| @@ -5,308 +5,8 @@ | ||
| 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 | -/** | |
| 18 | - * Validate a client-supplied session id (plan-mxchat-20260731-d42bec). | |
| 19 | - * | |
| 20 | - * sanitize_text_field() — which every session_id read site used before this — | |
| 21 | - * preserves '/' and '..'. Harmless where the value is only an option or | |
| 22 | - * transient key suffix, but mxchat_send_delayed_transcript() interpolates it | |
| 23 | - * into a filesystem path, so '../../../../path/x' wrote, emailed and deleted a | |
| 24 | - * file outside the uploads dir. | |
| 25 | - * | |
| 26 | - * REJECTS rather than rewrites: a silently-stripped id would orphan the | |
| 27 | - * conversation it belongs to, which is harder to diagnose than a clean refusal. | |
| 28 | - * Returns '' for anything malformed, so call sites fall into the empty-session | |
| 29 | - * error paths they already have. | |
| 30 | - * | |
| 31 | - * The generator only ever emits 'mxchat_chat_' + 32 hex chars | |
| 32 | - * (class-mxchat-integrator.php, js/chat-script.js), so this is not restrictive | |
| 33 | - * in practice. Length ceiling is deliberate — session ids are also used as | |
| 34 | - * option-name suffixes, and WP option names cap at 191 chars. | |
| 35 | - * | |
| 36 | - * @param mixed $raw Raw request value. | |
| 37 | - * @return string The id if well-formed, '' otherwise. | |
| 38 | - */ | |
| 39 | -public static function sanitize_session_id($raw) { | |
| 40 | - if (!is_scalar($raw)) { | |
| 41 | - return ''; | |
| 42 | - } | |
| 43 | - $val = trim((string) $raw); | |
| 44 | - if ($val === '') { | |
| 45 | - return ''; | |
| 46 | - } | |
| 47 | - return preg_match('/\A[A-Za-z0-9_-]{1,128}\z/', $val) ? $val : ''; | |
| 48 | -} | |
| 49 | - | |
| 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 | - * Most recipients the Notification Email field will accept (plan 2f131a). | |
| 190 | - * A settings field is not a mailing list. | |
| 191 | - */ | |
| 192 | -const NOTIFICATION_EMAIL_MAX = 5; | |
| 193 | - | |
| 194 | -/** | |
| 195 | - * Parse the Notification Email field into a list of recipients (plan 2f131a). | |
| 196 | - * | |
| 197 | - * THE TRAP THIS EXISTS TO CLOSE: sanitize_email() cannot be the validator for | |
| 198 | - * this field, because its output for the failing input is VALID. WordPress | |
| 199 | - * strips the separator and the surplus '@' and concatenates the remains: | |
| 200 | - * | |
| 201 | - * support@acme.com, sales@acme.com -> support@acme.comsalesacme.com | |
| 202 | - * | |
| 203 | - * and is_email() then returns true on that. So every guard in the plugin passed, | |
| 204 | - * the address was stored, the autosave ticked green, and both the new-session | |
| 205 | - * notification and the auto-emailed transcript went to a domain that does not | |
| 206 | - * exist — with no error anywhere. Validating the RAW part BEFORE sanitizing is | |
| 207 | - * the whole point; reversing those two lines silently restores the bug. | |
| 208 | - * | |
| 209 | - * All-or-nothing by design: if any entry is bad the caller must store NOTHING. | |
| 210 | - * A partial accept — keeping the good addresses and dropping the bad one — is | |
| 211 | - * the same defect in a new costume, because the owner still believes everyone | |
| 212 | - * on their list is being notified. | |
| 213 | - * | |
| 214 | - * @param mixed $raw Raw field value, exactly as submitted. | |
| 215 | - * @return array{emails: string[], error: string} Empty emails + empty error | |
| 216 | - * means the field was empty. | |
| 217 | - */ | |
| 218 | -public static function parse_notification_emails($raw) { | |
| 219 | - $out = array('emails' => array(), 'error' => ''); | |
| 220 | - | |
| 221 | - if (!is_scalar($raw)) { | |
| 222 | - $out['error'] = __('The notification email could not be read.', 'mxchat'); | |
| 223 | - return $out; | |
| 224 | - } | |
| 225 | - | |
| 226 | - $raw = trim((string) $raw); | |
| 227 | - if ($raw === '') { | |
| 228 | - return $out; // genuinely empty — the caller falls back to admin_email | |
| 229 | - } | |
| 230 | - | |
| 231 | - $seen = array(); | |
| 232 | - foreach (preg_split('/[,;]/', $raw) as $part) { | |
| 233 | - $part = trim($part); | |
| 234 | - if ($part === '') { | |
| 235 | - // A trailing or doubled separator carries no address, so skipping it | |
| 236 | - // cannot silently drop a recipient. This is the ONLY thing tolerated. | |
| 237 | - continue; | |
| 238 | - } | |
| 239 | - | |
| 240 | - // RAW first. See the note above — order is load-bearing. | |
| 241 | - $clean = is_email($part) ? sanitize_email($part) : ''; | |
| 242 | - if ($clean === '' || !is_email($clean)) { | |
| 243 | - return array( | |
| 244 | - 'emails' => array(), | |
| 245 | - 'error' => sprintf( | |
| 246 | - /* translators: %s: the email address the owner typed. */ | |
| 247 | - __('"%s" is not a valid email address, so nothing was saved. Separate multiple addresses with a comma.', 'mxchat'), | |
| 248 | - esc_html($part) | |
| 249 | - ), | |
| 250 | - ); | |
| 251 | - } | |
| 252 | - | |
| 253 | - $key = strtolower($clean); | |
| 254 | - if (isset($seen[$key])) { | |
| 255 | - continue; // same address twice would simply mail them twice | |
| 256 | - } | |
| 257 | - $seen[$key] = true; | |
| 258 | - $out['emails'][] = $clean; | |
| 259 | - } | |
| 260 | - | |
| 261 | - if (count($out['emails']) > self::NOTIFICATION_EMAIL_MAX) { | |
| 262 | - return array( | |
| 263 | - 'emails' => array(), | |
| 264 | - 'error' => sprintf( | |
| 265 | - /* translators: %d: maximum number of notification recipients. */ | |
| 266 | - __('Enter at most %d email addresses, separated by commas.', 'mxchat'), | |
| 267 | - self::NOTIFICATION_EMAIL_MAX | |
| 268 | - ), | |
| 269 | - ); | |
| 270 | - } | |
| 271 | - | |
| 272 | - return $out; | |
| 273 | -} | |
| 274 | - | |
| 275 | -/** | |
| 276 | - * The stored recipient list, ready to hand to wp_mail() (plan 2f131a). | |
| 277 | - * | |
| 278 | - * Fallback rule, and it is narrow on purpose: an EMPTY field falls back to the | |
| 279 | - * site admin address, because that is the documented behaviour and an owner who | |
| 280 | - * never filled the field in still wants their notifications. A field holding | |
| 281 | - * something unusable does NOT fall back — it sends nowhere, exactly as before | |
| 282 | - * this plan. Falling back on bad input would mean a typo silently redirects a | |
| 283 | - * store's transcripts to a different mailbox than the one on screen. | |
| 284 | - * | |
| 285 | - * @param array|null $options mxchat_transcripts_options, or null to read it. | |
| 286 | - * @return string[] Recipients; empty means do not send. | |
| 287 | - */ | |
| 288 | -public static function notification_recipients($options = null) { | |
| 289 | - if (!is_array($options)) { | |
| 290 | - $options = get_option('mxchat_transcripts_options', array()); | |
| 291 | - if (!is_array($options)) { | |
| 292 | - $options = array(); | |
| 293 | - } | |
| 294 | - } | |
| 295 | - | |
| 296 | - $raw = isset($options['mxchat_notification_email']) ? $options['mxchat_notification_email'] : ''; | |
| 297 | - $raw = is_scalar($raw) ? trim((string) $raw) : ''; | |
| 298 | - | |
| 299 | - if ($raw === '') { | |
| 300 | - $admin = get_option('admin_email'); | |
| 301 | - return is_email($admin) ? array($admin) : array(); | |
| 302 | - } | |
| 303 | - | |
| 304 | - $parsed = self::parse_notification_emails($raw); | |
| 305 | - return $parsed['error'] === '' ? $parsed['emails'] : array(); | |
| 306 | -} | |
| 307 | - | |
| 308 | -/** | |
| 309 | 9 | * Centralized embedding model registry. Single source of truth for dimensions |
| 310 | 10 | * and provider, so model-switch protection logic doesn't drift across files. |
| 311 | 11 | */ |
| 312 | 12 | public static function embedding_model_registry() { |
| @@ -324,12 +24,8 @@ | ||
| 324 | 24 | return isset($registry[$model]) ? (int) $registry[$model]['dims'] : 0; |
| 325 | 25 | } |
| 326 | 26 | |
| 327 | 27 | public static function embedding_model_label($model) { |
| 328 | - if (is_string($model) && strpos($model, 'custom:') === 0) { | |
| 329 | - /* translators: %s: the embedding model name configured on the custom provider */ | |
| 330 | - return sprintf(__('%s (custom provider)', 'mxchat'), substr($model, 7)); | |
| 331 | - } | |
| 332 | 28 | $registry = self::embedding_model_registry(); |
| 333 | 29 | return isset($registry[$model]) ? $registry[$model]['label'] : $model; |
| 334 | 30 | } |
| 335 | 31 | |
| @@ -352,110 +48,8 @@ | ||
| 352 | 48 | } |
| 353 | 49 | } |
| 354 | 50 | |
| 355 | 51 | /** |
| 356 | - * The model name the custom-provider embedding path will send, mirroring the | |
| 357 | - * fallback chain the request itself uses: dedicated custom embedding model, | |
