OpenKB currently converts .docx files via MarkItDown() and then runs extract_base64_images(...) on the generated markdown.
However, for .docx files with embedded images, MarkItDown needs keep_data_uris enabled to preserve full base64 image content. Otherwise the output may contain incomplete image refs like , which means OpenKB cannot extract and persist those images afterward.
Relevant upstream context:
Suggested fix:
- enable
keep_data_uris in the MarkItDown-based .docx conversion path before calling extract_base64_images(...)
OpenKB currently converts
.docxfiles viaMarkItDown()and then runsextract_base64_images(...)on the generated markdown.However, for
.docxfiles with embedded images, MarkItDown needskeep_data_urisenabled to preserve full base64 image content. Otherwise the output may contain incomplete image refs like, which means OpenKB cannot extract and persist those images afterward.Relevant upstream context:
Suggested fix:
keep_data_urisin the MarkItDown-based.docxconversion path before callingextract_base64_images(...)