new debug functions + 2000 chunk size

documents_prep.py

@@ -174,31 +174,33 @@ def format_table_header(doc_id, table_identifier, table_num, table_title, sectio
     content += f"ТАБЛИЦА: {table_identifier}\n"
     
     # Extract and emphasize the connection type if present
-    connection_type = ''
     if table_title:
         content += f"НАЗВАНИЕ ТАБЛИЦЫ: {table_title}\n"
         
-        # Parse type from title - ADD MORE VARIANTS
+        # Parse type from title (e.g., "С-25" from "Тип сварного соединения С-25")
         import re
-        type_match = re.search(r'[СУUTC]-?\s*\d+(?:-\d+)?', table_title)
+        type_match = re.search(r'[СУUTC]-?\d+(?:-\d+)?', table_title)
         if type_match:
-            connection_type = type_match.group(0).replace(' ', '')
-            # Normalize: always use С (Cyrillic)
-            connection_type = connection_type.replace('C', 'С').replace('c', 'С')
+            connection_type = type_match.group(0)
             content += f"ТИП СОЕДИНЕНИЯ: {connection_type}\n"
-            # ADD SEARCHABLE KEYWORDS
-            content += f"КЛЮЧЕВЫЕ СЛОВА: {connection_type} тип сварного соединения\n"
     
-    # Also check table_identifier for type
-    if not connection_type and table_identifier:
-        import re
-        type_match = re.search(r'[СУUTC]-?\s*\d+', table_identifier)
-        if type_match:
-            connection_type = type_match.group(0).replace(' ', '')
-            connection_type = connection_type.replace('C', 'С')
-            content += f"ТИП СОЕДИНЕНИЯ: {connection_type}\n"
-            content += f"КЛЮЧЕВЫЕ СЛОВА: {connection_type} тип сварного соединения\n"
+    if table_num and table_num != table_identifier:
+        content += f"НОМЕР ТАБЛИЦЫ: {table_num}\n"
+    
+    if section:
+        content += f"РАЗДЕЛ ДОКУМЕНТА: {section}\n"
     
+    content += f"\n{'='*70}\n"
+    
+    # Add headers with better formatting
+    if headers:
+        content += "СТОЛБЦЫ ТАБЛИЦЫ:\n"
+        for i, h in enumerate(headers, 1):
+            content += f"  {i}. {h}\n"
+        content += "\n"
+    
+    content += "ДАННЫЕ ТАБЛИЦЫ:\n"
+    return content
 
 
 def format_single_row(row, idx):