| 358 | - * else the custom chat model, else 'default'. Single source shared by | |
| 359 | - * generate_embedding_custom() and the mismatch-warning "selected" side so the | |
| 360 | - * two can never drift (plan ae02cb). | |
| 361 | - */ | |
| 362 | -public static function resolve_custom_embedding_model($options) { | |
| 363 | - if (isset($options['custom_provider_embedding_model']) && trim((string) $options['custom_provider_embedding_model']) !== '') { | |
| 364 | - return trim((string) $options['custom_provider_embedding_model']); | |
| 365 | - } | |
| 366 | - if (isset($options['custom_provider_model']) && trim((string) $options['custom_provider_model']) !== '') { | |
| 367 | - return trim((string) $options['custom_provider_model']); | |
| 368 | - } | |
| 369 | - return 'default'; | |
| 370 | -} | |
| 371 | - | |
| 372 | -/** | |
| 373 | - * The EFFECTIVE selected embedding model — what the next embed will actually | |
| 374 | - * use. With custom-provider embeddings on this is the custom identity in the | |
| 375 | - * same 'custom:<model>' form stamp_active_embedding_model() records, not the | |
| 376 | - * inert standard dropdown value. Mismatch-warning comparisons must read this, | |
| 377 | - * never $options['embedding_model'] directly — the dropdown cannot be | |
| 378 | - * deselected, so reading it raw flags every correctly-configured custom setup. | |
| 379 | - */ | |
| 380 | -public static function get_selected_embedding_model($options = null) { | |
| 381 | - if (!is_array($options)) { | |
| 382 | - $options = get_option('mxchat_options', array()); | |
| 383 | - } | |
| 384 | - if (isset($options['custom_provider_for_embeddings']) && $options['custom_provider_for_embeddings'] === 'on') { | |
| 385 | - return 'custom:' . self::resolve_custom_embedding_model($options); | |
| 386 | - } | |
| 387 | - return $options['embedding_model'] ?? ''; | |
| 388 | -} | |
| 389 | - | |
| 390 | -/** | |
| 391 | - * Extract the 11-character YouTube video ID from a URL, or '' if the URL is | |
| 392 | - * not a single-video YouTube link. Single source of truth for both the KB | |
| 393 | - * ingestion side and the chat render side — do not duplicate this parsing. | |
| 394 | - * Channel, playlist, and search URLs deliberately return '' (only a URL that | |
| 395 | - * identifies one video can be embedded). | |
| 396 | - */ | |
| 397 | -public static function parse_youtube_id($url) { | |
| 398 | - if (!is_string($url) || $url === '' || !preg_match('#^https?://#i', $url)) { | |
| 399 | - return ''; | |
| 400 | - } | |
| 401 | - $host = strtolower((string) wp_parse_url($url, PHP_URL_HOST)); | |
| 402 | - $host = preg_replace('/^(www|m)\./', '', $host); | |
| 403 | - $path = (string) wp_parse_url($url, PHP_URL_PATH); | |
| 404 | - $id = ''; | |
| 405 | - if ($host === 'youtu.be') { | |
| 406 | - $segments = explode('/', ltrim($path, '/')); | |
| 407 | - $id = $segments[0] ?? ''; | |
| 408 | - } elseif (in_array($host, array('youtube.com', 'youtube-nocookie.com'), true)) { | |
| 409 | - if (preg_match('#^/(?:shorts|embed|live|v)/([A-Za-z0-9_-]+)#', $path, $m)) { | |
| 410 | - $id = $m[1]; | |
| 411 | - } elseif ($path === '/watch') { | |
| 412 | - parse_str((string) wp_parse_url($url, PHP_URL_QUERY), $query_vars); | |
| 413 | - $id = isset($query_vars['v']) ? (string) $query_vars['v'] : ''; | |
| 414 | - } | |
| 415 | - } | |
| 416 | - $id = preg_replace('/[^A-Za-z0-9_-]/', '', (string) $id); | |
| 417 | - return (strlen($id) === 11) ? $id : ''; | |
| 418 | -} | |
| 419 | - | |
| 420 | -/** | |
| 421 | - * plan-mxchat-20260813-f52492 — video-card gating. | |
| 422 | - * | |
| 423 | - * Two standalone options (NOT mxchat_options — they skip the sanitize/autosave | |
| 424 | - * traps entirely), read here so the gate in the integrator and the fields on | |
| 425 | - * Knowledge -> Chunking & Retrieval can never disagree about a default. | |
| 426 | - * | |
| 427 | - * Master switch. Default ON: the card is existing behavior, and this is an | |
| 428 | - * opt-out for owners who never want one, not a new feature to opt into. | |
| 429 | - */ | |
| 430 | -public static function video_embed_enabled() { | |
| 431 | - return get_option('mxchat_video_embed_enabled', 'on') === 'on'; | |
| 432 | -} | |
| 433 | - | |
| 434 | -/** | |
| 435 | - * The video card's OWN confidence floor, as a 0-1 cosine — deliberately not | |
| 436 | - * the site-wide Similarity Threshold (default 35). "Good enough to quote in | |
| 437 | - * the answer" and "good enough to put a video on screen" are different | |
| 438 | - * questions: retrieval is allowed to be generous because the model still | |
| 439 | - * decides what to say, whereas the card is asserted to the visitor with no | |
| 440 | - * such filter. Stored as an int percentage to match the site-wide slider's | |
| 441 | - * convention; the default (55) sits above it on purpose. | |
| 442 | - * | |
| 443 | - * MXCHAT_VIDEO_EMBED_THRESHOLD_DEFAULT is the single source of that number. | |
| 444 | - */ | |
| 445 | -public static function video_embed_threshold() { | |
| 446 | - $stored = get_option('mxchat_video_embed_threshold', null); | |
| 447 | - $percent = ($stored === null || $stored === '') | |
| 448 | - ? MXCHAT_VIDEO_EMBED_THRESHOLD_DEFAULT | |
| 449 | - : (int) $stored; | |
| 450 | - if ($percent < 0) { $percent = 0; } | |
| 451 | - if ($percent > 100) { $percent = 100; } | |
| 452 | - // Cast: PHP evaluates 100/100 to int(1), so an unclamped return type would | |
| 453 | - // vary with the stored value. Callers compare against a cosine — keep it float. | |
| 454 | - return (float) $percent / 100; | |
| 455 | -} | |
| 456 | - | |
| 457 | -/** | |
| 458 | 52 | * UPDATED: Submit or update content (and its embedding) in the database. |
| 459 | 53 | * Stores in Pinecone if enabled, otherwise stores in WordPress DB. |
| 460 | 54 | * |
| 461 | 55 | * @param string $content The content to be embedded. |
| @@ -472,18 +66,10 @@ | ||
| 472 | 66 | |
| 473 | 67 | //error_log('[MXCHAT-DB] Starting database submission for URL: ' . $source_url . ' (Bot: ' . $bot_id . ', Type: ' . $content_type . ')'); |
| 474 | 68 | //error_log('[MXCHAT-DB] Content length: ' . strlen($content) . ' bytes'); |
| 475 | 69 | |
| 476 | - // Sanitize the source URL. Internal identities (mxchat:// manual docs, | |
| 477 | - // upload:// file uploads) are NOT web URLs — esc_url_raw EMPTIES them | |
| 478 | - // because its protocol list is statically cached and effectively | |
| 479 | - // unfilterable (measured, plan 945406 / 0485e5) — sanitize those as text | |
| 480 | - // so upserts stay keyed to a stable identity across re-imports. | |
| 481 | - if (preg_match('#^(mxchat|upload)://#i', $source_url)) { | |
| 482 | - $source_url = sanitize_text_field($source_url); | |
| 483 | - } else { | |
| 484 | - $source_url = esc_url_raw($source_url); | |
| 485 | - } | |
| 70 | + // Sanitize the source URL | |
| 71 | + $source_url = esc_url_raw($source_url); | |
| 486 | 72 | |
| 487 | 73 | // Sanitize content_type |
| 488 | 74 | $content_type = sanitize_key($content_type); |
| 489 | 75 | if (empty($content_type)) { |
| @@ -498,16 +84,9 @@ | ||
| 498 | 84 | // Check if chunking should be applied |
| 499 | 85 | $chunker = MxChat_Chunker::from_settings(); |
| 500 | 86 | if ($chunker->should_chunk($safe_content)) { |
| 501 | 87 | //error_log('[MXCHAT-DB] Content exceeds chunk threshold, using chunked submission'); |
| 502 | - $chunk_result = self::submit_chunked_content($safe_content, $source_url, $api_key, $bot_id, $content_type, $chunker); | |
| 503 | - // Vector Store mirror gets the entry WHOLE — one file per KB entry, | |
| 504 | - // OpenAI chunks server-side; local chunking is our own embedding | |
| 505 | - // concern and never reaches the store (plan 15b5c6). | |
| 506 | - if ($chunk_result === true && class_exists('MxChat_Vectorstore_Manager')) { | |
| 507 | - MxChat_Vectorstore_Manager::sync_upsert_entry($source_url, $safe_content, $bot_id, $content_type); | |
| 508 | - } | |
| 509 | - return $chunk_result; | |
| 88 | + return self::submit_chunked_content($safe_content, $source_url, $api_key, $bot_id, $content_type, $chunker); | |
| 510 | 89 | } |
| 511 | 90 | |
| 512 | 91 | // UPDATED: Generate the embedding using bot-specific configuration |
| 513 | 92 | $embedding_vector = self::generate_embedding($content, $api_key, $bot_id); |
| @@ -512,13 +91,10 @@ | ||
| 512 | 91 | // UPDATED: Generate the embedding using bot-specific configuration |
| 513 | 92 | $embedding_vector = self::generate_embedding($content, $api_key, $bot_id); |
| 514 | 93 | |
| 515 | 94 | if (!is_array($embedding_vector)) { |
| 516 | - // Surface the provider's real reason instead of a fixed string (4a7c0a). | |
| 517 | - $reason = is_wp_error($embedding_vector) | |
| 518 | - ? $embedding_vector->get_error_message() | |
| 519 | - : 'Failed to generate embedding for content'; | |
