl_cp_to_utf8($cp)] = $base; | |
| 2491 | + } | |
| 2492 | + } | |
| 2493 | + return $map; | |
| 2494 | +} | |
| 2495 | + | |
| 2496 | +/** | |
| 2497 | + * Codepoint to UTF-8 without ext-intl / mbstring entity tricks (PHP 7.2 floor). | |
| 2498 | + */ | |
| 2499 | +private static function pdf_rtl_cp_to_utf8($cp) { | |
| 2500 | + if ($cp < 0x80) { | |
| 2501 | + return chr($cp); | |
| 2502 | + } | |
| 2503 | + if ($cp < 0x800) { | |
| 2504 | + return chr(0xC0 | ($cp >> 6)) . chr(0x80 | ($cp & 0x3F)); | |
| 2505 | + } | |
| 2506 | + if ($cp < 0x10000) { | |
| 2507 | + return chr(0xE0 | ($cp >> 12)) . chr(0x80 | (($cp >> 6) & 0x3F)) . chr(0x80 | ($cp & 0x3F)); | |
| 2508 | + } | |
| 2509 | + return chr(0xF0 | ($cp >> 18)) . chr(0x80 | (($cp >> 12) & 0x3F)) . chr(0x80 | (($cp >> 6) & 0x3F)) . chr(0x80 | ($cp & 0x3F)); | |
| 1368 | 2510 | } |
| 1369 | 2511 | } |
