fix(docx): tolerate styles missing type

packages/markitdown/src/markitdown/converter_utils/docx/pre_process.py

@@ -7,6 +7,9 @@
 
 from .math.omml import OMML_NS, oMath2Latex
 
+WORD_NS = "{http://schemas.openxmlformats.org/wordprocessingml/2006/main}"
+WORD_STYLE_TYPE_ATTR = WORD_NS + "type"
+
 MATH_ROOT_TEMPLATE = "".join(
     (
         "<w:document ",
@@ -115,6 +118,14 @@ def _pre_process_math(content: bytes) -> bytes:
     return str(soup).encode()
 
 
+def _pre_process_styles(content: bytes) -> bytes:
+    root = ET.fromstring(content)
+    for style in list(root.findall(f".//{WORD_NS}style")):
+        if WORD_STYLE_TYPE_ATTR not in style.attrib:
+            root.remove(style)
+    return ET.tostring(root, encoding="utf-8", xml_declaration=True)
+
+
 def pre_process_docx(input_docx: BinaryIO) -> BinaryIO:
     """
     Pre-processes a DOCX file with provided steps.
@@ -136,6 +147,9 @@ def pre_process_docx(input_docx: BinaryIO) -> BinaryIO:
         "word/footnotes.xml",
         "word/endnotes.xml",
     ]
+    pre_process_style_files = [
+        "word/styles.xml",
+    ]
     with zipfile.ZipFile(input_docx, mode="r") as zip_input:
         files = {name: zip_input.read(name) for name in zip_input.namelist()}
         with zipfile.ZipFile(output_docx, mode="w") as zip_output:
@@ -150,6 +164,12 @@ def pre_process_docx(input_docx: BinaryIO) -> BinaryIO:
                     except Exception:
                         # If there is an error in processing the content, write the original content
                         zip_output.writestr(name, content)
+                elif name in pre_process_style_files:
+                    try:
+                        updated_content = _pre_process_styles(content)
+                        zip_output.writestr(name, updated_content)
+                    except Exception:
+                        zip_output.writestr(name, content)
                 else:
                     zip_output.writestr(name, content)
     output_docx.seek(0)

packages/markitdown/tests/test_module_misc.py

@@ -3,7 +3,9 @@
 import os
 import re
 import shutil
+import zipfile
 import pytest
+from xml.etree import ElementTree as ET
 from unittest.mock import MagicMock
 
 from markitdown._uri_utils import parse_data_uri, file_uri_to_path
@@ -274,6 +276,38 @@ def test_docx_equations() -> None:
     assert block_equations, "No block equations found in the document."
 
 
+def test_docx_missing_style_type_does_not_crash(tmp_path) -> None:
+    docx_file = os.path.join(TEST_FILES_DIR, "test.docx")
+    malformed_docx = tmp_path / "missing-style-type.docx"
+
+    namespace = "http://schemas.openxmlformats.org/wordprocessingml/2006/main"
+    style_type_attr = f"{{{namespace}}}type"
+
+    with zipfile.ZipFile(docx_file, mode="r") as zip_input:
+        with zipfile.ZipFile(malformed_docx, mode="w") as zip_output:
+            zip_output.comment = zip_input.comment
+            removed_style_type = False
+
+            for item in zip_input.infolist():
+                content = zip_input.read(item.filename)
+                if item.filename == "word/styles.xml":
+                    root = ET.fromstring(content)
+                    for style in root.findall(f".//{{{namespace}}}style"):
+                        if style_type_attr in style.attrib:
+                            del style.attrib[style_type_attr]
+                            removed_style_type = True
+                            break
+                    content = ET.tostring(root, encoding="utf-8", xml_declaration=True)
+
+                zip_output.writestr(item, content)
+
+    assert removed_style_type
+
+    result = MarkItDown().convert(str(malformed_docx))
+
+    assert "AutoGen: Enabling Next-Gen LLM Applications" in result.text_content
+
+
 def test_input_as_strings() -> None:
     markitdown = MarkItDown()