OmniSeek
Copyright 2026 The OmniSeek authors

This product is licensed under the Apache License, Version 2.0 (see LICENSE).

------------------------------------------------------------------------------
Third-party dependency licenses
------------------------------------------------------------------------------

The CORE install (`pip install -e .`) depends only on permissively-licensed
packages (Apache-2.0 / MIT / BSD / PSF / Unlicense), e.g. mcp, httpx,
beautifulsoup4, lxml, feedparser, playwright, curl_cffi, pydantic, numpy,
sentence-transformers, yt-dlp, python-pptx/docx, openpyxl, pillow,
rapidocr-onnxruntime.

OPTIONAL extras pull in heavier or differently-licensed packages; install them
only if you accept their terms:

  - omniseek[pdf]  -> PyMuPDF (fitz) is AGPL-3.0 (or a commercial license
    from Artifex). It is NOT part of the core install precisely so the default
    engine stays Apache-clean. If you install [pdf] AND redistribute a
    network-accessible service built on it, AGPL section 13 may require you to
    offer the corresponding source to your users. Choose this extra knowingly.

  - omniseek[asr]  -> funasr + imageio-ffmpeg (transcription). funasr pulls
    PyTorch (large). imageio-ffmpeg vendors an ffmpeg binary; ffmpeg components
    carry their own (LGPL/GPL) terms depending on build.

  - omniseek[walled] -> patchright (stealth Playwright fork) + xhshow (an
    anti-bot request signer). The xhshow-backed source is OPT-IN and is NOT in
    the default source pack; see SECURITY.md (access-tier posture)
    and the per-source legal notes before enabling any walled/grey source.

------------------------------------------------------------------------------
Operator responsibility
------------------------------------------------------------------------------

This project publishes an ENGINE; it does not run a service for you and does not
adjudicate whether your access to any given source is lawful where you deploy it.
YOU are responsible for the sources you enable, the sites you point this at, your
own credentials/accounts, and compliance with each site's Terms of Service and
the law of your jurisdiction(s). Walled/login sources are OFF by default and
require you to bring your own account. See LICENSE sections 7-8 (no warranty,
no liability) and SECURITY.md.
