Back to changelog

Word Files Return Plain Text

Chunks from Word (.docx) files now contain plain text, whichever processing_type indexed them.

Before, a Word file indexed with basic, or with auto when it chose basic, returned its chunk text wrapped in a record. The text started with {'page_number': 1, 'content': '... and had \n written out in place of line breaks. That wrapped text was also what Captain embedded, so search worked less well on those files.

Now Word files return plain text with basic, auto and advanced. Files indexed with advanced, and PDFs, were never affected.

Do I need to change anything?

  • Word files indexed with basic or auto before 2026-11-21: they keep the wrapped text until they are indexed again. To refresh them:
    • Uploaded files: upload them again with skip_existing=false.
    • Files indexed from a bucket: index the bucket, or that file, again with overwrite_existing=true and skip_existing=false.
  • If you wrote code to unwrap that record: keep it until every affected file has been indexed again, so it handles both forms in the meantime. Then remove it.
  • Word files indexed with advanced, and PDFs: nothing to do.