Dify Extractor

Dify Extractor converts uploaded documents into Markdown plus structured document records for
RAG pipelines and workflows.

Supported formats

• PDF ()
• Microsoft Word ()
• Microsoft PowerPoint ()
• Microsoft Excel (, )
• Markdown (, , )
• HTML (, )
• CSV ()
• JSON ()
• YAML (, )
• Plain-text formats such as , , , , , , and

Text MIME types are also accepted when a file has no recognized extension. Unsupported binary
formats return a clear error instead of being decoded as text.

Outputs

• The text output contains the complete extracted Markdown.
• contains format-appropriate records: rows for CSV/Excel, pages for PDF, sections for
  Markdown, and a whole-file record for the other formats.
• is emitted when embedded or referenced images are successfully uploaded to Dify.

Malformed, empty, undecodable, or resource-limit-breaking files return one error message and do
not emit partial document variables.

Image and archive safeguards

Remote Markdown and DOCX images are imported on a best-effort basis. Only HTTP(S) URLs are used,
with a 30-second request timeout, a maximum of 20 images, a 15 MiB per-image limit, and a 100 MiB
cumulative limit. A failed image download or upload does not discard otherwise extractable text.

ZIP-based Office files are rejected before parsing if they contain more than 10,000 entries,
expand beyond 200 MiB, or exceed a 1,000:1 compression ratio.

Development

The plugin targets Python 3.12 and uses for dependency management.