Metadata-Version: 2.4
Name: lia-platform
Version: 0.0.2
Summary: Lia — propose-only AI sentinel family for repository audits. Ships lia-scan, a local secret-hygiene check that detects credentials accidentally committed to a git history so they can be rotated before abuse. Lia sentinels never auto-act: every finding is a proposal for human review.
Requires-Python: >=3.12
Requires-Dist: lia-core>=0.0.1
Description-Content-Type: text/markdown

# lia-platform

Umbrella package for the **Lia sentinel family** — propose-only AI auditors that
review a repository and surface findings for a human to approve. Lia never
auto-acts: every outward action is a proposal awaiting human review
(`lia-core`, `lia-gro`, `lia-code`, …).

## v0.0.2 — secret-hygiene preview

This release ships one real tool: **`lia-scan`** — a local secret-hygiene check
that detects credentials accidentally committed to a repository's git history
(API keys, tokens, `.env` material, private keys) and prints a one-line severity
score so you can rotate them before they're abused. It is a preview of the Lia
auto-engineer, not the full auditor (which arrives in later releases).

## Install

```bash
uv pip install lia-platform
# or: pip install lia-platform
```

Requires Python ≥ 3.12. Pulls in `lia-core` — the family's constitutional core
(6 invariants: never-auto-act, capability-confinement, immutable audit-log,
delivery-contract, manifest-conformance, semver; guarded executor; registry).

## Use

```bash
lia-scan                  # check the current repo
lia-scan /path/to/repo    # check a specific repo
python -m lia --version
```

Exit codes: `0` = clean, `1` = potential credentials found, `2` = not a git
repository / git error.

## Status

`0.0.2` pre-freeze. The Lia platform is under active development; the full
auditor + sentinel family arrive in later releases.