| @@ -3482,22 +3482,9 @@ | ||
| 3482 | 3482 | } |
| 3483 | 3483 | |
| 3484 | 3484 | // Get bot_id from request |
| 3485 | 3485 | $bot_id = isset($_POST['bot_id']) ? sanitize_key($_POST['bot_id']) : 'default'; |
| 3486 | - | |
| 3487 | - // ACF→PDF extraction is opt-in per import batch. Persist the last-used value so users | |
| 3488 | - // don't re-check on every batch; the default is OFF for installs that haven't set it. | |
| 3489 | - $extract_acf_pdfs = !empty($_POST['extract_acf_pdfs']) && $_POST['extract_acf_pdfs'] !== 'false'; | |
| 3490 | - $mxchat_options = get_option('mxchat_options', array()); | |
| 3491 | - if (!is_array($mxchat_options)) { | |
| 3492 | - $mxchat_options = array(); | |
| 3493 | - } | |
| 3494 | - $prior_default = !empty($mxchat_options['acf_pdf_extract_default']); | |
| 3495 | - if ($prior_default !== $extract_acf_pdfs) { | |
| 3496 | - $mxchat_options['acf_pdf_extract_default'] = $extract_acf_pdfs ? 1 : 0; | |
| 3497 | - update_option('mxchat_options', $mxchat_options); | |
| 3498 | - } | |
| 3499 | - | |
| 3486 | + | |
| 3500 | 3487 | // Process only ONE post at a time to avoid request size issues |
| 3501 | 3488 | $post_id = reset($post_ids); |
| 3502 | 3489 | $post = get_post($post_id); |
| 3503 | 3490 | |
| @@ -3603,57 +3590,23 @@ | ||
| 3603 | 3590 | } |
| 3604 | 3591 | |
| 3605 | 3592 | // ADD ACF FIELDS SUPPORT |
| 3606 | 3593 | $acf_fields = $this->mxchat_get_acf_fields_for_post($post_id); |
| 3607 | - $pdf_extracted_count = 0; | |
| 3608 | 3594 | if (!empty($acf_fields)) { |
| 3609 | 3595 | $acf_content_parts = array(); |
| 3610 | - $pdf_attachment_ids = array(); | |
| 3611 | - | |
| 3596 | + | |
| 3612 | 3597 | foreach ($acf_fields as $field_name => $field_value) { |
| 3613 | 3598 | $formatted_value = $this->mxchat_format_acf_field_value($field_value, $field_name, $post_id); |
| 3614 | - | |
| 3599 | + | |
| 3615 | 3600 | if (!empty($formatted_value)) { |
| 3616 | 3601 | $field_label = ucwords(str_replace('_', ' ', $field_name)); |
| 3617 | 3602 | $acf_content_parts[] = $field_label . ": " . $formatted_value; |
| 3618 | 3603 | } |
| 3619 | - | |
| 3620 | - // Walk this field's value tree for any PDF attachment references and queue them for extraction. | |
| 3621 | - // Only when the user opted into ACF→PDF extraction for this batch; otherwise the ACF text | |
| 3622 | - // still lands in the KB but the heavier PDF parsing is skipped. | |
| 3623 | - if ($extract_acf_pdfs) { | |
| 3624 | - $this->mxchat_collect_pdf_attachment_ids_from_acf_value($field_value, $pdf_attachment_ids); | |
| 3625 | - } | |
| 3626 | 3604 | } |
| 3627 | - | |
| 3605 | + | |
| 3628 | 3606 | if (!empty($acf_content_parts)) { |
| 3629 | 3607 | $content .= "\n\n" . implode("\n", $acf_content_parts); |
| 3630 | 3608 | } |
| 3631 | - | |
| 3632 | - // Extract text from each unique PDF found in ACF fields and append as a labeled section | |
| 3633 | - if ($extract_acf_pdfs && !empty($pdf_attachment_ids)) { | |
| 3634 | - $pdf_attachment_ids = array_unique(array_filter(array_map('intval', $pdf_attachment_ids))); | |
| 3635 | - $pdf_sections = array(); | |
| 3636 | - foreach ($pdf_attachment_ids as $att_id) { | |
| 3637 | - $pdf_text = $this->mxchat_extract_pdf_text_by_attachment_id($att_id); | |
| 3638 | - if (!empty($pdf_text)) { | |
| 3639 | - $pdf_title = get_the_title($att_id); | |