| 520 | - return new WP_Error('embedding_failed', $reason); | |
| 95 | + //error_log('[MXCHAT-DB] Error: Embedding generation failed'); | |
| 96 | + return new WP_Error('embedding_failed', 'Failed to generate embedding for content'); | |
| 521 | 97 | } |
| 522 | 98 | |
| 523 | 99 | //error_log('[MXCHAT-DB] Embedding generated successfully'); |
| 524 | 100 | |
| @@ -525,30 +101,14 @@ | ||
| 525 | 101 | // UPDATED: Check if Pinecone is enabled for this specific bot |
| 526 | 102 | if (self::is_pinecone_enabled_for_bot($bot_id)) { |
| 527 | 103 | //error_log('[MXCHAT-DB] Pinecone is enabled for bot ' . $bot_id . ' - using Pinecone storage'); |
| 528 | 104 | // Store in Pinecone only |
| 529 | - $result = self::store_in_pinecone_only($embedding_vector, $content, $source_url, $vector_id, $bot_id, $content_type); | |
| 530 | - // If this URL previously stored as CHUNKED content and the new content fits in a | |
| 531 | - // single vector, the upsert above only overwrote the base id — the old | |
| 532 | - // md5(url)_chunk_N vectors would keep serving the stale text. Sweep them. | |
| 533 | - // Only for a real source_url: chunk ids derive from it, and md5('') is shared | |
| 534 | - // by legacy URL-less entries so a blind sweep there could hit other entries. | |
| 535 | - if ($result === true && !empty($source_url)) { | |
| 536 | - self::cleanup_pinecone_chunk_stragglers($source_url, $bot_id); | |
| 537 | - } | |
| 538 | - if ($result === true && class_exists('MxChat_Vectorstore_Manager')) { | |
| 539 | - MxChat_Vectorstore_Manager::sync_upsert_entry($source_url, $safe_content, $bot_id, $content_type); | |
| 540 | - } | |
| 541 | - return $result; | |
| 105 | + return self::store_in_pinecone_only($embedding_vector, $content, $source_url, $vector_id, $bot_id, $content_type); | |
| 542 | 106 | } else { |
| 543 | 107 | //error_log('[MXCHAT-DB] Pinecone not enabled for bot ' . $bot_id . ' - using WordPress storage'); |
| 544 | 108 | // Store in WordPress database only |
| 545 | 109 | $embedding_vector_serialized = maybe_serialize($embedding_vector); |
| 546 | - $result = self::store_in_wordpress_db($safe_content, $source_url, $embedding_vector_serialized, $table_name, $content_type); | |
| 547 | - if ($result === true && class_exists('MxChat_Vectorstore_Manager')) { | |
| 548 | - MxChat_Vectorstore_Manager::sync_upsert_entry($source_url, $safe_content, $bot_id, $content_type); | |
| 549 | - } | |
| 550 | - return $result; | |
| 110 | + return self::store_in_wordpress_db($safe_content, $source_url, $embedding_vector_serialized, $table_name, $content_type); | |
| 551 | 111 | } |
| 552 | 112 | } |
| 553 | 113 | |
| 554 | 114 | /** |
| @@ -664,15 +224,11 @@ | ||
| 664 | 224 | // Check if this is truly manual content (no URL at all) vs a real URL that filter_var rejects |
| 665 | 225 | // filter_var(FILTER_VALIDATE_URL) rejects valid URLs with encoded chars, non-ASCII, fragments, etc. |
| 666 | 226 | // Use a looser check: if it starts with http(s):// or has a scheme, it's a URL |
| 667 | 227 | $has_url_scheme = !empty($source_url) && preg_match('#^https?://#i', $source_url); |
| 668 | - // upload:// identities (admin document/PDF uploads, plan 0485e5) are stable | |
| 669 | - // and deduplicable — treat them like URLs so a re-upload UPDATES the row | |
| 670 | - // instead of minting a fresh manual identity (which would duplicate). | |
| 671 | - $has_stable_identity = $has_url_scheme || (!empty($source_url) && preg_match('#^upload://#i', $source_url)); | |
| 672 | 228 | // Treat legacy mxchat.ai source URLs as manual — old bug assigned the site URL to manual entries |
| 673 | 229 | $is_legacy_mxchat_url = $has_url_scheme && strpos($source_url, 'mxchat.ai') !== false; |
| 674 | - $is_manual_content = empty($source_url) || $source_url === '' || !$has_stable_identity || $is_legacy_mxchat_url; | |
| 230 | + $is_manual_content = empty($source_url) || $source_url === '' || !$has_url_scheme || $is_legacy_mxchat_url; | |
| 675 | 231 | |
| 676 | 232 | if ($is_manual_content) { |
| 677 | 233 | // Generate unique identifier for manual content to prevent overwrites |
| 678 | 234 | $source_url = 'mxchat://manual-content/' . time() . '-' . wp_generate_password(8, false); |
| @@ -796,11 +352,10 @@ | ||
| 796 | 352 | // ===== UPDATED: Handle manual content with unique vector IDs ===== |
| 797 | 353 | if ($vector_id) { |
| 798 | 354 | // Use provided vector ID |
| 799 | 355 | //error_log('[MXCHAT-PINECONE-MAIN] Using provided vector ID: ' . $vector_id); |
| 800 | - } elseif (!empty($url) && preg_match('#^(https?|upload)://#i', $url)) { | |
| 801 | - // For URLs — and stable upload:// identities (plan 0485e5) — use an | |
| 802 | - // identity-derived ID so a re-import upserts the same vector. | |
| 356 | + } elseif (!empty($url) && preg_match('#^https?://#i', $url)) { | |
| 357 | + // For URLs, use URL-based ID (existing behavior) | |
| 803 | 358 | $vector_id = md5($url); |
| 804 | 359 | //error_log('[MXCHAT-PINECONE-MAIN] Generated vector ID from URL: ' . $vector_id); |
| 805 | 360 | } else { |
| 806 | 361 | // For manual content (empty/no URL scheme), generate unique ID |
| @@ -918,104 +473,8 @@ | ||
| 918 | 473 | return true; |
| 919 | 474 | } |
| 920 | 475 | |
| 921 | 476 | /** |
| 922 | - * Caller-side pre-flight for KB ingestion: can an embedding request be made | |
| 923 | - * with these options, and which API key should travel downstream? | |
| 924 | - * | |
| 925 | - * Custom-provider-aware — generate_embedding() below routes to the custom | |
| 926 | - * endpoint FIRST and ignores the passed cloud key entirely when | |
| 927 | - * custom_provider_for_embeddings is on, so on that branch the only real | |
| 928 | - * requirement is a Base URL. Ingestion callers that gated on a cloud API key | |
| 929 | - * were killing keyless custom-embeddings sites (local Ollama / LM Studio | |
| 930 | - * class) before the embed layer could route (plan cbd5fd). | |
| 931 | - * | |
| 932 | - * NOTE: reads $options['embedding_model'] raw on purpose — this mirrors | |
| 933 | - * generate_embedding()'s own routing read, NOT the mismatch-banner's | |
| 934 | - * "selected" chain (get_selected_embedding_model). The helper must predict | |
| 935 | - * what the very next embed call will do, byte-for-byte. | |
| 936 | - * | |
| 937 | - * Decision only — callers keep their own error-surfacing shape (admin-notice | |
| 938 | - * transient + redirect, wp_send_json_error, WP_Error, silent return). | |
| 939 | - * | |
| 940 | - * @param array|null $options Resolved options (bot-specific where the caller | |
| 941 | - * has them); null loads the default bot's options. | |
| 942 | - * @return array { | |
| 943 | - * @type bool $ok Whether ingestion can proceed. | |
| 944 | - * @type string $api_key Key to pass downstream ('' on the custom branch — | |
| 945 | - * generate_embedding() ignores it there). | |
| 946 | - * @type string $reason Human-readable blocker; '' when $ok. | |
| 947 | - * @type string $provider Short provider label ('OpenAI', 'Voyage AI', | |
| 948 | - * 'Google Gemini', 'Custom Provider'). | |
| 949 | - * } | |
| 950 | - */ | |
| 951 | -public static function embedding_preflight($options = null) { | |
| 952 | - if (!is_array($options)) { | |
| 953 | - $options = get_option('mxchat_options'); | |
| 954 | - $options = is_array($options) ? $options : array(); | |
| 955 | - } | |
| 956 | - | |
| 957 | - // Custom branch mirrors generate_embedding()'s routing order (custom first). | |
| 958 | - if (isset($options['custom_provider_for_embeddings']) && $options['custom_provider_for_embeddings'] === 'on') { | |
| 959 | - $base_url = isset($options['custom_provider_base_url']) ? rtrim(trim((string) $options['custom_provider_base_url']), '/') : ''; | |
| 960 | - if ($base_url === '') { | |
| 961 | - return array( | |
| 962 | - 'ok' => false, | |
| 963 | - 'api_key' => '', | |
| 964 | - // Same string generate_embedding_custom() returns for this state. | |
| 965 | - 'reason' => __('Custom provider Base URL is not configured.', 'mxchat'), | |
| 966 | - 'provider' => 'Custom Provider', | |
| 967 | - ); | |
| 968 | - } | |
| 969 | - return array('ok' => true, 'api_key' => '', 'reason' => '', 'provider' => 'Custom Provider'); | |
| 970 | - } | |
| 971 | - | |
| 972 | - $selected_model = $options['embedding_model'] ?? 'text-embedding-ada-002'; | |
| 973 | - if (strpos($selected_model, 'voyage') === 0) { | |
| 974 | - $api_key = $options['voyage_api_key'] ?? ''; | |
| 975 | - $provider = 'Voyage AI'; | |
| 976 | - } elseif (strpos($selected_model, 'gemini-embedding') === 0) { | |
| 977 | - $api_key = $options['gemini_api_key'] ?? ''; | |
| 978 | - $provider = 'Google Gemini'; | |
| 979 | - } else { | |
| 980 | - $api_key = $options['api_key'] ?? ''; | |
| 981 | - $provider = 'OpenAI'; | |
| 982 | - } | |
