Suggestion: include Xberg (formerly Kreuzberg) in the benchmark

It might be interesting to include Xberg in the comparison. It seems particularly comparable to anydoc since it also has a Rust core, runs locally, supports the same Office/document formats, and can produce Markdown without requiring hosted services.

For the fairest comparison, Xberg could probably be run with the simplest extraction/Markdown backend/configuration i.e. without optional OCR, ML layout, or enrichment features unless those are also enabled for the other tools.

Would you be open to adding it as another benchmark backend and posting the results?

The issue does not name files or tests; start by locating the benchmark backend configuration and the existing comparison runs. Add Xberg with its simplest extraction/Markdown configuration, keeping optional OCR, ML layout, and enrichment disabled unless the other tools use them, then post the results.

Written by the indexing model from the issue text.