| 3640 | - $pdf_url = wp_get_attachment_url($att_id); | |
| 3641 | - $header = 'PDF Attachment'; | |
| 3642 | - if (!empty($pdf_title)) { | |
| 3643 | - $header .= ': ' . $pdf_title; | |
| 3644 | - } | |
| 3645 | - if (!empty($pdf_url)) { | |
| 3646 | - $header .= ' (' . $pdf_url . ')'; | |
| 3647 | - } | |
| 3648 | - $pdf_sections[] = $header . "\n" . $pdf_text; | |
| 3649 | - $pdf_extracted_count++; | |
| 3650 | - } | |
| 3651 | - } | |
| 3652 | - if (!empty($pdf_sections)) { | |
| 3653 | - $content .= "\n\nAttached PDFs:\n" . implode("\n\n", $pdf_sections); | |
| 3654 | - } | |
| 3655 | - } | |
| 3656 | 3609 | } |
| 3657 | 3610 | |
| 3658 | 3611 | // ADD CUSTOM POST META SUPPORT (whitelisted non-ACF meta fields) |
| 3659 | 3612 | $custom_meta = $this->mxchat_get_whitelisted_post_meta($post_id); |
| @@ -3783,9 +3736,8 @@ | ||
| 3783 | 3736 | 'title' => $post->post_title, |
| 3784 | 3737 | 'operation_type' => $operation_type, |
| 3785 | 3738 | 'vector_id' => $vector_id, |
| 3786 | 3739 | 'acf_fields_found' => $acf_field_count, |
| 3787 | - 'pdf_extracted_count' => (int) $pdf_extracted_count, | |
| 3788 | 3740 | 'content_preview' => substr($content, 0, 100) . '...', |
| 3789 | 3741 | 'bot_id' => $bot_id |
| 3790 | 3742 | )); |
| 3791 | 3743 | exit; |
| @@ -5120,197 +5072,8 @@ | ||
| 5120 | 5072 | return implode(', ', array_filter($text_parts)); |
| 5121 | 5073 | } |
| 5122 | 5074 | |
| 5123 | 5075 | /** |
| 5124 | - * Walk an ACF field value tree and collect attachment IDs for any value that | |
| 5125 | - * resolves to a PDF in the WordPress media library. Handles the three shapes | |
| 5126 | - * ACF returns for File/Image/URL fields (array with ID+url, integer attachment ID, | |
| 5127 | - * plain URL string), and recurses through repeater/group/flexible content. | |
| 5128 | - * | |
| 5129 | - * @param mixed $value The ACF field value (any depth) | |
| 5130 | - * @param array $out Accumulator (passed by reference) for attachment IDs | |
| 5131 | - * @param int $depth Recursion guard | |
| 5132 | - */ | |
| 5133 | -private function mxchat_collect_pdf_attachment_ids_from_acf_value($value, &$out, $depth = 0) { | |
| 5134 | - if ($depth > 6) { | |
| 5135 | - return; // prevent runaway recursion on circular/very-deep structures | |
| 5136 | - } | |
| 5137 | - | |
| 5138 | - if (empty($value)) { | |
| 5139 | - return; | |
| 5140 | - } | |
| 5141 | - | |
| 5142 | - // Array shapes: ACF File/Image return value=array; repeaters/groups are arrays of arrays | |
| 5143 | - if (is_array($value)) { | |
| 5144 | - // Direct File/Image-style array (has 'url' and usually 'ID' + 'mime_type') | |
| 5145 | - $looks_like_attachment = isset($value['url']) || isset($value['ID']) || isset($value['id']); | |
| 5146 | - if ($looks_like_attachment) { | |
| 5147 | - $att_id = 0; | |
| 5148 | - if (!empty($value['ID']) && is_numeric($value['ID'])) { | |
| 5149 | - $att_id = (int) $value['ID']; | |
| 5150 | - } elseif (!empty($value['id']) && is_numeric($value['id'])) { | |
| 5151 | - $att_id = (int) $value['id']; | |
| 5152 | - } elseif (!empty($value['url']) && is_string($value['url'])) { | |
| 5153 | - $att_id = (int) attachment_url_to_postid($value['url']); | |
| 5154 | - } | |
| 5155 | - | |
| 5156 | - $is_pdf = false; | |
| 5157 | - if (!empty($value['mime_type']) && $value['mime_type'] === 'application/pdf') { | |
