@fengrru/healix

Self-healing error classifier with dual hash matching and Hamming distance fuzzy matching.

Install

npm install @fengrru/healix

Quick Start

import { ErrorClassifier, RepairMemoryEngine } from "@fengrru/healix"

// Classify errors
const classifier = new ErrorClassifier()
const category = classifier.classify("ENOENT: no such file or directory") // "not_found"

// Learn repair rules
const engine = new RepairMemoryEngine()
engine.addRule({
  id: "r1",
  category: "not_found",
  exact_hash: "...",
  fuzzy_hash: "...",
  conditions: [{ type: "file_pattern", value: "*.ts" }],
  action: { type: "create_file", params: {} },
  success_rate: 0.9,
  occurrence_count: 10,
})

// Match rules
const match = engine.matchRules(errorInfo)

Error Categories

• not_found — file/directory not found
• permission — access denied
• timeout — operation timed out
• syntax — parse/compile errors
• resource — memory/disk limits
• network — connection failures
• research_failed — LLM/tool failures
• unknown — unrecognized errors

Features

• Dual hash matching: exact + fuzzy (Hamming distance)
• Specificity scoring: AND conditions weighted higher
• Self-learning: success_rate tracks rule effectiveness
• Time decay: retention decays over time

Documentation

• API Reference — TypeDoc-generated API docs
• Source Code
• Examples

License

MIT