fix: handle None values in extracted table cells when converting to markdown and improve empty table detection.

data_cutter.py

@@ -17,13 +17,15 @@ def load_documents():
             tables = page.extract_tables() or []
             table_texts = []
             for table in tables:
-                if not table:
+                if not table or not table[0]:
                     continue
-                # Convert table to markdown
-                md_table = "| " + " | ".join(table[0]) + " |\n"
-                md_table += "| " + " | ".join(["---"]*len(table[0])) + " |\n"
+                # Convert table to markdown - handle None values in cells
+                header_row = [str(cell) if cell is not None else "" for cell in table[0]]
+                md_table = "| " + " | ".join(header_row) + " |\n"
+                md_table += "| " + " | ".join(["---"]*len(header_row)) + " |\n"
                 for row in table[1:]:
-                    md_table += "| " + " | ".join([str(cell) if cell else "" for cell in row]) + " |\n"
+                    row_cells = [str(cell) if cell is not None else "" for cell in row]
+                    md_table += "| " + " | ".join(row_cells) + " |\n"
                 table_texts.append(md_table)
             full_page = text + "\n\n" + "\n\n".join(table_texts)
             documents.append(Document(page_content=full_page, metadata={"page": i+1}))