| 5158 | - $is_pdf = true; | |
| 5159 | - } elseif (!empty($value['subtype']) && strtolower((string) $value['subtype']) === 'pdf') { | |
| 5160 | - $is_pdf = true; | |
| 5161 | - } elseif (!empty($value['url']) && is_string($value['url']) && $this->mxchat_url_looks_like_pdf($value['url'])) { | |
| 5162 | - $is_pdf = true; | |
| 5163 | - } elseif ($att_id && get_post_mime_type($att_id) === 'application/pdf') { | |
| 5164 | - $is_pdf = true; | |
| 5165 | - } | |
| 5166 | - | |
| 5167 | - if ($is_pdf && $att_id && get_post_mime_type($att_id) === 'application/pdf') { | |
| 5168 | - $out[] = $att_id; | |
| 5169 | - } | |
| 5170 | - // An array node that represents one attachment doesn't contain other | |
| 5171 | - // attachments inside it — done with this branch. | |
| 5172 | - return; | |
| 5173 | - } | |
| 5174 | - | |
| 5175 | - // Recurse: repeater rows, flexible-content layouts, groups, etc. | |
| 5176 | - foreach ($value as $sub) { | |
| 5177 | - $this->mxchat_collect_pdf_attachment_ids_from_acf_value($sub, $out, $depth + 1); | |
| 5178 | - } | |
| 5179 | - return; | |
| 5180 | - } | |
| 5181 | - | |
| 5182 | - // Plain numeric attachment ID (ACF File field set to "Return: ID") | |
| 5183 | - if (is_numeric($value)) { | |
| 5184 | - $att_id = (int) $value; | |
| 5185 | - if ($att_id > 0 && get_post_mime_type($att_id) === 'application/pdf') { | |
| 5186 | - $out[] = $att_id; | |
| 5187 | - } | |
| 5188 | - return; | |
| 5189 | - } | |
| 5190 | - | |
| 5191 | - // Plain string — URL pointing at a PDF (ACF File field set to "Return: URL", or a custom URL/text field) | |
| 5192 | - if (is_string($value)) { | |
| 5193 | - $trimmed = trim($value); | |
| 5194 | - if ($trimmed !== '' && $this->mxchat_url_looks_like_pdf($trimmed)) { | |
| 5195 | - $att_id = (int) attachment_url_to_postid($trimmed); | |
| 5196 | - if ($att_id > 0 && get_post_mime_type($att_id) === 'application/pdf') { | |
| 5197 | - $out[] = $att_id; | |
| 5198 | - } | |
| 5199 | - } | |
| 5200 | - return; | |
| 5201 | - } | |
| 5202 | -} | |
| 5203 | - | |
| 5204 | -/** | |
| 5205 | - * Heuristic: does this URL/string look like a PDF reference? | |
| 5206 | - * Tolerates query strings and fragments (#page=2). | |
| 5207 | - */ | |
| 5208 | -private function mxchat_url_looks_like_pdf($url) { | |
| 5209 | - if (!is_string($url) || $url === '') { | |
| 5210 | - return false; | |
| 5211 | - } | |
| 5212 | - // Strip query + fragment before checking extension | |
| 5213 | - $path = preg_replace('/[?#].*$/', '', $url); | |
| 5214 | - return (bool) preg_match('/\.pdf$/i', $path); | |
| 5215 | -} | |
| 5216 | - | |
| 5217 | -/** | |
| 5218 | - * Extract text from a PDF attachment by ID using the bundled Smalot parser. | |
| 5219 | - * Reads the file directly from disk via get_attached_file (no HTTP fetch). | |
| 5220 | - * Result is cached on the attachment as post_meta keyed by file mtime so we | |
| 5221 | - * only parse the same PDF once unless the file changes on disk. | |
| 5222 | - * | |
| 5223 | - * @param int $attachment_id | |
| 5224 | - * @return string Extracted plain text, or '' on failure. | |
| 5225 | - */ | |
| 5226 | -private function mxchat_extract_pdf_text_by_attachment_id($attachment_id) { | |
| 5227 | - $attachment_id = (int) $attachment_id; | |
| 5228 | - if ($attachment_id <= 0) { | |
| 5229 | - return ''; | |
| 5230 | - } | |
