SecureVector Guardian
Copyright 2026 SecureVector

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

------------------------------------------------------------------------------
Originality
------------------------------------------------------------------------------
The source code in this repository was authored for SecureVector. The model is
trained from scratch on SecureVector's own labeled corpus; it does NOT
incorporate, fine-tune, or distribute any third-party pretrained model weights
(no DeBERTa, MiniLM, Model2Vec, Llama Guard, or other external checkpoints).
No third-party datasets are used.

The pure-Python inference runtime (src/svguardian/model/pure_infer.py)
reimplements the documented feature-extraction behavior of scikit-learn's
TfidfVectorizer (word and char_wb n-gram analyzers, sublinear tf, l2
normalization) so exported models score identically without an ML dependency.
scikit-learn is (c) the scikit-learn developers, BSD 3-Clause License
(https://github.com/scikit-learn/scikit-learn/blob/main/COPYING); this
attribution covers any expressive similarity between the reimplementation and
the original.

------------------------------------------------------------------------------
Third-party dependencies (used as libraries, not modified or vendored)
------------------------------------------------------------------------------
This software depends on the following open-source libraries, each used under
its own permissive license:

  - scikit-learn   BSD 3-Clause License
  - NumPy          BSD 3-Clause License
  - SciPy          BSD 3-Clause License
  - PyYAML         MIT License
  - joblib         BSD 3-Clause License

These libraries are not redistributed in source form by this repository; they
are installed independently via the Python package manager. Their respective
license texts are available with their distributions.
