@whatitbroke/ai

Optional AI Augmented Reasoning Layer with Multi-Provider Connectors and Zero-Leak Redaction Gate

@whatitbroke/ai is an opt-in augmented reasoning package for WhatItBroke. It connects your debug context to LLM providers (Google Gemini, OpenAI, Anthropic Claude) to provide human-like explanations and alternative patch recommendations for obscure or complex crashes.

🔒 Privacy & Zero-Leak Redaction Gate

CRITICAL GUARANTEE: WhatItBroke never sends raw code or runtime data externally by default. Even when @whatitbroke/ai is active, every single prompt passes through the strict Redactor first:

• All passwords, API tokens, Authorization headers, and cookies are stripped.
• Database connection string credentials and JWTs are sanitized.
• Only normalized execution paths and anonymized code snippets reach the LLM.

Installation

npm install @whatitbroke/ai @whatitbroke/shared

Supported Providers

• Google Gemini: gemini-1.5-flash, gemini-1.5-pro
• OpenAI: gpt-4o-mini, gpt-4o
• Anthropic Claude: claude-3-5-sonnet
• Mock Provider: Zero network calls, perfect for offline development and CI/CD pipelines.

Usage

1. Using the Factory (createAIAnalyzer)

import { createAIAnalyzer } from '@whatitbroke/ai';

// Example: Google Gemini
const analyzer = createAIAnalyzer({
  provider: 'gemini',
  apiKey: process.env.GEMINI_API_KEY,
  model: 'gemini-1.5-flash'
});

// Example: OpenAI
const openaiAnalyzer = createAIAnalyzer({
  provider: 'openai',
  apiKey: process.env.OPENAI_API_KEY,
  model: 'gpt-4o-mini'
});

// Analyze sanitized debug context:
const result = await analyzer.analyze(debugContext);

console.log(result.explanation);
console.log(result.patch); // Unified code diff
console.log(`AI Confidence: ${result.confidence}%`);

2. Custom Prompt Formatting (PromptBuilder)

If you want to inspect or customize the sanitized prompt sent to the LLM:

import { PromptBuilder } from '@whatitbroke/ai';

const prompt = PromptBuilder.buildDebuggingPrompt(debugContext);
console.log(prompt);

Configuration (whatitbroke.config.json)

You can enable this layer directly in your CLI configuration file:

{
  "framework": "node",
  "ai": {
    "enabled": true,
    "provider": "gemini",
    "model": "gemini-1.5-flash",
    "apiKey": "process.env.GEMINI_API_KEY"
  }
}

License

MIT © WhatItBroke Team