Metadata-Version: 2.4
Name: xenmask
Version: 0.0.0
Summary: Reversible PHI/PII de-identification round-trip for agentic LLM pipelines (placeholder — under active design)
Author: Mansib Miraj
License-Expression: Apache-2.0
Project-URL: Homepage, https://github.com/mansibmiraj/xenmask
Keywords: de-identification,deidentification,pii,phi,pseudonymization,privacy,llm,hipaa,pdpa,informatics
Classifier: Development Status :: 1 - Planning
Classifier: Intended Audience :: Healthcare Industry
Classifier: Intended Audience :: Science/Research
Classifier: Programming Language :: Python :: 3
Classifier: Topic :: Scientific/Engineering :: Medical Science Apps.
Classifier: Topic :: Security
Requires-Python: >=3.9
Description-Content-Type: text/markdown

# xenmask

**Reversible PHI/PII de-identification round-trip for agentic LLM pipelines.**

Mask sensitive fields to typed, referentially-consistent surrogate tokens before they
reach an LLM; unmask the model's output afterward — with the round-trip enforced as a
verifiable invariant (an unmapped token in the output is an error, never shipped).

> ⚠️ **0.0.0 placeholder.** This name is reserved while the v1 API is under active
> design. Not yet functional. Follow the repo for the first real release.

## Why

Static de-identification destroys the linkage you need to keep working. `xenmask`
de-identifies *reversibly* and *locally* (no SaaS, no data egress), so you can run
agentic intelligence over clinical/casework data and still get real answers back.

Licensed under Apache-2.0.
