NOTICE

Extralit HF Space NOTICE
========================

This NOTICE file provides licensing and usage-boundary information for third‑party
software included in (or used by) this repository / container image, with special
attention to components licensed under the GNU Affero General Public License
(AGPL) and how they are isolated from the core Extralit server codebase.

-------------------------------------------------------------------------------
Project Licensing Summary
-------------------------------------------------------------------------------
1. Core Extralit Server & Related Source (outside the isolated extraction
   microservice) are licensed under the Apache License, Version 2.0 (Apache-2.0).

2. The PDF Extraction Microservice located in this repository (see `extract.py`
   under `extralit_ocr/`) intentionally imports and uses:
   - PyMuPDF (a/k/a `fitz`) – Licensed under the GNU Affero General Public
     License, version 3 (AGPL-3.0).
   - pymupdf4llm – Also distributed under AGPL-3.0 (it depends on PyMuPDF and
     provides higher-level LLM / Markdown extraction utilities).

3. No other Extralit Apache-2.0 modules import or link against PyMuPDF or
   pymupdf4llm. Their usage is *confined to a distinct, optional runtime
   process* (a FastAPI application) intended to run as an isolated service
   (e.g., a separate process invoked via Procfile or a Unix Domain Socket).

-------------------------------------------------------------------------------
Isolation / Usage Boundary
-------------------------------------------------------------------------------
- The AGPL-licensed libraries (PyMuPDF, pymupdf4llm) are only loaded within
  the dedicated extraction microservice defined by `extract.py`.
- The main Extralit application code under Apache-2.0 can function without
  directly importing or distributing derivative code based on PyMuPDF; it
  communicates over an internal HTTP or Unix Domain Socket boundary.
- This design aims to limit AGPL copyleft obligations *to the components that
  actually combine with or run AGPL code*, while keeping the primary server
  codebase under Apache-2.0. (Consult qualified legal counsel for definitive
  compliance guidance—this document is not legal advice.)

If you deploy the extraction service publicly over a network (SaaS scenario),
AGPL Section 13 requires that you offer the Corresponding Source for the
AGPL-covered components (and any modifications you made to them).

-------------------------------------------------------------------------------
Obtaining Source Code for AGPL Components
-------------------------------------------------------------------------------
Upstream sources (unmodified in this repository) are available at:
- PyMuPDF:       https://github.com/pymupdf/PyMuPDF
- pymupdf4llm:   https://github.com/pymupdf/pymupdf4llm

The minimal glue / integration code we authored that *uses* these libraries
and produces hierarchical Markdown extraction is contained in:
- `extralit_ocr/extract.py`

That file is provided under Apache-2.0 (for our original portions). However,
because it imports and runs AGPL libraries in-process, distributing a compiled
artifact or offering a network service that executes that code may trigger AGPL
obligations (notably providing corresponding source, including our modifications
to that file and any other changes that form a combined work). We have not
modified the PyMuPDF or pymupdf4llm libraries themselves inside this repository.

-------------------------------------------------------------------------------
How to Remove / Replace AGPL Components
-------------------------------------------------------------------------------
If you prefer not to use AGPL-licensed code:
1. Exclude the extraction microservice from your build (remove or comment out
   its Procfile entry, or do not run `uvicorn app:app ...`).
2. Replace it with an alternative PDF text extraction mechanism using permissive
   licenses (e.g., pdfminer.six, Apache Tika, or other tools) and adjust the
   calling code accordingly.
3. Ensure no remaining imports of `fitz` or `pymupdf4llm` persist in the Apache-2.0
   layers of your deployment.

-------------------------------------------------------------------------------
Runtime Notes
-------------------------------------------------------------------------------
- The extraction process may run via a Unix Domain Socket (UDS) for local-only
  access, reinforcing that only internal trusted components can invoke AGPL
  functionality.
- Environment variables (documented in `extract.py`) control size limits, logging,
  and optional writing of extracted Markdown for debugging.

-------------------------------------------------------------------------------
Trademark & Attribution
-------------------------------------------------------------------------------
"PyMuPDF" and related marks are the property of their respective owners.
All third-party licenses and trademarks are acknowledged.

-------------------------------------------------------------------------------
NO WARRANTY
-------------------------------------------------------------------------------
All software is provided "AS IS", without warranty of any kind, express or
implied, including but not limited to the warranties of merchantability,
fitness for a particular purpose, and noninfringement.

-------------------------------------------------------------------------------
Contact
-------------------------------------------------------------------------------
For questions about how this repository structures license boundaries or to
request a copy of any corresponding source we have modified relating to the
AGPL components, contact: legal@extralit.ai (placeholder – adjust as needed).

-------------------------------------------------------------------------------
License Text References
-------------------------------------------------------------------------------
- Apache License 2.0: https://www.apache.org/licenses/LICENSE-2.0
- AGPL-3.0: https://www.gnu.org/licenses/agpl-3.0.en.html

End of NOTICE