What is npm (and pip)?

// 01What a package manager actually does

A package is a bundle of code someone else wrote and published. A package manager is the tool that downloads and installs those bundles for you. It also keeps track of which ones you have, so you can remove or update them later.

The difference is that package managers work for code libraries — pieces of code that other programs use internally — rather than apps you open with a click. You install a library once, and then any code in your project can use it.

// 02npm vs pip — how to know which one to use

The rule is simple: follow the language. If the guide you're following is a Python project, you'll use pip. If it's a JavaScript/Node.js project, you'll use npm. The guide will make it obvious — the install command in the README tells you which one.

npm — for JavaScript / Node.js

npm stands for Node Package Manager. It ships with Node.js, so if you install Node.js you automatically have npm too. Almost all web-based AI tools, Next.js apps, and browser-extension projects use npm.

pip — for Python

pip is Python's package manager. It comes bundled with Python 3. Most AI/ML tools — Whisper, LangChain, Stable Diffusion, PyTorch — are Python packages installed with pip.

// 03The two files that describe what's installed

Every project that uses packages has one of these files. They're not code — they're manifests: lists of which packages the project needs and at what version. When someone else clones your project, they run one command and the manifest tells their package manager exactly what to install.

package.json (npm projects)

This file lives in the root of every npm project. When you see a repo with a package.json, you know: run npm install and you'll have everything. You rarely edit this by hand — npm install <package-name> adds the entry automatically.

{
  "name": "my-ai-project",
  "dependencies": {
    "openai": "^1.3.5",
    "next": "^14.0.0"
  }
}

requirements.txt (pip projects)

Same idea for Python. One package per line, optional version pinned. Run pip install -r requirements.txt and the whole project installs. Most AI tool repos include this file.

openai>=1.3.0
whisper==1.1.10
torch>=2.0.0
numpy

// 04How to read an install command

Every install command follows the same pattern. Once you see it, you can read any guide's install section without needing to think.

npm   install   openai   @1.3.5
 ↑       ↑        ↑        ↑
tool  action  package  version (optional)

pip   install   openai==1.3.5
 ↑       ↑        ↑
tool  action  package + version

The version is optional. Without it, you get the latest stable version, which is almost always what you want. Guides only pin a version when a newer one breaks something — they'll tell you why.

• →npm install <package> — installs into the current project folder
• →npm install -g <package> — installs globally (available everywhere, not just this project). Rare — only when a guide says -g.
• →pip install <package> — installs into the current Python environment
• →pip install --upgrade <package> — upgrades an existing package

// 05Now you. Run your first install command.

Five minutes. We'll check if npm and pip are already on your machine (they probably are), then install one real package each. No project needed — we're just practicing the motion.