Metadata-Version: 2.4
Name: intentseal-core
Version: 0.1.0
Summary: IntentSeal Core Engine: prompt-injection firewall and action guard for LLM agents
Author-email: Sankalp Wanjari <sankalpwanjari85@gmail.com>
License-Expression: Apache-2.0
Project-URL: Homepage, https://github.com/sankalp2515/intentseal
Project-URL: Source, https://github.com/sankalp2515/intentseal
Project-URL: Documentation, https://github.com/sankalp2515/intentseal/blob/main/docs/INTEGRATION.md
Project-URL: Issues, https://github.com/sankalp2515/intentseal/issues
Keywords: prompt injection,llm security,ai agents,guardrails,firewall,rag
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Security
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.11
Description-Content-Type: text/markdown
License-File: LICENSE
License-File: NOTICE
Requires-Dist: httpx>=0.27
Requires-Dist: pydantic>=2.6
Requires-Dist: pyyaml>=6.0
Requires-Dist: python-dotenv>=1.0
Requires-Dist: anyascii>=0.3
Provides-Extra: engine
Requires-Dist: beautifulsoup4>=4.12; extra == "engine"
Requires-Dist: pypdf>=4; extra == "engine"
Requires-Dist: pdfplumber>=0.11; extra == "engine"
Requires-Dist: python-docx>=1.1; extra == "engine"
Requires-Dist: ftfy>=6.1; extra == "engine"
Requires-Dist: confusable-homoglyphs>=3.2; extra == "engine"
Requires-Dist: detect-secrets>=1.4; extra == "engine"
Provides-Extra: classifiers
Requires-Dist: onnxruntime>=1.17; extra == "classifiers"
Requires-Dist: tokenizers>=0.15; extra == "classifiers"
Requires-Dist: huggingface_hub>=0.23; extra == "classifiers"
Requires-Dist: numpy>=1.26; extra == "classifiers"
Provides-Extra: ocr
Requires-Dist: rapidocr-onnxruntime>=1.3; extra == "ocr"
Requires-Dist: pillow>=10; extra == "ocr"
Provides-Extra: postgres
Requires-Dist: psycopg[binary]>=3.2; extra == "postgres"
Provides-Extra: anthropic
Requires-Dist: anthropic>=1.12; extra == "anthropic"
Provides-Extra: all
Requires-Dist: intentseal-core[classifiers,engine,ocr]; extra == "all"
Dynamic: license-file

# intentseal-core

The engine behind [IntentSeal](https://pypi.org/project/intentseal/), a prompt-injection firewall for AI agents:
content extraction for every format (PDF, Word, HTML, email, Markdown, JSON, code, images via OCR), normalisation and
decoding, a three-tier detector (structural rules, a local classifier, an AI judge), session risk and data-flow
tracking, the action guard and the decision log.

Most users install the SDK instead:

```bash
pip install intentseal                 # SDK, remote mode
pip install "intentseal[embedded]"     # SDK with this engine in-process
```

Extras of this package: `engine` (extraction, normalisation, secret scan), `classifiers` (local ONNX classifier),
`ocr`, `postgres`, `anthropic`, `all`.

It also installs the `intentseal` command: `intentseal init` writes a starter policy; `intentseal providers check`
lists the configured model providers and tests each key (key values are never printed).

Source and documentation: https://github.com/sankalp2515/intentseal · Licensed under the Apache License 2.0.