| 983 | - | |
| 984 | - if (empty($api_key)) { | |
| 985 | - return array( | |
| 986 | - 'ok' => false, | |
| 987 | - 'api_key' => '', | |
| 988 | - 'reason' => sprintf( | |
| 989 | - /* translators: %s: embedding provider name */ | |
| 990 | - __('%s API key is not configured. Please add your API key in the settings before submitting content.', 'mxchat'), | |
| 991 | - $provider | |
| 992 | - ), | |
| 993 | - 'provider' => $provider, | |
| 994 | - ); | |
| 995 | - } | |
| 996 | - | |
| 997 | - return array('ok' => true, 'api_key' => $api_key, 'reason' => '', 'provider' => $provider); | |
| 998 | -} | |
| 999 | - | |
| 1000 | -/** | |
| 1001 | - * Public QUERY-side entry point (plan 876edb). The chat pipeline's | |
| 1002 | - * MxChat_Integrator::mxchat_generate_embedding() adapter routes through here | |
| 1003 | - * so the query and index sides share ONE provider-routing implementation — | |
| 1004 | - * the same endpoints, request bodies, and stamping semantics. The Integrator | |
| 1005 | - * keeps its own error vocabulary by translating the WP_Error this returns | |
| 1006 | - * (see the structured error data on every failure path below). | |
| 1007 | - * | |
| 1008 | - * @param string $text The text to be embedded. | |
| 1009 | - * @param string $api_key Caller-resolved API key (per-bot on the query side). | |
| 1010 | - * @param string $bot_id The bot ID for multi-bot support. | |
| 1011 | - * @return array|WP_Error The embedding vector, or WP_Error carrying the reason. | |
| 1012 | - */ | |
| 1013 | -public static function generate_query_embedding($text, $api_key, $bot_id = 'default') { | |
| 1014 | - return self::generate_embedding($text, $api_key, $bot_id); | |
| 1015 | -} | |
| 1016 | - | |
| 1017 | -/** | |
| 1018 | 477 | * UPDATED: Generate an embedding for the given text using bot-specific configuration. |
| 1019 | 478 | * |
| 1020 | 479 | * @param string $text The text to be embedded. |
| 1021 | 480 | * @param string $api_key The API key used for generating embeddings. |
| @@ -1036,13 +495,9 @@ | ||
| 1036 | 495 | // vectors and query vectors come from the same model. Default-off behavior |
| 1037 | 496 | // below is untouched. |
| 1038 | 497 | if (isset($options['custom_provider_for_embeddings']) && $options['custom_provider_for_embeddings'] === 'on') { |
| 1039 | 498 | $custom = self::generate_embedding_custom($text, $options); |
| 1040 | - // The custom path already returns a human-readable error string — | |
| 1041 | - // carry it instead of collapsing to null (plan 4a7c0a). The 'custom' | |
| 1042 | - // branch marker lets the Integrator adapter map the string back onto | |
| 1043 | - // its own error codes (876edb). | |
| 1044 | - return is_array($custom) ? $custom : new WP_Error('embedding_failed', (string) $custom, array('branch' => 'custom')); | |
| 499 | + return is_array($custom) ? $custom : null; | |
| 1045 | 500 | } |
| 1046 | 501 | |
| 1047 | 502 | $selected_model = $options['embedding_model'] ?? 'text-embedding-ada-002'; |
| 1048 | 503 | |
| @@ -1054,13 +509,10 @@ | ||
| 1054 | 509 | $endpoint = 'https://generativelanguage.googleapis.com/v1beta/models/' . $selected_model . ':embedContent'; |
| 1055 | 510 | $api_key = $options['gemini_api_key'] ?? ''; |
| 1056 | 511 | } else { |
| 1057 | 512 | $endpoint = 'https://api.openai.com/v1/embeddings'; |
| 1058 | - // Prefer the caller-resolved key when one was passed — the query side | |
| 1059 | - // resolves per-bot keys at its call sites (integrator adapter, 876edb). | |
| 1060 | - // Index callers pass the preflight key, which equals this options read, | |
| 1061 | - // so nothing changes for them. | |
| 1062 | - $api_key = !empty($api_key) ? $api_key : ($options['api_key'] ?? ''); | |
| 513 | + // Use the bot-specific API key or fallback to passed API key | |
| 514 | + $api_key = $options['api_key'] ?? $api_key; | |
| 1063 | 515 | } |
| 1064 | 516 | |
| 1065 | 517 | // Prepare request body based on provider |
| 1066 | 518 | if (strpos($selected_model, 'gemini-embedding') === 0) { |
| @@ -1111,22 +563,14 @@ | ||
| 1111 | 563 | |
| 1112 | 564 | $response = wp_remote_post($endpoint, $args); |
| 1113 | 565 | |
| 1114 | 566 | if (is_wp_error($response)) { |
| 1115 | - $message = 'Embedding request failed (connection): ' . $response->get_error_message(); | |
| 1116 | - if (class_exists('MxChat_Admin')) { | |
| 1117 | - MxChat_Admin::mxchat_log_debug('embedding_error', $message, array('model' => $selected_model, 'bot_id' => $bot_id)); | |
| 1118 | - } | |
| 1119 | - return new WP_Error('embedding_failed', $message, array( | |
| 1120 | - 'branch' => 'cloud', | |
| 1121 | - 'kind' => 'connection', | |
| 1122 | - 'reason' => $response->get_error_message(), | |
| 1123 | - 'model' => $selected_model, | |
| 1124 | - )); | |
| 567 | + //error_log('Error generating embedding for bot ' . $bot_id . ': ' . $response->get_error_message()); | |
| 568 | + return null; | |
| 1125 | 569 | } |
| 1126 | - | |
| 570 | + | |
| 1127 | 571 | $response_body = json_decode(wp_remote_retrieve_body($response), true); |
| 1128 | - | |
| 572 | + | |
| 1129 | 573 | // Handle different response formats based on provider |
| 1130 | 574 | if (strpos($selected_model, 'gemini-embedding') === 0) { |
| 1131 | 575 | // Gemini API response format |
| 1132 | 576 | if (isset($response_body['embedding']['values']) && is_array($response_body['embedding']['values'])) { |
| @@ -1132,9 +576,10 @@ | ||
| 1132 | 576 | if (isset($response_body['embedding']['values']) && is_array($response_body['embedding']['values'])) { |
| 1133 | 577 | self::stamp_active_embedding_model($selected_model); |
| 1134 | 578 | return $response_body['embedding']['values']; |
| 1135 | 579 | } else { |
| 1136 | - return self::embedding_failure_error($response, $selected_model, $api_key, $bot_id); | |
| 580 | + //error_log('Invalid response received from Gemini embedding API for bot ' . $bot_id . ': ' . wp_json_encode($response_body)); | |
| 581 | + return null; | |
| 1137 | 582 | } |
| 1138 | 583 | } else { |
| 1139 | 584 | // OpenAI/Voyage API response format |
| 1140 | 585 | if (isset($response_body['data'][0]['embedding']) && is_array($response_body['data'][0]['embedding'])) { |
| @@ -1140,68 +585,15 @@ | ||
| 1140 | 585 | if (isset($response_body['data'][0]['embedding']) && is_array($response_body['data'][0]['embedding'])) { |
| 1141 | 586 | self::stamp_active_embedding_model($selected_model); |
| 1142 | 587 | return $response_body['data'][0]['embedding']; |
| 1143 | 588 | } else { |
| 1144 | - return self::embedding_failure_error($response, $selected_model, $api_key, $bot_id); | |
| 589 | + //error_log('Invalid response received from embedding API for bot ' . $bot_id . ': ' . wp_json_encode($response_body)); | |
| 590 | + return null; | |
| 1145 | 591 | } |
| 1146 | 592 | } |
| 1147 | 593 | } |
| 1148 | 594 | |
| 1149 | 595 | /** |
| 1150 | - * Build a WP_Error carrying the embedding provider's REAL failure reason, | |
| 1151 | - * and record it in the Debug Mode log. Previously every failure path | |
| 1152 | - * returned bare null, so customers saw only "Failed to generate embedding | |
| 1153 | - * for content" / "Failed to store any chunks" with no cause (plan 4a7c0a). | |
| 1154 | - * | |
| 1155 | - * The API key never appears in provider response bodies (it travels in the | |
| 1156 | - * request headers), but the reason is scrubbed for it anyway before it can | |
| 1157 | - * reach a notice or the debug log. | |
| 1158 | - */ | |
| 1159 | -private static function embedding_failure_error($response, $selected_model, $api_key, $bot_id) { | |
| 1160 | - $status = (int) wp_remote_retrieve_response_code($response); | |
| 1161 | - $raw = (string) wp_remote_retrieve_body($response); | |
| 1162 | - $decoded = json_decode($raw, true); | |
| 1163 | - | |
| 1164 | - // Provider error shapes: OpenAI + Gemini use {"error":{"message":…}}; | |
| 1165 | - // Voyage uses {"detail":…}. | |
| 1166 | - $reason = ''; | |
| 1167 | - if (is_array($decoded)) { | |
| 1168 | - if (isset($decoded['error']['message']) && is_string($decoded['error']['message'])) { | |
| 1169 | - $reason = $decoded['error']['message']; | |
| 1170 | - } elseif (isset($decoded['detail']) && is_string($decoded['detail'])) { | |
| 1171 | - $reason = $decoded['detail']; | |
| 1172 | - } | |
| 1173 | - } | |
| 1174 | - if ($reason === '') { | |
| 1175 | - $reason = ($raw !== '') ? substr($raw, 0, 200) : 'empty or malformed response'; | |
| 1176 | - } | |
| 1177 | - if (is_string($api_key) && $api_key !== '') { | |
| 1178 | - $reason = str_replace($api_key, '[redacted]', $reason); | |
| 1179 | - } | |
| 1180 | - $reason = substr($reason, 0, 300); | |
| 1181 | - $message = sprintf('Embedding failed (%s, HTTP %d): %s', $selected_model, $status, $reason); | |
| 1182 | - | |
| 1183 | - if (class_exists('MxChat_Admin')) { | |
