mdcx
Copyright 2026 Jorge Ellena G.

This product includes software developed by Jorge Ellena G.

Anyone redistributing this software, with or without modification, must retain this
notice, as required by section 4 of the Apache 2.0 licence.

HOW IT WAS WRITTEN

Conceived and directed by Jorge Ellena G.; programmed with the assistance of Claude
(Anthropic). Every design decision in this package was made against measurements over a
real document collection, and those measurements are recorded alongside the decisions
they justify.

Copyright is held by Jorge Ellena G. Assistance from a tool does not create legal
co-authorship, any more than a compiler or an editor does; it is declared here because it
is accurate and because it is worth knowing how something one adopts was made.

DEPENDENCIES

The package relies on the following libraries, each under its own licence:

  Docling y docling-core (IBM)          MIT
  pypdfium2                             BSD-3-Clause / Apache-2.0
  python-docx, openpyxl, python-pptx    MIT
  onnxruntime, rapidocr-onnxruntime     MIT / Apache-2.0
  cryptography                          Apache-2.0 / BSD-3-Clause
  mcp (Model Context Protocol SDK)      MIT

PyMuPDF was deliberately avoided: its AGPL-3.0 licence would require any program using
it to be published under AGPL, including one offered only as a network service.
pypdfium2 covers the same needs without that condition.