| 5231 | - if (get_post_mime_type($attachment_id) !== 'application/pdf') { | |
| 5232 | - return ''; | |
| 5233 | - } | |
| 5234 | - | |
| 5235 | - $pdf_path = get_attached_file($attachment_id); | |
| 5236 | - if (empty($pdf_path) || !file_exists($pdf_path) || !is_readable($pdf_path)) { | |
| 5237 | - return ''; | |
| 5238 | - } | |
| 5239 | - | |
| 5240 | - // Raw-file size cap. Parsing very large PDFs can OOM the request; skip with a log entry | |
| 5241 | - // and let the rest of the ACF content land in the KB. Filterable for users who need it bigger. | |
| 5242 | - $default_max_bytes = 25 * 1024 * 1024; | |
| 5243 | - $max_bytes = (int) apply_filters('mxchat_acf_pdf_max_bytes', $default_max_bytes, $attachment_id, $pdf_path); | |
| 5244 | - if ($max_bytes > 0) { | |
| 5245 | - $file_size = @filesize($pdf_path); | |
| 5246 | - if ($file_size !== false && $file_size > $max_bytes) { | |
| 5247 | - error_log(sprintf( | |
| 5248 | - '[mxchat] ACF PDF skipped (over size cap): attachment %d "%s" %d bytes > cap %d', | |
| 5249 | - $attachment_id, | |
| 5250 | - basename($pdf_path), | |
| 5251 | - $file_size, | |
| 5252 | - $max_bytes | |
| 5253 | - )); | |
| 5254 | - return ''; | |
| 5255 | - } | |
| 5256 | - } | |
| 5257 | - | |
| 5258 | - $mtime = @filemtime($pdf_path); | |
| 5259 | - $cache_meta_key = '_mxchat_acf_pdf_text_v1'; | |
| 5260 | - $cached = get_post_meta($attachment_id, $cache_meta_key, true); | |
| 5261 | - if (is_array($cached) && isset($cached['mtime'], $cached['text']) && (int) $cached['mtime'] === (int) $mtime) { | |
| 5262 | - return (string) $cached['text']; | |
| 5263 | - } | |
| 5264 | - | |
| 5265 | - $text = ''; | |
| 5266 | - try { | |
| 5267 | - if (function_exists('mxchat_load_pdf_parser')) { | |
| 5268 | - mxchat_load_pdf_parser(); | |
| 5269 | - } | |
| 5270 | - if (!class_exists('\\Smalot\\PdfParser\\Parser')) { | |
| 5271 | - return ''; | |
| 5272 | - } | |
| 5273 | - $parser = new \Smalot\PdfParser\Parser(); | |
| 5274 | - $pdf = $parser->parseFile($pdf_path); | |
| 5275 | - $pages = $pdf->getPages(); | |
| 5276 | - $page_texts = array(); | |
| 5277 | - foreach ($pages as $page) { | |
| 5278 | - $page_text = ''; | |
| 5279 | - try { | |
| 5280 | - $page_text = $page->getText(); | |
| 5281 | - } catch (\Exception $e) { | |
| 5282 | - $page_text = ''; | |
| 5283 | - } | |
| 5284 | - if (!empty($page_text)) { | |
| 5285 | - $page_texts[] = $page_text; | |
| 5286 | - } | |
| 5287 | - } | |
| 5288 | - $text = trim(implode("\n\n", $page_texts)); | |
| 5289 | - } catch (\Exception $e) { | |
| 5290 | - error_log('[mxchat] ACF PDF extraction failed for attachment ' . $attachment_id . ': ' . $e->getMessage()); | |
| 5291 | - return ''; | |
| 5292 | - } catch (\Throwable $e) { | |
| 5293 | - error_log('[mxchat] ACF PDF extraction error for attachment ' . $attachment_id . ': ' . $e->getMessage()); | |
| 5294 | - return ''; | |
| 5295 | - } | |
| 5296 | - | |
| 5297 | - // Cap per-PDF text to avoid blowing up the embedding payload on enormous PDFs. | |
| 5298 | - // The chunker downstream will still split this into multiple vectors. | |
| 5299 | - $max_len = (int) apply_filters('mxchat_acf_pdf_text_max_length', 50000); | |
| 5300 | - if ($max_len > 0 && strlen($text) > $max_len) { | |
| 5301 | - $text = substr($text, 0, $max_len); | |
| 5302 | - } | |
| 5303 | - | |
