SnipSmart

Extract clean JSON or HTML/XML snippets from messy AI outputs or raw text. A tiny, dependency‑free utility built for MCPs, agents, and AI‑native applications.

✨ Features

• snipSmart -- Forgiving parser that extracts JSON or HTML/XML without throwing

• snipSmartOrThrow -- Strict variant that throws unless parsing is fully successful

• Handles:

  • Missing braces
  • Markdown fences
  • Leading/trailing noise
  • Broken or partial tag structures

📦 Installation

bash

npm install snipsmart

🔧 Usage

SnipSmart exposes two functions:

• snipSmart → forgiving, never throws
• snipSmartOrThrow → strict, throws on invalid or partial results

✅ Option 1: Use snipSmart (recommended)

js

import snipSmart from "snipsmart";

const input = `Broken JSON: { "name": "Alice", "age": 30 `;

const result = snipSmart(input, "json"); // or { format: "json" }

console.log(result);
/*
{
  status: "check",
  comments: "Added missing closing brace. Parsing passed",
  data: { name: "Alice", age: 30 },
  raw: null
}
*/

snipSmart never throws --- perfect for AI pipelines and debugging.

✅ Option 2: Use snipSmartOrThrow (strict)

js

import { snipSmartOrThrow } from "snipsmart";

try {
  const data = snipSmartOrThrow(`{ "x": 1 `, "json");
  console.log(data);
} catch (err) {
  console.log("Error:", err.message);
}

This variant throws unless:

Code

status === "success"

Ideal for validation, tests, and deterministic workflows.

✅ Extract HTML/XML

js

const tagInput = `
Random text
<div><p>Hello <b>World</b></p></div>
More text
`;

const result = snipSmart(tagInput, "tag");

console.log(result.data);
// "<div><p>Hello <b>World</b></p></div>"

📄 Return Object

Both JSON and TAG modes return:

ts

{
  status: "success" | "check" | "fail",
  comments: string,
  data: any | string | null,
  raw: string | null
}

• success → fully valid
• check → recovered with minor fixes (e.g., missing brace)
• fail → extraction failed

snipSmartOrThrow throws on "check" and "fail".

🧪 Examples

The /examples folder contains runnable examples.

Run one:

bash

node examples/use-snipSmart.js

🧩 Test Cases

The /testCases folder includes curated cases for:

• JSON extraction
• Tag extraction
• Unified parsing behavior

Files include:

• snipJson-cases.md
• snipByTag-cases.md
• snipSmart-cases.md

🚀 Roadmap

Planned future extractors:

• snipCsv -- Extract CSV tables
• snipCode -- Extract code blocks
• Additional formats for broader AI‑output parsing

🤝 Contributing

Ideas, edge cases, or improvements are welcome. Open an issue or submit a pull request!

🔓 License

MIT © Shiny Sherbina Free for personal and commercial use --- just keep the license notice.

🌟 Why SnipSmart?

AI models often wrap structured data in explanations, markdown fences, or conversational fluff. SnipSmart extracts only the useful part, making it ideal for:

• MCP tools
• Agent pipelines
• LLM‑driven backends
• Any workflow requiring clean, predictable structured output

Lightweight, deterministic, and built for real‑world AI engineering.