react-native-signal-scorer-lite

A lightweight signal-scoring Nitro module for React Native. It inspects a numeric window of time-series samples and returns trend direction, volatility, confidence, and a composite score — without any Rust dependency. The algorithm lives directly in C++ (with a pure-TS fallback), so it is simple to build and easy to reason about.

What it does

Given a numeric signal array, it returns:

• mean — average value
• slope — trend direction
• volatility — standard deviation
• score — composite in [0, 1]
• confidence — in [0, 1]
• prediction — steady, accelerating, or volatile

Differences from the full signal-scorer

• No Rust — the scoring math is implemented directly in C++ (cpp/HybridSignalScorerLite.cpp).
• No JNI bridge — only the Nitro HybridObject path is provided.
• No named registry — stateless, so there is no reset() method.
• Same public API — evaluate, score, evaluateBatch, and scoreSignal.

Architecture

TypeScript (spec) -> C++ HybridObject (algorithm inlined)
                    -> TS fallback (when native unavailable)

Getting started

npm install
npm run typecheck

API

import { signalScorerLite, scoreSignal } from 'react-native-signal-scorer-lite'

const result = signalScorerLite.score([0.2, 0.5, 0.7, 1.1, 1.3])
// { prediction: 'accelerating', confidence: 0.8, score: 0.85, slope: 0.275, volatility: 0.4, mean: 0.76 }

const batch = signalScorerLite.evaluateBatch([0.2, 0.5, 0.7, 1.1, 1.3], 3)
// Array<SignalPrediction>