| 1184 | - MxChat_Admin::mxchat_log_debug('embedding_error', $message, array( | |
| 1185 | - 'model' => $selected_model, | |
| 1186 | - 'status' => $status, | |
| 1187 | - 'bot_id' => $bot_id, | |
| 1188 | - )); | |
| 1189 | - } | |
| 1190 | - | |
| 1191 | - // Structured data so the Integrator's query-side adapter can rebuild its | |
| 1192 | - // typed error contract (auth/rate-limit/quota/invalid-response) without a | |
| 1193 | - // second transport implementation (876edb). Additive — message unchanged. | |
| 1194 | - return new WP_Error('embedding_failed', $message, array( | |
| 1195 | - 'branch' => 'cloud', | |
| 1196 | - 'status' => $status, | |
| 1197 | - 'error_type' => (is_array($decoded) && isset($decoded['error']['type']) && is_string($decoded['error']['type'])) ? $decoded['error']['type'] : '', | |
| 1198 | - 'reason' => $reason, | |
| 1199 | - 'model' => $selected_model, | |
| 1200 | - )); | |
| 1201 | -} | |
| 1202 | - | |
| 1203 | -/** | |
| 1204 | 596 | * Generate an embedding via a Custom (OpenAI-compatible) provider's /embeddings route. |
| 1205 | 597 | * Shared by every embedding entry point so the KNOWLEDGE-BASE INDEX side and the |
| 1206 | 598 | * QUERY side route through the same model when the opt-in |
| 1207 | 599 | * 'custom_provider_for_embeddings' setting is on. Mirrors the query-path logic in |
| @@ -1229,11 +621,12 @@ | ||
| 1229 | 621 | $api_key = isset($options['custom_provider_api_key']) ? trim((string) $options['custom_provider_api_key']) : ''; |
| 1230 | 622 | $auth_scheme = isset($options['custom_provider_auth_scheme']) ? $options['custom_provider_auth_scheme'] : 'bearer'; |
| 1231 | 623 | $api_version = isset($options['custom_provider_api_version']) ? trim((string) $options['custom_provider_api_version']) : ''; |
| 1232 | 624 | |
| 1233 | - // Embedding model: shared resolver (dedicated embedding model -> chat model | |
| 1234 | - // -> 'default') — the mismatch warning's "selected" side reads the same chain. | |
| 1235 | - $model = self::resolve_custom_embedding_model($options); | |
| 625 | + // Embedding model: prefer the dedicated custom_provider_embedding_model, fall back to the chat model. | |
| 626 | + $model = (isset($options['custom_provider_embedding_model']) && trim((string) $options['custom_provider_embedding_model']) !== '') | |
| 627 | + ? trim((string) $options['custom_provider_embedding_model']) | |
| 628 | + : ((isset($options['custom_provider_model']) && trim((string) $options['custom_provider_model']) !== '') ? trim((string) $options['custom_provider_model']) : 'default'); | |
| 1236 | 629 | |
| 1237 | 630 | $embed_url = $base_url . '/embeddings'; |
| 1238 | 631 | if (!empty($api_version)) { |
| 1239 | 632 | $embed_url .= (strpos($embed_url, '?') === false ? '?' : '&') . 'api-version=' . rawurlencode($api_version); |
| @@ -1253,13 +646,9 @@ | ||
| 1253 | 646 | 'body' => wp_json_encode(['input' => $text, 'model' => $model]), |
| 1254 | 647 | 'timeout' => 60, |
| 1255 | 648 | ]); |
| 1256 | 649 | if (is_wp_error($response)) { |
| 1257 | - return self::log_custom_embedding_failure( | |
| 1258 | - 'Connection error when generating embeddings (custom provider): ' . $response->get_error_message(), | |
| 1259 | - $model, | |
| 1260 | - $api_key | |
| 1261 | - ); | |
| 650 | + return 'Connection error when generating embeddings (custom provider): ' . $response->get_error_message(); | |
| 1262 | 651 | } |
| 1263 | 652 | |
| 1264 | 653 | $status = wp_remote_retrieve_response_code($response); |
| 1265 | 654 | $body = json_decode(wp_remote_retrieve_body($response), true); |
| @@ -1264,14 +653,9 @@ | ||
| 1264 | 653 | $status = wp_remote_retrieve_response_code($response); |
| 1265 | 654 | $body = json_decode(wp_remote_retrieve_body($response), true); |
| 1266 | 655 | if ($status !== 200) { |
| 1267 | 656 | $msg = isset($body['error']['message']) ? $body['error']['message'] : 'HTTP ' . $status; |
| 1268 | - return self::log_custom_embedding_failure( | |
| 1269 | - 'Custom embedding endpoint error: ' . $msg, | |
| 1270 | - $model, | |
| 1271 | - $api_key, | |
| 1272 | - (int) $status | |
| 1273 | - ); | |
| 657 | + return 'Custom embedding endpoint error: ' . $msg; | |
| 1274 | 658 | } |
| 1275 | 659 | if (isset($body['data'][0]['embedding']) && is_array($body['data'][0]['embedding'])) { |
| 1276 | 660 | // Stamp the custom model identity so the active-embedding-model mismatch |
| 1277 | 661 | // warning reflects the real (custom) model rather than the built-in setting. |
| @@ -1277,42 +661,12 @@ | ||
| 1277 | 661 | // warning reflects the real (custom) model rather than the built-in setting. |
| 1278 | 662 | self::stamp_active_embedding_model('custom:' . $model); |
| 1279 | 663 | return $body['data'][0]['embedding']; |
| 1280 | 664 | } |
| 1281 | - return self::log_custom_embedding_failure('Invalid embedding response from custom provider.', $model, $api_key); | |
| 665 | + return 'Invalid embedding response from custom provider.'; | |
| 1282 | 666 | } |
| 1283 | 667 | |
| 1284 | 668 | /** |
| 1285 | - * Record a custom-provider embedding failure in the Debug Mode log, then | |
| 1286 | - * return the message unchanged so callers keep their string-error contract. | |
| 1287 | - * The cloud branch has logged its failures since 4a7c0a; the custom branch | |
| 1288 | - * never did, so chat-side failures on Custom-provider installs were | |
| 1289 | - * invisible to Debug Mode despite the 3.2.18 readme saying otherwise | |
| 1290 | - * (plan 71e4b6). Same scrub-then-log shape as embedding_failure_error(). | |
| 1291 | - * | |
| 1292 | - * @param string $message Human-readable failure (the caller's return value). | |
| 1293 | - * @param string $model Resolved custom embedding model. | |
| 1294 | - * @param string $api_key Scrubbed out of the logged message if it ever appears. | |
| 1295 | - * @param int $status HTTP status when one was received, 0 otherwise. | |
| 1296 | - * @return string The (scrubbed) message. | |
| 1297 | - */ | |
| 1298 | -private static function log_custom_embedding_failure($message, $model, $api_key, $status = 0) { | |
| 1299 | - if (is_string($api_key) && $api_key !== '') { | |
| 1300 | - $message = str_replace($api_key, '[redacted]', $message); | |
| 1301 | - } | |
| 1302 | - | |
| 1303 | - if (class_exists('MxChat_Admin')) { | |
| 1304 | - $context = array('model' => 'custom:' . $model); | |
| 1305 | - if ($status > 0) { | |
| 1306 | - $context['status'] = $status; | |
| 1307 | - } | |
| 1308 | - MxChat_Admin::mxchat_log_debug('embedding_error', $message, $context); | |
| 1309 | - } | |
| 1310 | - | |
| 1311 | - return $message; | |
| 1312 | -} | |
| 1313 | - | |
| 1314 | -/** | |
| 1315 | 669 | * Submit content as multiple chunks |
| 1316 | 670 | * |
| 1317 | 671 | * Splits large content into chunks, generates embeddings for each, |
| 1318 | 672 | * and stores them with chunk metadata for later reassembly. |
| @@ -1331,26 +685,10 @@ | ||
| 1331 | 685 | |
| 1332 | 686 | //error_log('[MXCHAT-CHUNK-DEBUG] Starting chunked submission for: ' . $source_url); |
| 1333 | 687 | //error_log('[MXCHAT-CHUNK-DEBUG] Content length: ' . strlen($content) . ' chars'); |
| 1334 | 688 | |
| 1335 | - // URL-less (manual) content needs a minted identity BEFORE chunk ids are derived: | |
| 1336 | - // every chunk id is md5(source_url)_chunk_N, so with source_url = '' EVERY long | |
| 1337 | - // manual document shared the md5('') prefix — and the clean-slate delete below | |
| 1338 | - // wiped the PREVIOUS manual entry's chunks each time a new one was added. The | |
| 1339 | - // single-vector paths already mint (mxchat:// in WP, manual_* in Pinecone); this | |
| 1340 | - // was the one storage path that didn't. Keep an mxchat:// identity if the caller | |
| 1341 | - // already carries one. | |
| 1342 | - if (strpos($source_url, 'mxchat://') !== 0 && strpos($source_url, 'upload://') !== 0 && !preg_match('#^https?://#i', $source_url)) { | |
| 1343 | - $source_url = 'mxchat://manual-content/' . time() . '-' . wp_generate_password(8, false); | |
| 1344 | - } | |
| 1345 | - | |
| 1346 | - // First, delete any existing chunks for this URL (clean slate). | |
| 1347 | - // Mirror-suspended: this is a re-store, not an entry removal — the Vector | |
| 1348 | - // Store file is replaced (or kept, on hash match) by the caller's | |
| 1349 | - // sync_upsert_entry after storage succeeds. | |
| 1350 | - self::$vectorstore_mirror_suspended = true; | |
| 689 | + // First, delete any existing chunks for this URL (clean slate) | |
| 1351 | 690 | $delete_result = self::delete_chunks_for_url($source_url, $bot_id); |
| 1352 | - self::$vectorstore_mirror_suspended = false; | |
| 1353 | 691 | if (is_wp_error($delete_result)) { |
| 1354 | 692 | //error_log('[MXCHAT-CHUNK-DEBUG] Warning: Failed to delete existing chunks: ' . $delete_result->get_error_message()); |
