Metadata-Version: 2.4
Name: hbp100
Version: 0.1.5
Summary: Ultra-light privacy firewall for LLM prompts
Author-email: Dipanjan Dutta <relic.of.sky@gmail.com>
License-Expression: MIT
Project-URL: Repository, https://github.com/Xpheonix0/humming-bird-/
Keywords: privacy,firewall,nlp,ml,pii
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Security
Classifier: Topic :: Text Processing
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: joblib>=0.1.0
Requires-Dist: scikit-learn>=1.0.0
Dynamic: license-file

# hbp100

Ultra-light privacy firewall for LLM prompts.

`hbp100` detects and masks common personally identifiable information (PII)
before text is sent to an LLM or another downstream service.

## Install

```bash
pip install hbp100
```

## Usage

```python
from hbp100 import sanitize

result = sanitize("Email john@example.com and OTP 123456")

print(result.text)
# Email [EMAIL_1] and OTP [OTP_1]

print(result.metadata)
# {'[OTP_1]': '123456', '[EMAIL_1]': 'john@example.com'}
```

## What It Masks

- Email addresses, phone numbers, URLs with tokens, API keys, bearer tokens,
  JWTs, passwords, OTPs, PINs, credit cards, IBANs, UPI IDs, government IDs,
  account identifiers, patient IDs, IMEI numbers, addresses, and similar data.
- Context-aware years: horoscope and calendar-conversion years can be kept
  while identity-related years are masked.

## License

MIT
