Package Exports

@uvrn/drift
@uvrn/drift/dist/index.js

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (@uvrn/drift) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

@uvrn/drift

Temporal decay scoring for UVRN verification receipts.

Models how a claim's confidence score degrades over time using configurable decay curves — because a receipt from 6 hours ago on a fast-moving claim shouldn't carry the same weight as one from 6 minutes ago.

Package provides: computeDrift, DriftMonitor, computeDriftFromInput; built-in profiles and decay curves; types (DriftReceipt, DriftSnapshot, AgentDriftReceipt, etc.). Score-to-status (STABLE / DRIFTING / CRITICAL).

You provide: Receipt (or drift input) to score; optionally your own decay profile. No storage — you consume threshold events (e.g. alert, feed agent/canon).

Install

npm install @uvrn/drift @uvrn/core @uvrn/score

@uvrn/core and @uvrn/score are required peer dependencies. @uvrn/score owns the canonical V-Score weights; @uvrn/drift imports WEIGHTS from it.

Quick start

import { computeDrift, DRIFT_PROFILES } from '@uvrn/drift';

const result = computeDrift(receipt, DRIFT_PROFILES.fast);

console.log(result.drift.decayed_score); // e.g. 61 (was 88, 24h ago)
console.log(result.drift.status);        // 'DRIFTING'
console.log(result.drift.delta);         // -27

Decay curves

Curve	Behaviour	Good for
`LINEAR`	Fixed pts/hr drop	Slow-moving, long-lived claims
`SIGMOID`	Holds, then hard drop at midpoint	Claims with binary freshness — current or not
`EXPONENTIAL`	Immediate aggressive decay	Fast-moving claims, volatile data

Built-in profiles

fast · moderate · threshold_short · threshold_long · slow · archival · default

These are generic starting points covering each curve type at different speeds. Build your own DriftProfile for domain-specific tuning:

import type { DriftProfile } from '@uvrn/drift';

const myCustomProfile: DriftProfile = {
  name: 'my_domain_profile',
  curve: 'SIGMOID',
  rate: 48,               // midpoint at 48 hours
  staleAfterHours: 120,
  scoreFloor: 10,
};

const result = computeDrift(receipt, myCustomProfile);

Continuous monitoring

import { DriftMonitor, DRIFT_PROFILES } from '@uvrn/drift';

const monitor = new DriftMonitor({
  intervalMs: 60_000, // check every minute
  onThreshold: (event) => {
    console.log(`${event.receipt_id}: ${event.from} → ${event.to}`);
    // emit alert receipt, notify @uvrn/agent, etc.
  },
});

monitor.watch(receipt, DRIFT_PROFILES.threshold_short);
monitor.start();

V-Score formula

V-Score = (Completeness × 0.35) + (Parity × 0.35) + (Freshness × 0.30)

Weights are owned by @uvrn/score. Drift decays the Freshness component only (via computeDrift). Completeness and Parity are re-scored by @uvrn/agent when new sources are fetched.

License

MIT · uvrn.org