| 1355 | 693 | // Continue anyway - we'll overwrite with upsert |
| 1356 | 694 | } |
| @@ -1370,11 +708,8 @@ | ||
| 1370 | 708 | return new WP_Error('chunking_failed', 'Content could not be split into chunks'); |
| 1371 | 709 | } |
| 1372 | 710 | |
| 1373 | 711 | $errors = array(); |
| 1374 | - $embed_failures = 0; | |
| 1375 | - $first_embed_reason = ''; | |
| 1376 | - $first_store_reason = ''; | |
| 1377 | 712 | $is_pinecone = self::is_pinecone_enabled_for_bot($bot_id); |
| 1378 | 713 | |
| 1379 | 714 | foreach ($chunks as $index => $chunk_text) { |
| 1380 | 715 | // Generate chunk metadata |
| @@ -1414,17 +749,10 @@ | ||
| 1414 | 749 | // Generate embedding for this chunk |
| 1415 | 750 | $embedding_vector = self::generate_embedding($chunk_text, $api_key, $bot_id); |
| 1416 | 751 | |
| 1417 | 752 | if (!is_array($embedding_vector)) { |
| 1418 | - // Track embedding failures separately from storage failures, and | |
| 1419 | - // keep the first provider reason seen — the two failure classes | |
| 1420 | - // have opposite remedies (API key vs Pinecone/DB) (plan 4a7c0a). | |
| 1421 | - $embed_failures++; | |
| 1422 | - $reason = is_wp_error($embedding_vector) ? $embedding_vector->get_error_message() : ''; | |
| 1423 | - if ($reason !== '' && $first_embed_reason === '') { | |
| 1424 | - $first_embed_reason = $reason; | |
| 1425 | - } | |
| 1426 | - $errors[] = new WP_Error('embedding_failed', 'Failed to generate embedding for chunk ' . $index . ($reason !== '' ? ' — ' . $reason : '')); | |
| 753 | + $errors[] = new WP_Error('embedding_failed', 'Failed to generate embedding for chunk ' . $index); | |
| 754 | + //error_log('[MXCHAT-CHUNK] Failed to generate embedding for chunk ' . $index); | |
| 1427 | 755 | continue; |
| 1428 | 756 | } |
| 1429 | 757 | |
| 1430 | 758 | if ($is_pinecone) { |
| @@ -1454,44 +782,20 @@ | ||
| 1454 | 782 | } |
| 1455 | 783 | |
| 1456 | 784 | if (is_wp_error($result)) { |
| 1457 | 785 | $errors[] = $result; |
| 1458 | - if ($first_store_reason === '') { | |
| 1459 | - $first_store_reason = $result->get_error_message(); | |
| 1460 | - } | |
| 786 | + //error_log('[MXCHAT-CHUNK] Failed to store chunk ' . $index . ': ' . $result->get_error_message()); | |
| 1461 | 787 | } |
| 1462 | 788 | } |
| 1463 | 789 | |
| 1464 | 790 | if (count($errors) === $total_chunks) { |
| 1465 | - // Say WHICH stage failed — "failed to store" used to cover pure | |
| 1466 | - // embedding failures too, sending customers to debug Pinecone when | |
| 1467 | - // the problem was their embedding API key (plan 4a7c0a). | |
| 1468 | - if ($embed_failures === $total_chunks) { | |
| 1469 | - return new WP_Error('chunking_failed', | |
| 1470 | - 'Failed to store any chunks — every chunk failed to embed' | |
| 1471 | - . ($first_embed_reason !== '' ? ': ' . $first_embed_reason : '') | |
| 1472 | - . ' Check the embedding provider API key and model under MxChat Settings.'); | |
| 1473 | - } | |
| 1474 | - if ($embed_failures === 0) { | |
| 1475 | - return new WP_Error('chunking_failed', | |
| 1476 | - 'Failed to store any chunks — embeddings generated but storage failed' | |
| 1477 | - . ($first_store_reason !== '' ? ': ' . $first_store_reason : '') | |
| 1478 | - . ' Check the knowledge base storage (Pinecone index or database).'); | |
| 1479 | - } | |
| 1480 | - return new WP_Error('chunking_failed', sprintf( | |
| 1481 | - 'Failed to store any chunks — %d failed to embed%s and %d failed to store%s', | |
| 1482 | - $embed_failures, | |
| 1483 | - $first_embed_reason !== '' ? ' (' . $first_embed_reason . ')' : '', | |
| 1484 | - $total_chunks - $embed_failures, | |
| 1485 | - $first_store_reason !== '' ? ' (' . $first_store_reason . ')' : '' | |
| 1486 | - )); | |
| 791 | + return new WP_Error('chunking_failed', 'Failed to store any chunks'); | |
| 1487 | 792 | } |
| 1488 | 793 | |
| 1489 | 794 | if (!empty($errors)) { |
| 1490 | - $detail = $first_embed_reason !== '' ? $first_embed_reason : $first_store_reason; | |
| 795 | + //error_log('[MXCHAT-CHUNK] Completed with ' . count($errors) . ' errors out of ' . $total_chunks . ' chunks'); | |
| 1491 | 796 | return new WP_Error('chunking_partial_failure', |
| 1492 | - sprintf('Failed to store %d of %d chunks', count($errors), $total_chunks) | |
| 1493 | - . ($detail !== '' ? ' — first error: ' . $detail : '')); | |
| 797 | + sprintf('Failed to store %d of %d chunks', count($errors), $total_chunks)); | |
| 1494 | 798 | } |
| 1495 | 799 | |
| 1496 | 800 | //error_log('[MXCHAT-CHUNK] Successfully stored all ' . $total_chunks . ' chunks'); |
| 1497 | 801 | return true; |
| @@ -1614,14 +918,8 @@ | ||
| 1614 | 918 | */ |
| 1615 | 919 | public static function delete_chunks_for_url($source_url, $bot_id = 'default') { |
| 1616 | 920 | //error_log('[MXCHAT-CHUNK-DELETE] Deleting chunks for URL: ' . $source_url); |
| 1617 | 921 | |
| 1618 | - // Entry removal — mirror it to the Vector Store (unless a storage routine | |
| 1619 | - // is mid-re-store, see $vectorstore_mirror_suspended). | |
| 1620 | - if (!self::$vectorstore_mirror_suspended && class_exists('MxChat_Vectorstore_Manager')) { | |
| 1621 | - MxChat_Vectorstore_Manager::sync_delete_entry($source_url, $bot_id); | |
| 1622 | - } | |
| 1623 | - | |
| 1624 | 922 | if (self::is_pinecone_enabled_for_bot($bot_id)) { |
| 1625 | 923 | return self::delete_pinecone_chunks_by_url($source_url, $bot_id); |
| 1626 | 924 | } else { |
| 1627 | 925 | return self::delete_wordpress_chunks_by_url($source_url); |
| @@ -1630,109 +928,8 @@ | ||
| 1630 | 928 | |
| 1631 | 929 | /** |
| 1632 | 930 | * Delete all chunks for a URL from Pinecone |
| 1633 | 931 | */ |
| 1634 | -/** | |
| 1635 | - * Delete leftover md5(url)_chunk_N vectors after a URL's content was re-stored | |
| 1636 | - * as a SINGLE vector (content shrank below the chunk threshold on edit/re-import). | |
| 1637 | - * Unlike delete_pinecone_chunks_by_url this leaves the base id alone — the caller | |
| 1638 | - * just upserted the new content there. Failure is logged, not fatal: the save | |
| 1639 | - * itself succeeded, and the next save retries the sweep. | |
| 1640 | - */ | |
| 1641 | -private static function cleanup_pinecone_chunk_stragglers($source_url, $bot_id) { | |
| 1642 | - if ($bot_id === 'default' || !class_exists('MxChat_Multi_Bot_Manager')) { | |
| 1643 | - $pinecone_options = get_option('mxchat_pinecone_addon_options'); | |
| 1644 | - $api_key = $pinecone_options['mxchat_pinecone_api_key'] ?? ''; | |
| 1645 | - $host = $pinecone_options['mxchat_pinecone_host'] ?? ''; | |
| 1646 | - $namespace = $pinecone_options['mxchat_pinecone_namespace'] ?? ''; | |
| 1647 | - } else { | |
| 1648 | - $bot_pinecone_config = apply_filters('mxchat_get_bot_pinecone_config', array(), $bot_id); | |
| 1649 | - if (empty($bot_pinecone_config)) { | |
| 1650 | - $pinecone_options = get_option('mxchat_pinecone_addon_options'); | |
| 1651 | - $api_key = $pinecone_options['mxchat_pinecone_api_key'] ?? ''; | |
| 1652 | - $host = $pinecone_options['mxchat_pinecone_host'] ?? ''; | |
| 1653 | - $namespace = $pinecone_options['mxchat_pinecone_namespace'] ?? ''; | |
| 1654 | - } else { | |
| 1655 | - $api_key = $bot_pinecone_config['api_key'] ?? ''; | |
| 1656 | - $host = $bot_pinecone_config['host'] ?? ''; | |
| 1657 | - $namespace = $bot_pinecone_config['namespace'] ?? ''; | |
| 1658 | - } | |
| 1659 | - } | |
| 1660 | - | |
| 1661 | - if (empty($host) || empty($api_key)) { | |
| 1662 | - return; | |
| 1663 | - } | |
| 1664 | - | |
| 1665 | - $stragglers = array(); | |
| 1666 | - | |
| 1667 | - // Pinecone /vectors/list is a GET endpoint with query-string parameters (a POST | |
| 1668 | - // answers 200-with-an-empty-body, which reads as "no stragglers"). | |
| 1669 | - $query_params = array( | |
| 1670 | - 'prefix' => md5($source_url) . '_chunk_', | |
| 1671 | - 'limit' => 100, | |
| 1672 | - ); | |
| 1673 | - if (!empty($namespace)) { | |
| 1674 | - $query_params['namespace'] = $namespace; | |
| 1675 | - } | |
| 1676 | - | |
| 1677 | - $list_url = "https://{$host}/vectors/list?" . http_build_query($query_params); | |
| 1678 | - | |
| 1679 | - do { | |
| 1680 | - $list_response = wp_remote_get($list_url, array( | |
| 1681 | - 'headers' => array( | |
| 1682 | - 'Api-Key' => $api_key, | |
| 1683 | - 'accept' => 'application/json', | |
| 1684 | - ), | |
| 1685 | - 'timeout' => 30, | |
| 1686 | - )); | |
| 1687 | - | |
| 1688 | - if (is_wp_error($list_response) || wp_remote_retrieve_response_code($list_response) !== 200) { | |
| 1689 | - break; | |
| 1690 | - } | |
| 1691 | - | |
| 1692 | - $list_data = json_decode(wp_remote_retrieve_body($list_response), true); | |
| 1693 | - if (!empty($list_data['vectors'])) { | |
| 1694 | - foreach ($list_data['vectors'] as $vector) { | |
| 1695 | - if (isset($vector['id'])) { | |
| 1696 | - $stragglers[] = $vector['id']; | |
