From cd66dc168ab679f57c0f4a7947c03563ac376134 Mon Sep 17 00:00:00 2001
From: RussellLuo <luopeng.he@gmail.com>
Date: Thu, 2 Jul 2026 17:48:52 +0800
Subject: [PATCH] fix(converter): preserve docx embedded images from markitdown

Fixes #162.
---
 openkb/converter.py     |  2 +-
 tests/test_converter.py | 30 ++++++++++++++++++++++++++++++
 2 files changed, 31 insertions(+), 1 deletion(-)

diff --git a/openkb/converter.py b/openkb/converter.py
index fc3c96b5..a904ea22 100644
--- a/openkb/converter.py
+++ b/openkb/converter.py
@@ -230,7 +230,7 @@ def convert_document(
             markdown = convert_pdf_with_images(src, doc_name, images_dir)
         else:
             # Non-PDF, non-MD: use markitdown (docx, pptx, html, etc.)
-            mid = MarkItDown()
+            mid = MarkItDown(keep_data_uris=True)
             result = mid.convert(str(src))
             markdown = result.text_content
             markdown = extract_base64_images(markdown, doc_name, images_dir)
diff --git a/tests/test_converter.py b/tests/test_converter.py
index e7c95d4d..bc9ed22c 100644
--- a/tests/test_converter.py
+++ b/tests/test_converter.py
@@ -130,6 +130,36 @@ def test_long_pdf_returns_is_long_doc(self, kb_dir, tmp_path):
         assert result.raw_path is not None
 
 
+# ---------------------------------------------------------------------------
+# convert_document — MarkItDown-backed formats
+# ---------------------------------------------------------------------------
+
+
+class TestConvertDocumentMarkItDown:
+    def test_docx_conversion_enables_keep_data_uris(self, kb_dir, tmp_path):
+        src = tmp_path / "report.docx"
+        src.write_bytes(b"fake docx")
+
+        mock_result = MagicMock()
+        mock_result.text_content = "![](data:image/png;base64,abc123)"
+
+        with (
+            patch("openkb.converter.MarkItDown") as mock_markitdown,
+            patch("openkb.converter.extract_base64_images", return_value="converted markdown") as mock_extract,
+        ):
+            mock_markitdown.return_value.convert.return_value = mock_result
+
+            result = convert_document(src, kb_dir)
+
+        mock_markitdown.assert_called_once_with(keep_data_uris=True)
+        mock_markitdown.return_value.convert.assert_called_once_with(str(src))
+        mock_extract.assert_called_once()
+        assert result.skipped is False
+        assert result.is_long_doc is False
+        assert result.source_path is not None
+        assert result.source_path.read_text(encoding="utf-8") == "converted markdown"
+
+
 # ---------------------------------------------------------------------------
 # _registry_path
 # ---------------------------------------------------------------------------