kintsugi-math

The mathematics of beautiful error recovery — for JavaScript/TypeScript.

In kintsugi, broken pottery is repaired with gold lacquer, making the cracks beautiful rather than hidden. This library brings that philosophy to software: transform errors into golden seams, reassemble fragmented data, map crack propagation graphs, measure wabi-sabi aesthetics, and test resilience.

Install

npm install kintsugi-math

Quick Start

import {
  goldenRepair,
  repairTraceback,
  FragmentCollection,
  entropyOfErrors,
  goldenRatio,
  aestheticScore,
  buildCrackGraph,
  findGoldenJoints,
  shortestCrackPath,
  injectFaults,
  measureRecovery,
} from "kintsugi-math";

// Transform an error into a golden seam
const seam = goldenRepair(new TypeError("not a function"));
console.log(seam.severity);    // number 0-10
console.log(seam.position);    // "TypeError: not a function"
console.log(seam.repairQuality); // "Ensure correct argument types..."

// Reassemble fragmented data
const coll = new FragmentCollection();
coll.add([1, 2, 3], 0.9, 0);
coll.add([7, 8], 0.7, 5);
const whole = coll.reassemble(); // fills gaps with interpolation

// Measure wabi-sabi aesthetics
const entropy = entropyOfErrors(["TypeError", "RangeError", "TypeError"]);
const beauty = goldenRatio(1.618, 1.0); // ≈ 1.0
const score = aestheticScore([1, 2, null, 4]); // 0 < score < 1

// Build crack propagation graphs
const graph = buildCrackGraph([
  { source: "database", target: "api" },
  { source: "api", target: "frontend" },
]);
const joints = findGoldenJoints(graph); // highest-impact repair points
const path = shortestCrackPath(graph, "database", "frontend");

// Test resilience
const { results, faultCount } = injectFaults(() => compute(), 0.2);
const report = measureRecovery(recoverFromError, [err1, err2, err3]);
console.log(report.recoveryRate); // 0-1

API

Types

• GoldenSeam — { severity, position, repairQuality } — A repair point extracted from an error.
• Fragment — { data, confidence, offset } — A piece of partial data.
• CrackGraph — { nodes, edges, errorCounts, edgeWeights } — Directed error propagation graph.
• GoldenJoint — { node, impact, beauty } — High-impact repair point in a crack graph.
• WabiSabiReport — { entropy, ratio, score, imperfections, description } — Imperfection metrics.
• ResilienceReport — { recoveryRate, meanTimeToRepair, aestheticQuality, totalFaults, successfulRecoveries } — Recovery stats.

goldenRepair

• goldenRepair(error) → GoldenSeam
• repairTraceback(error) → GoldenSeam[]
• severityScore(error) → number (0–10)

FragmentCollection

• new FragmentCollection()
• .add(data, confidence, offset)
• .sort() — sort by confidence (descending)
• .reasemble() → number[] — fills gaps with confidence-weighted interpolation
• .length, .totalConfidence, .averageConfidence

wabiSabi

• entropyOfErrors(errors) → number (Shannon entropy in bits)
• goldenRatio(recovered, lost) → number (0–1, closeness to φ)
• aestheticScore(data) → number (0–1)

cracks

• buildCrackGraph(errors) → CrackGraph
• findGoldenJoints(graph) → GoldenJoint[]
• shortestCrackPath(graph, source, target) → string[] | null

resilience

• injectFaults(fn, faultRate) → { results, faultCount }
• measureRecovery(system, faults) → ResilienceReport

License

MIT