| 1697 | - } | |
| 1698 | - } | |
| 1699 | - } | |
| 1700 | - | |
| 1701 | - $next_token = $list_data['pagination']['next'] ?? ''; | |
| 1702 | - if (empty($next_token)) { | |
| 1703 | - break; | |
| 1704 | - } | |
| 1705 | - | |
| 1706 | - $query_params['paginationToken'] = $next_token; | |
| 1707 | - $list_url = "https://{$host}/vectors/list?" . http_build_query($query_params); | |
| 1708 | - } while (true); | |
| 1709 | - | |
| 1710 | - if (empty($stragglers)) { | |
| 1711 | - return; | |
| 1712 | - } | |
| 1713 | - | |
| 1714 | - $delete_body = array('ids' => $stragglers); | |
| 1715 | - if (!empty($namespace)) { | |
| 1716 | - $delete_body['namespace'] = $namespace; | |
| 1717 | - } | |
| 1718 | - | |
| 1719 | - $delete_response = wp_remote_post("https://{$host}/vectors/delete", array( | |
| 1720 | - 'headers' => array( | |
| 1721 | - 'Api-Key' => $api_key, | |
| 1722 | - 'accept' => 'application/json', | |
| 1723 | - 'content-type' => 'application/json' | |
| 1724 | - ), | |
| 1725 | - 'body' => wp_json_encode($delete_body), | |
| 1726 | - 'timeout' => 30 | |
| 1727 | - )); | |
| 1728 | - | |
| 1729 | - if ((is_wp_error($delete_response) || wp_remote_retrieve_response_code($delete_response) !== 200) | |
| 1730 | - && class_exists('MxChat_Admin') && method_exists('MxChat_Admin', 'mxchat_log_debug')) { | |
| 1731 | - 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))); | |
| 1732 | - } | |
| 1733 | -} | |
| 1734 | - | |
| 1735 | 932 | private static function delete_pinecone_chunks_by_url($source_url, $bot_id) { |
| 1736 | 933 | // Get Pinecone configuration |
| 1737 | 934 | if ($bot_id === 'default' || !class_exists('MxChat_Multi_Bot_Manager')) { |
| 1738 | 935 | $pinecone_options = get_option('mxchat_pinecone_addon_options'); |
| @@ -1866,297 +1063,6 @@ | ||
| 1866 | 1063 | } |
| 1867 | 1064 | |
| 1868 | 1065 | //error_log('[MXCHAT-CHUNK-DELETE] Deleted ' . $result . ' rows from WordPress DB'); |
| 1869 | 1066 | return true; |
| 1870 | -} | |
| 1871 | - | |
| 1872 | -/** | |
| 1873 | - * Hybrid keyword boost (plan-38ffa1): detect whether the WP-DB knowledge | |
| 1874 | - * table can serve the keyword leg via a MySQL FULLTEXT index, creating the | |
| 1875 | - * index if needed. Detection runs once and caches the answer in the | |
| 1876 | - * mxchat_hybrid_keyword_capability option ('fulltext' | 'like'); pass | |
| 1877 | - * $force to re-detect. LIKE is the graceful fallback for shared hosts | |
| 1878 | - * whose ALTER fails — the feature works either way, FULLTEXT just ranks | |
| 1879 | - * better and scales. | |
| 1880 | - * | |
| 1881 | - * @param bool $force Re-run detection even if a cached answer exists. | |
| 1882 | - * @return string 'fulltext' or 'like' | |
| 1883 | - */ | |
| 1884 | -public static function mxchat_hybrid_detect_capability($force = false) { | |
| 1885 | - $cached = get_option('mxchat_hybrid_keyword_capability', ''); | |
| 1886 | - if (!$force && in_array($cached, array('fulltext', 'like'), true)) { | |
| 1887 | - return $cached; | |
| 1888 | - } | |
| 1889 | - | |
| 1890 | - global $wpdb; | |
| 1891 | - $table = $wpdb->prefix . 'mxchat_system_prompt_content'; | |
| 1892 | - | |
| 1893 | - $index_exists = $wpdb->get_var("SHOW INDEX FROM {$table} WHERE Key_name = 'mxchat_content_ft'"); | |
| 1894 | - if (!$index_exists) { | |
| 1895 | - // Suppress the visible error on hosts where this is not permitted — | |
| 1896 | - // failure is an expected, handled outcome (LIKE fallback). | |
| 1897 | - $suppress = $wpdb->suppress_errors(true); | |
| 1898 | - $wpdb->query("ALTER TABLE {$table} ADD FULLTEXT INDEX mxchat_content_ft (article_content)"); | |
| 1899 | - $wpdb->suppress_errors($suppress); | |
| 1900 | - $index_exists = $wpdb->get_var("SHOW INDEX FROM {$table} WHERE Key_name = 'mxchat_content_ft'"); | |
| 1901 | - } | |
| 1902 | - | |
| 1903 | - $capability = $index_exists ? 'fulltext' : 'like'; | |
| 1904 | - update_option('mxchat_hybrid_keyword_capability', $capability); | |
| 1905 | - return $capability; | |
| 1906 | -} | |
| 1907 | - | |
| 1908 | -/** | |
| 1909 | - * Public entry for re-embedding already-stored content in place (wp mxchat | |
| 1910 | - * rtl-repair, plan d1e6f7). Thin wrapper so the repair CLI gets the exact | |
| 1911 | - * provider routing the import path uses — the repaired vector must come from | |
| 1912 | - * the same model family the bot indexes with, or retrieval stays broken. | |
| 1913 | - */ | |
| 1914 | -public static function regenerate_embedding($text, $api_key, $bot_id = 'default') { | |
| 1915 | - return self::generate_embedding($text, $api_key, $bot_id); | |
| 1916 | -} | |
| 1917 | - | |
| 1918 | -/** | |
| 1919 | - * Restore logical character order in PDF-extracted RTL text (plan 32bf9e). | |
| 1920 | - * | |
| 1921 | - * The bundled Smalot parser only un-reverses text runs tagged with the | |
| 1922 | - * ReversedChars marked-content operator (Word emits it; LibreOffice and most | |
| 1923 | - * other producers do not), so their Hebrew/Arabic PDFs extract in visual | |
| 1924 | - * (reversed) order and embed/search as garbage. This is OUR post-processing | |
| 1925 | - * seam over getText() — the parser itself is never patched (it gets replaced | |
| 1926 | - * wholesale on library updates). | |
| 1927 | - * | |
| 1928 | - * Heuristic and deliberately conservative, per line: | |
| 1929 | - * - lines without strong RTL codepoints are untouched (a fully-Latin line in | |
| 1930 | - * an RTL document therefore stays as extracted — accepted limitation); | |
| 1931 | - * - Arabic presentation forms are a definitive visual-order signal (they only | |
| 1932 | - * appear in shaped output): de-shape to base letters and reverse; | |
| 1933 | - * - otherwise flip only on positive evidence — Hebrew final-letter position | |
| 1934 | - * (a sofit at word START only happens in reversed text) or sentence | |
| 1935 | - * punctuation position (leading in visual order, trailing in logical); | |
| 1936 | - * - ambiguous lines are left alone: a conservative miss beats corrupting a | |
| 1937 | - * Word-produced extraction the parser already handled (the double-flip | |
| 1938 | - * guard this plan's approval named mandatory). | |
| 1939 | - * | |
| 1940 | - * @param string $text One extracted page string, straight from getText(). | |
| 1941 | - * @param string $context Caller tag for the Debug Mode entry (site + page). | |
| 1942 | - * @return string Text with RTL lines restored to logical order. | |
| 1943 | - */ | |
| 1944 | -public static function normalize_pdf_rtl($text, $context = '') { | |
| 1945 | - if (!is_string($text) || '' === $text) { | |
| 1946 | - return $text; | |
| 1947 | - } | |
| 1948 | - // Escape hatch for sites whose PDFs already extract logically. | |
| 1949 | - if (!apply_filters('mxchat_pdf_rtl_normalize', true, $text)) { | |
| 1950 | - return $text; | |
| 1951 | - } | |
| 1952 | - // Fast bail: nothing RTL anywhere in the page. | |
| 1953 | - 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)) { | |
| 1954 | - return $text; | |
| 1955 | - } | |
| 1956 | - | |
| 1957 | - $parts = preg_split('/(\R)/u', $text, -1, PREG_SPLIT_DELIM_CAPTURE); | |
| 1958 | - if (false === $parts) { | |
| 1959 | - return $text; | |
| 1960 | - } | |
| 1961 | - | |
| 1962 | - $flipped_lines = 0; | |
| 1963 | - $deshaped_lines = 0; | |
| 1964 | - foreach ($parts as $i => $part) { | |
| 1965 | - if ('' === $part || preg_match('/^\R$/u', $part)) { | |
| 1966 | - continue; | |
| 1967 | - } | |
| 1968 | - $was_flipped = false; | |
| 1969 | - $was_deshaped = false; | |
| 1970 | - $new = self::pdf_rtl_normalize_line($part, $was_flipped, $was_deshaped); | |
| 1971 | - if ($new !== $part) { | |
| 1972 | - $parts[$i] = $new; | |
| 1973 | - } | |
| 1974 | - if ($was_flipped) { | |
| 1975 | - $flipped_lines++; | |
| 1976 | - } | |
| 1977 | - if ($was_deshaped) { | |
| 1978 | - $deshaped_lines++; | |
| 1979 | - } | |
| 1980 | - } | |
| 1981 | - | |
| 1982 | - if (($flipped_lines || $deshaped_lines) && class_exists('MxChat_Admin')) { | |
| 1983 | - MxChat_Admin::mxchat_log_debug('pdf_rtl_normalized', 'RTL PDF text restored to logical order', array( | |
| 1984 | - 'context' => (string) $context, | |
| 1985 | - 'lines_flipped' => $flipped_lines, | |
| 1986 | - 'lines_deshaped' => $deshaped_lines, | |
| 1987 | - 'decision' => 'visual-order extraction detected', | |
| 1988 | - )); | |
| 1989 | - } | |
| 1990 | - | |
| 1991 | - return implode('', $parts); | |
| 1992 | -} | |
| 1993 | - | |
| 1994 | -/** | |
| 1995 | - * Normalize one line. Sets $flipped/$deshaped for the caller's debug entry. | |
| 1996 | - */ | |
| 1997 | -private static function pdf_rtl_normalize_line($line, &$flipped, &$deshaped) { | |
| 1998 | - $flipped = false; | |
| 1999 | - $deshaped = false; | |
| 2000 | - | |
| 2001 | - 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)) { | |