| 5304 | - update_post_meta($attachment_id, $cache_meta_key, array( | |
| 5305 | - 'mtime' => (int) $mtime, | |
| 5306 | - 'text' => $text, | |
| 5307 | - )); | |
| 5308 | - | |
| 5309 | - return $text; | |
| 5310 | -} | |
| 5311 | - | |
| 5312 | -/** | |
| 5313 | 5076 | * Handle ACF save - fires after ACF fields are saved |
| 5314 | 5077 | * This ensures ACF field data is available when syncing to knowledge base |
| 5315 | 5078 | */ |
| 5316 | 5079 | public function mxchat_handle_acf_save($post_id) { |
| @@ -5553,9 +5316,8 @@ | ||
| 5553 | 5316 | // ADD ACF FIELDS SUPPORT (matches ajax_mxchat_process_selected_content behavior) |
| 5554 | 5317 | $acf_fields = $this->mxchat_get_acf_fields_for_post($post_id); |
| 5555 | 5318 | if (!empty($acf_fields)) { |
| 5556 | 5319 | $acf_content_parts = array(); |
| 5557 | - $pdf_attachment_ids = array(); | |
| 5558 | 5320 | |
| 5559 | 5321 | foreach ($acf_fields as $field_name => $field_value) { |
| 5560 | 5322 | $formatted_value = $this->mxchat_format_acf_field_value($field_value, $field_name, $post_id); |
| 5561 | 5323 | if (!empty($formatted_value)) { |
| @@ -5562,43 +5324,12 @@ | ||
| 5562 | 5324 | // Convert field name to readable label |
| 5563 | 5325 | $field_label = ucwords(str_replace(['_', '-'], ' ', $field_name)); |
| 5564 | 5326 | $acf_content_parts[] = $field_label . ": " . $formatted_value; |
| 5565 | 5327 | } |
| 5566 | - | |
| 5567 | - $this->mxchat_collect_pdf_attachment_ids_from_acf_value($field_value, $pdf_attachment_ids); | |
| 5568 | 5328 | } |
| 5569 | 5329 | |
| 5570 | 5330 | if (!empty($acf_content_parts)) { |
| 5571 | 5331 | $final_content .= "\n\n" . implode("\n", $acf_content_parts); |
| 5572 | - } | |
| 5573 | - | |
| 5574 | - // Gate the auto-sync PDF-extraction loop behind an opt-in option. | |
| 5575 | - // Mirrors the per-batch checkbox the manual content selector has; the | |
| 5576 | - // 25 MB size cap lives in the shared extractor so it applies in both | |
| 5577 | - // paths regardless. Default OFF — re-parsing every ACF PDF on every | |
| 5578 | - // editor save is expensive and most sites don't want it. | |
| 5579 | - $autosync_extract_acf_pdfs = get_option('mxchat_auto_sync_acf_pdfs', '0') === '1'; | |
| 5580 | - if ($autosync_extract_acf_pdfs && !empty($pdf_attachment_ids)) { | |
| 5581 | - $pdf_attachment_ids = array_unique(array_filter(array_map('intval', $pdf_attachment_ids))); | |
| 5582 | - $pdf_sections = array(); | |
| 5583 | - foreach ($pdf_attachment_ids as $att_id) { | |
| 5584 | - $pdf_text = $this->mxchat_extract_pdf_text_by_attachment_id($att_id); | |
| 5585 | - if (!empty($pdf_text)) { | |
| 5586 | - $pdf_title = get_the_title($att_id); | |
| 5587 | - $pdf_url = wp_get_attachment_url($att_id); | |
| 5588 | - $header = 'PDF Attachment'; | |
| 5589 | - if (!empty($pdf_title)) { | |
| 5590 | - $header .= ': ' . $pdf_title; | |
| 5591 | - } | |
| 5592 | - if (!empty($pdf_url)) { | |
| 5593 | - $header .= ' (' . $pdf_url . ')'; | |
| 5594 | - } | |
| 5595 | - $pdf_sections[] = $header . "\n" . $pdf_text; | |
| 5596 | - } | |
| 5597 | - } | |
| 5598 | - if (!empty($pdf_sections)) { | |
| 5599 | - $final_content .= "\n\nAttached PDFs:\n" . implode("\n\n", $pdf_sections); | |
| 5600 | - } | |
| 5601 | 5332 | } |
| 5602 | 5333 | } |
| 5603 | 5334 | |
| 5604 | 5335 | // ADD CUSTOM POST META SUPPORT (whitelisted non-ACF meta fields) |
