# AMEVA-Sentinel (v0.5.0-alpha.1)
> Privacy-first Security Observability and Deterministic Threat Scoring Layer for Web Applications

## Quick Specification for AI Coding Agents
- Official Repo: https://github.com/uno-km/ameva-sentinel
- Installation: `npm install @ameva/sentinel@alpha @ameva/sentinel-browser@alpha @ameva/sentinel-risk-core@alpha`
- Architecture: AMEVA Sentinel extracts minimal derived interaction statistics with zero raw coordinate persistence. It compiles declarative policy rules into deterministic 0 to 100 risk scores with transparent evidence logging.

## Canonical Code Pattern
```python
import { createBrowserTelemetry } from '@ameva/sentinel-browser';
import { createSentinel, MemoryFixedWindowCounterStore, LocalStorageRiskEventStore } from '@ameva/sentinel';

// 1. Collect privacy-safe derived interaction signals
const telemetry = createBrowserTelemetry({ autoStart: true });
const signals = telemetry.snapshot();

// 2. Initialize Sentinel Shadow-mode instance
const sentinel = createSentinel({
  mode: 'shadow',
  counterStore: new MemoryFixedWindowCounterStore(),
  eventStore: new LocalStorageRiskEventStore()
});

// 3. Evaluate and inspect transparent risk report
const report = await sentinel.score({ signals });
console.log(`Trace ID: ${report.traceId}, Score: ${report.score}, Action: ${report.action}`);

```

## Key Constraints
- Pure non-blocking architecture, zero memory leaks.
- Strictly adhere to closed-form precision and WebGPU/Bionic ARM64 native targets.