| 2002 | - return $line; | |
| 2003 | - } | |
| 2004 | - | |
| 2005 | - $has_forms = (bool) preg_match('/[\x{FB50}-\x{FDFF}\x{FE70}-\x{FEFF}]/u', $line); | |
| 2006 | - $work = $line; | |
| 2007 | - if ($has_forms) { | |
| 2008 | - $work = strtr($work, self::pdf_rtl_deshape_map()); | |
| 2009 | - $deshaped = ($work !== $line); | |
| 2010 | - } | |
| 2011 | - | |
| 2012 | - $verdict = 'ambiguous'; | |
| 2013 | - if ($has_forms) { | |
| 2014 | - // Shaped glyph codepoints only exist in visual-order output. | |
| 2015 | - $verdict = 'visual'; | |
| 2016 | - } else { | |
| 2017 | - // Strong-direction dominance gate first: an LTR-dominant line with an | |
| 2018 | - // embedded RTL word is not flip material. | |
| 2019 | - $rtl_count = preg_match_all('/[\x{0590}-\x{05FF}\x{0600}-\x{06FF}\x{0750}-\x{077F}]/u', $work, $m_rtl); | |
| 2020 | - $ltr_count = preg_match_all('/[A-Za-z]/u', $work, $m_ltr); | |
| 2021 | - if ($rtl_count < 1 || $rtl_count <= $ltr_count) { | |
| 2022 | - return $line; | |
| 2023 | - } | |
| 2024 | - | |
| 2025 | - // Hebrew final letters (ך ם ן ף ץ) end words in logical text; one at | |
| 2026 | - // a word START (Hebrew letter follows, none precedes) is reversal | |
| 2027 | - // evidence. Positional, so it survives the line being reversed. | |
| 2028 | - $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); | |
| 2029 | - $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, $m_t); | |
| 2030 | - if ($sofit_initial > $sofit_terminal) { | |
| 2031 | - $verdict = 'visual'; | |
| 2032 | - } elseif ($sofit_terminal > $sofit_initial) { | |
| 2033 | - $verdict = 'logical'; | |
| 2034 | - } else { | |
| 2035 | - // Sentence punctuation lands at the visual LEFT edge of an RTL | |
| 2036 | - // line, i.e. the START of a visual-order extraction. | |
| 2037 | - $trimmed = trim($work); | |
| 2038 | - $starts_punct = (bool) preg_match('/^[.?!:;,]/u', $trimmed); | |
| 2039 | - $ends_punct = (bool) preg_match('/[.?!:;,]$/u', $trimmed); | |
| 2040 | - if ($starts_punct && !$ends_punct) { | |
| 2041 | - $verdict = 'visual'; | |
| 2042 | - } elseif ($ends_punct && !$starts_punct) { | |
| 2043 | - $verdict = 'logical'; | |
| 2044 | - } | |
| 2045 | - } | |
| 2046 | - } | |
| 2047 | - | |
| 2048 | - if ('visual' !== $verdict) { | |
| 2049 | - // Ambiguous or logical: hand back the original line UNLESS we | |
| 2050 | - // de-shaped (de-shaping alone is always safe — same letters, same | |
| 2051 | - // order, un-ligated). | |
| 2052 | - return $deshaped ? $work : $line; | |
| 2053 | - } | |
| 2054 | - | |
| 2055 | - $flipped = true; | |
| 2056 | - return self::pdf_rtl_flip_line($work); | |
| 2057 | -} | |
| 2058 | - | |
| 2059 | -/** | |
| 2060 | - * Reverse a visual-order line into logical order: full character reversal, | |
| 2061 | - * mirror paired punctuation, then re-reverse embedded LTR runs (Latin words | |
| 2062 | - * and digit sequences, incl. Arabic-Indic digits) so they stay readable. | |
| 2063 | - */ | |
| 2064 | -private static function pdf_rtl_flip_line($line) { | |
| 2065 | - $chars = preg_split('//u', $line, -1, PREG_SPLIT_NO_EMPTY); | |
| 2066 | - if (false === $chars) { | |
| 2067 | - return $line; | |
| 2068 | - } | |
| 2069 | - $reversed = implode('', array_reverse($chars)); | |
| 2070 | - $reversed = strtr($reversed, array( | |
| 2071 | - '(' => ')', ')' => '(', | |
| 2072 | - '[' => ']', ']' => '[', | |
| 2073 | - '{' => '}', '}' => '{', | |
| 2074 | - '<' => '>', '>' => '<', | |
| 2075 | - )); | |
| 2076 | - $restored = preg_replace_callback( | |
| 2077 | - '/[0-9A-Za-z\x{0660}-\x{0669}\x{06F0}-\x{06F9}](?:[0-9A-Za-z\x{0660}-\x{0669}\x{06F0}-\x{06F9} .,\'"%\-:\/]*[0-9A-Za-z\x{0660}-\x{0669}\x{06F0}-\x{06F9}])?/u', | |
| 2078 | - function ($m) { | |
| 2079 | - $run = preg_split('//u', $m[0], -1, PREG_SPLIT_NO_EMPTY); | |
| 2080 | - return false === $run ? $m[0] : implode('', array_reverse($run)); | |
| 2081 | - }, | |
| 2082 | - $reversed | |
| 2083 | - ); | |
| 2084 | - return null === $restored ? $reversed : $restored; | |
| 2085 | -} | |
| 2086 | - | |
| 2087 | -/** | |
| 2088 | - * Arabic presentation forms (A + B) -> base letters. Built once from range | |
| 2089 | - * specs rather than ~120 hand-written literal entries; every codepoint in a | |
| 2090 | - * range maps to the same base sequence (isolated/final/initial/medial forms | |
| 2091 | - * of one letter are contiguous in the FE70 block). | |
| 2092 | - */ | |
| 2093 | -private static function pdf_rtl_deshape_map() { | |
| 2094 | - static $map = null; | |
| 2095 | - if (null !== $map) { | |
| 2096 | - return $map; | |
| 2097 | - } | |
| 2098 | - $ranges = array( | |
| 2099 | - // Form B harakat (each pair = standalone + tatweel-joined form). | |
| 2100 | - array(0xFE70, 0xFE71, array(0x064B)), array(0xFE72, 0xFE72, array(0x064C)), | |
| 2101 | - array(0xFE74, 0xFE74, array(0x064D)), array(0xFE76, 0xFE77, array(0x064E)), | |
| 2102 | - array(0xFE78, 0xFE79, array(0x064F)), array(0xFE7A, 0xFE7B, array(0x0650)), | |
| 2103 | - array(0xFE7C, 0xFE7D, array(0x0651)), array(0xFE7E, 0xFE7F, array(0x0652)), | |
| 2104 | - // Form B letters. | |
| 2105 | - array(0xFE80, 0xFE80, array(0x0621)), array(0xFE81, 0xFE82, array(0x0622)), | |
| 2106 | - array(0xFE83, 0xFE84, array(0x0623)), array(0xFE85, 0xFE86, array(0x0624)), | |
| 2107 | - array(0xFE87, 0xFE88, array(0x0625)), array(0xFE89, 0xFE8C, array(0x0626)), | |
| 2108 | - array(0xFE8D, 0xFE8E, array(0x0627)), array(0xFE8F, 0xFE92, array(0x0628)), | |
| 2109 | - array(0xFE93, 0xFE94, array(0x0629)), array(0xFE95, 0xFE98, array(0x062A)), | |
| 2110 | - array(0xFE99, 0xFE9C, array(0x062B)), array(0xFE9D, 0xFEA0, array(0x062C)), | |
| 2111 | - array(0xFEA1, 0xFEA4, array(0x062D)), array(0xFEA5, 0xFEA8, array(0x062E)), | |
| 2112 | - array(0xFEA9, 0xFEAA, array(0x062F)), array(0xFEAB, 0xFEAC, array(0x0630)), | |
| 2113 | - array(0xFEAD, 0xFEAE, array(0x0631)), array(0xFEAF, 0xFEB0, array(0x0632)), | |
| 2114 | - array(0xFEB1, 0xFEB4, array(0x0633)), array(0xFEB5, 0xFEB8, array(0x0634)), | |
| 2115 | - array(0xFEB9, 0xFEBC, array(0x0635)), array(0xFEBD, 0xFEC0, array(0x0636)), | |
| 2116 | - array(0xFEC1, 0xFEC4, array(0x0637)), array(0xFEC5, 0xFEC8, array(0x0638)), | |
| 2117 | - array(0xFEC9, 0xFECC, array(0x0639)), array(0xFECD, 0xFED0, array(0x063A)), | |
| 2118 | - array(0xFED1, 0xFED4, array(0x0641)), array(0xFED5, 0xFED8, array(0x0642)), | |
| 2119 | - array(0xFED9, 0xFEDC, array(0x0643)), array(0xFEDD, 0xFEE0, array(0x0644)), | |
| 2120 | - array(0xFEE1, 0xFEE4, array(0x0645)), array(0xFEE5, 0xFEE8, array(0x0646)), | |
| 2121 | - array(0xFEE9, 0xFEEC, array(0x0647)), array(0xFEED, 0xFEEE, array(0x0648)), | |
| 2122 | - array(0xFEEF, 0xFEF0, array(0x0649)), array(0xFEF1, 0xFEF4, array(0x064A)), | |
| 2123 | - // Form B lam-alef ligatures decompose to two letters. | |
| 2124 | - array(0xFEF5, 0xFEF6, array(0x0644, 0x0622)), array(0xFEF7, 0xFEF8, array(0x0644, 0x0623)), | |
| 2125 | - array(0xFEF9, 0xFEFA, array(0x0644, 0x0625)), array(0xFEFB, 0xFEFC, array(0x0644, 0x0627)), | |
| 2126 | - // Form A: Persian / Urdu letters in common use. | |
| 2127 | - array(0xFB56, 0xFB59, array(0x067E)), array(0xFB66, 0xFB69, array(0x0679)), | |
| 2128 | - array(0xFB7A, 0xFB7D, array(0x0686)), array(0xFB88, 0xFB89, array(0x0688)), | |
| 2129 | - array(0xFB8A, 0xFB8B, array(0x0698)), array(0xFB8E, 0xFB91, array(0x06A9)), | |
| 2130 | - array(0xFB92, 0xFB95, array(0x06AF)), array(0xFBA6, 0xFBA9, array(0x06C1)), | |
| 2131 | - array(0xFBAA, 0xFBAD, array(0x06BE)), array(0xFBAE, 0xFBAF, array(0x06D2)), | |
| 2132 | - array(0xFBFC, 0xFBFF, array(0x06CC)), | |
| 2133 | - ); | |
| 2134 | - $map = array(); | |
| 2135 | - foreach ($ranges as $range) { | |
| 2136 | - $base = ''; | |
| 2137 | - foreach ($range[2] as $cp) { | |
| 2138 | - $base .= self::pdf_rtl_cp_to_utf8($cp); | |
| 2139 | - } | |
| 2140 | - for ($cp = $range[0]; $cp <= $range[1]; $cp++) { | |
| 2141 | - $map[self::pdf_rtl_cp_to_utf8($cp)] = $base; | |
| 2142 | - } | |
| 2143 | - } | |
| 2144 | - return $map; | |
| 2145 | -} | |
| 2146 | - | |
| 2147 | -/** | |
| 2148 | - * Codepoint to UTF-8 without ext-intl / mbstring entity tricks (PHP 7.2 floor). | |
| 2149 | - */ | |
| 2150 | -private static function pdf_rtl_cp_to_utf8($cp) { | |
| 2151 | - if ($cp < 0x80) { | |
| 2152 | - return chr($cp); | |
| 2153 | - } | |
| 2154 | - if ($cp < 0x800) { | |
| 2155 | - return chr(0xC0 | ($cp >> 6)) . chr(0x80 | ($cp & 0x3F)); | |
| 2156 | - } | |
| 2157 | - if ($cp < 0x10000) { | |
| 2158 | - return chr(0xE0 | ($cp >> 12)) . chr(0x80 | (($cp >> 6) & 0x3F)) . chr(0x80 | ($cp & 0x3F)); | |
| 2159 | - } | |
| 2160 | - return chr(0xF0 | ($cp >> 18)) . chr(0x80 | (($cp >> 12) & 0x3F)) . chr(0x80 | (($cp >> 6) & 0x3F)) . chr(0x80 | ($cp & 0x3F)); | |
| 2161 | 1067 | } |
| 2162 | 1068 | } |
