Self-Contained TypeScript Programs Using Bun

If you hate Python as much as me it’s probably because of dependencies.

Roughly 23-319% of the time, when I run a Python app, it doesn’t work because of dependencies, and I end up trying to figure out which of my 17 real and virtual Python environments are actually active.

Using uv for everything is way better, but since I’m kind of moving my entire programming world to TypeScript, I’m now using bun‘s auto-install feature instead. And it’s actually a bit better.

Different Approaches, Same Goal

Python’s uv and Bun both solve the “self-contained app” problem by putting the requirements inside the program itself, but they do it in different ways:

Python uses special magical comments to declare dependencies:

# /// script
# dependencies = ["requests", "rich"]
# ///

import requests
# script continues...

It works great, but it feels super hack-y to me.

It feels like we’re smuggling in a dependency payload to trick Python into actually working for once.

Bun’s Approach: Just import it like normal

I like bun‘s approach much better. It just writes the imports out like it’s not embarrassed by them!

#!/usr/bin/env bun

// Just import what you need - Bun auto-installs!
import axios from 'axios';
import chalk from 'chalk';

console.log(chalk.cyan.bold('\n🚀 Bun Auto-Install Demo\n'));

// Fetch a random joke
try {
  console.log(chalk.yellow('Getting a random joke...'));
  const jokeResponse = await axios.get('https://official-joke-api.appspot.com/random_joke');
  const joke = jokeResponse.data;
  console.log(chalk.green(`\n${joke.setup}`));
  console.log(chalk.blue(`${joke.punchline} 😄\n`));
} catch (error) {
  console.log(chalk.red('Failed to fetch joke\n'));
}

// Fetch a random activity
try {
  console.log(chalk.yellow('Finding something to do...'));
  const activityResponse = await axios.get('https://bored-api.appbrewery.com/random');
  const activity = activityResponse.data;
  console.log(chalk.magenta(`Activity: ${activity.activity}`));
  console.log(chalk.dim(`Type: ${activity.type} | Participants: ${activity.participants}\n`));
} catch (error) {
  console.log(chalk.red('Failed to fetch activity\n'));
}

// Show a random number to prove it runs fresh
const randomNum = Math.floor(Math.random() * 1000);
console.log(chalk.green(`Random number: ${randomNum}`));

// Show that this runs fresh each time
console.log(chalk.dim('\nRun again for different results!'));
console.log(chalk.dim('No package.json or npm install needed 🎉\n'));

Running the Script

# Make it executable
chmod +x test.ts

# Run it directly - dependencies auto-install!
./test.ts

# Or just use bun
bun test.ts

The first time you run the script, bun automatically:

1. Detects the missing packages
2. Downloads and installs them
3. Caches them for future runs
4. Executes your script

No npm install, no package.json, no setup—just run it.

Example Output

🚀 Bun Auto-Install Demo

Getting a random joke...

Did you watch the new comic book movie?
It was very graphic! 😄

Finding something to do...
Activity: Explore a park you have never been to before
Type: recreational | Participants: 1

Random number: 294

Run again for different results!
No package.json or npm install needed 🎉

Oh, and it’s nuclear fast.

bun test.ts  0.08s user 0.06s system 29% cpu 0.469 total

My takeaway

This goes to a larger discussion around Python vs TypeScript, but I feel like this is another example where the latter is just a more natural, modern way of doing things.

TypeScript all the things.