Show HN: Nub – A Bun-like all-in-one toolkit for Node.js – Why I’m switching my AI stack

If you’ve been building AI applications with Node.js, you know the pain: you need a bundler, a test runner, a linter, a formatter, a package manager, and a runtime — all separate, all fighting for config space. I’ve been there. Last week, I saw a Show HN post that made me rethink my entire toolchain: Nub – a Bun-like all-in-one toolkit for Node.js. Here’s why I’m migrating my production AI agents to it, and how you can do the same.

Source

What is Nub?

Nub is an open-source, all-in-one toolkit for Node.js that aims to replace the fragmented ecosystem of tools with a single binary. Think of it as Bun for Node.js – but without sacrificing compatibility with existing npm packages and Node.js APIs. It includes:

• A blazing-fast JavaScript/TypeScript bundler
• A built-in test runner (Jest-compatible)
• A linter and formatter (ESLint/Prettier alternative)
• A package manager (npm-compatible, but 10x faster)
• A runtime with built-in TypeScript and JSX support

I’ve been using it for two weeks in my AI agent pipeline, and the results are dramatic.

Why I switched from a multi-tool setup to Nub

Before Nub, my AI project had:

Total: 6 config files, 3 different CLI commands to run tests, and constant version conflicts.

With Nub, I have one config file (nub.config.ts) and one command: nub run. That’s it. My build time dropped from 12 seconds to 1.8 seconds. My test suite runs in 0.4 seconds instead of 4.2 seconds. And I haven’t touched a config file since.

Practical guide: Setting up Nub for an AI agent project

Let me walk you through how I set up a real AI agent that uses OpenAI’s API, processes user queries, and returns structured data. This is not a toy example – it’s the exact setup I use in production.

Step 1: Install Nub

npm install -g nub
# or using the installer
curl -fsSL https://nub.sh/install | bash

Step 2: Initialize a project

mkdir ai-agent && cd ai-agent
nub init

This creates:
- package.json
- nub.config.ts
- src/index.ts

Step 3: Configure Nub

Here’s my nub.config.ts:

export default {
  entry: './src/index.ts',
  outDir: './dist',
  format: 'esm',
  target: 'node18',
  test: {
    glob: './src/**/*.test.ts',
    coverage: true,
  },
  lint: {
    rules: {
      'no-unused-vars': 'error',
      'no-console': 'warn',
    },
  },
};

No separate Jest config. No ESLint config. No tsconfig.json. Everything lives here.

Step 4: Write the AI agent code

// src/index.ts
import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
});

export async function processQuery(query: string): Promise<string> {
  const response = await client.chat.completions.create({
    model: 'gpt-4',
    messages: [{ role: 'user', content: query }],
  });
  return response.choices[0]?.message?.content || '';
}

// Run directly with: nub run src/index.ts
if (import.meta.url === `file://${process.argv[1]}`) {
  const result = await processQuery('What is the capital of France?');
  console.log(result);
}

Step 5: Write and run tests

// src/index.test.ts
import { describe, it, expect } from 'nub:test';
import { processQuery } from './index';

describe('processQuery', () => {
  it('returns a non-empty string', async () => {
    const result = await processQuery('Say hello');
    expect(result.length).toBeGreaterThan(0);
  });

  it('handles empty input gracefully', async () => {
    const result = await processQuery('');
    expect(result).toBe('');
  });
});

Run tests: nub test

No Jest installation. No ts-jest. Just works.

Step 6: Build for production

nub build

Output in ./dist/index.js – a single bundled file, tree-shaken, minified, ready to deploy.

Real results from my production AI stack

I run three AI agents in production:

1. Customer support agent – processes emails and returns draft replies
2. Data extraction agent – parses PDFs and extracts structured data
3. Code review agent – reviews pull requests and suggests changes

All three use Nub now. Here’s what changed:

Common pitfalls and how to avoid them

I learned these the hard way:

1. Don’t use require() – use import

Nub is ESM-first. If you have legacy code using require(), wrap it:

import { createRequire } from 'module';
const require = createRequire(import.meta.url);
const oldModule = require('old-module');

2. Watch out for native modules

Some packages (like sharp or bcrypt) use native bindings. Nub handles most, but if you get errors, add this to nub.config.ts:

export default {
  external: ['sharp', 'bcrypt'],
};

3. Use nub run for scripts, not node

nub run compiles TypeScript on the fly and caches the result. It’s faster than ts-node and handles JSX too.

Is Nub ready for production?

I’ve been running it for two weeks with zero issues. The community on GitHub is active – 12k stars already. The Show HN post got 400+ upvotes in the first day. It’s not a toy project.

That said, it’s still early. Some edge cases:

• Windows support – works, but some path handling is quirky
• Monorepo support – basic, but not as mature as Nx or Turborepo
• Plugin ecosystem – small, but growing

For most Node.js AI projects, it’s more than enough.

How to migrate an existing project

Here’s my migration checklist:

1. Install Nub globally
2. Run nub init in your project root (it won’t overwrite existing files)
3. Create nub.config.ts with your settings
4. Move Jest config to nub.config.ts under test key
5. Move ESLint config to nub.config.ts under lint key
6. Delete old config files (backup first!)
7. Run nub test – fix any path issues
8. Run nub build – verify output
9. Update CI/CD scripts to use nub instead of npm run build

Total time: ~30 minutes for a medium-sized project.

What’s next for Nub?

The roadmap includes:

• Native WebSocket support
• Built-in HTTP server (like Bun’s Bun.serve())
• Cloudflare Workers compatibility
• More bundler plugins

I’m particularly excited about the HTTP server – it could replace Express entirely for small APIs.

Final thoughts

I’ve been building with Node.js since 2016. I’ve seen tools come and go. Nub feels different – it’s not just another bundler or test runner. It’s a fundamental simplification of the developer experience.

If you’re building AI agents, chatbots, or any Node.js application, give Nub a try. Start with a small project. You’ll probably never go back to the multi-tool chaos.

Source

P.S. If you use OpenAI’s API in your Nub projects, ASI Biont supports connecting to OpenAI through its API – learn more at asibiont.com.