Clean LLM JSON

A tiny utility to clean and parse messy JSON outputs from LLMs (Large Language Models) like ChatGPT, Claude, or Gemini.

These models often wrap JSON in extra formatting like triple backticks or "```json\n{...}\n```", which breaks JSON.parse(). This package strips the noise and gives you a usable JavaScript object.

🚀 Installation

npm install cleanllmjson

🔧 Usage

import { cleanJSON } from 'cleanjson';

// Example with LLM-formatted JSON string
const input = "```json\n{\n  \"name\": \"Ali\",\n  \"age\": 22\n}\n```";
const result = cleanJSON(input);

console.log(result);
// Output: { name: 'Ali', age: 22 }

It also works with regular JSON strings and even plain objects (it just returns them as-is).

📦 What It Does

• Trims leading/trailing spaces and newlines
• Removes triple backtick formatting (json ... )
• Safely parses JSON
• Returns null if parsing fails

🧪 Example

cleanJSON(`{
  "success": true,
  "data": [1, 2, 3]
}`);

// → { success: true, data: [1, 2, 3] }

⚠️ Note

• Only use with trusted inputs. Don’t run cleanJSON() on unverified data from users without additional validation.
• If parsing fails, it logs the error and returns null.

📄 License

MIT © captainAlpha04