snipSmart – AI Cleanup Tool

shiny sherbinashiny sherbina
3 min read

AI has become our assistant in almost everything. Personally, I love using AI to learn — it helps me verify my understanding and feels almost like talking to a human.

But when it comes to GenAI-output, things can get frustrating. We just want clean output — no explanations, no extra text, just the data.


The Problem

Recently, I was building a candidate engagement bot. I needed to extract { skill: proficiency } pairs from a job description using AI. Despite my strict prompts —

"The output should strictly be a JSON. No additional explanations or text."

AI still responded with something like:

"Sure! Here is the JSON you asked: { skill: proficiency }"

So I had to write fallback logic just to extract the actual JSON and ignore the fluff.


The Solution: snipSmart

That’s when snipSmart was born.

It’s a lightweight, open-source utility that lets you extract clean JSON or HTML/XML snippets from messy AI-generated content. No dependencies. Just copy the function(s) you need and drop them into your project.

You pass in a string — it could be an AI response, raw content, or any noisy input — and the tool:

  • Cleans out irrelevant text

  • Extracts the valid JSON or tag-based content

  • Ensures it can be parsed safely (e.g., with JSON.parse)

If the result status is "success", you can confidently use it as-is in your code. No more parsing errors or weird edge cases.


How to Use snipSmart

There are 3 ways to use the tool, depending on what you need:

This unified function internally calls snipJson or snipByTag based on the format you provide ("json" or "tag").

import snipSmart from "./src/index.js";

const input = `Broken JSON: { "name": "Alice", "age": 30 `;
const result = snipSmart(input, { format: "json" });  // Or format : "tag"

console.log(result);

✅ Option 2: Use snipJson Directly

Use this if you only need JSON cleanup.

import snipJson from "./src/snipJson.js";

const input = `Here's the object you asked for: { "id": 1, "status": "active" } 😊`;
const result = snipJson(input);

console.log(result);

✅ Option 3: Use snipByTag Directly

Use this for HTML/XML snippets.

import snipByTag from "./src/snipByTag.js";

const input = `Some markup: <div><p>Hello <b>World</p>`;
const result = snipByTag(input);

console.log(result);

All functions return an object like:

{
  status: "success" | "fail" | "check",
  comments: "Info or error message",
  data: <extracted_snippet_or_null>,
  raw: <optional raw fragment>
}

🧪 Test Cases

Here are a few cases from the /testCases folder.

InputFormatOutputStatus
Hello, here's your data: { "a": 1, "b": 2 }json{ "a": 1, "b": 2 }
Here is the info: <note><to>User</to></note>tag<note><to>User</to></note>
{ "a": [1, 2 }, "b": 3jsonFails due to bracket mismatch
Just some text, nothing validjsonNo valid JSON structure found
<div><span>Text</div>tagMismatched closing tag: expected </span>but found </div>

You can view all test cases here


snipSmart is free, open-source, and built with simplicity in mind.
Feel free to use the code in your own projects, suggest improvements, or contribute new features.

More snips are in the pipeline — stay tuned!

Check out the GitHub repo

📜 This tool is licensed under the MIT License. If you reuse or share the code, please retain the license notice.

0
Subscribe to my newsletter

Read articles from shiny sherbina directly inside your inbox. Subscribe to the newsletter, and don't miss out.

Written by

shiny sherbina
shiny sherbina