Python Dependency Management Strategies — ELI5

Think of dependencies like ingredients in a recipe. Your Python project needs specific libraries to work, just like a cake needs flour, sugar, and eggs. Dependency management is the system that makes sure you always get the right ingredients in the right amounts.

Without a system, chaos creeps in. One developer uses version 2 of a library, another uses version 3, and the project crashes on a third computer because it grabbed version 4 automatically. This is the classic “works on my machine” disaster.

The solution has two parts. First, you write down what you need — a list of libraries and roughly which versions are okay. Second, a tool locks the exact versions that work together and saves that snapshot. Everyone on the team uses the same snapshot, so every computer gets identical ingredients.

Python has several tools for this: pip with requirements files is the oldest approach, Poetry wraps everything in one workflow, and uv is the newest and fastest option. They all solve the same problem with different levels of convenience.

The habit to build is: always pin your dependencies, always commit the lock file, and always install from the lock file in production. Skip any of those steps and you are rolling dice on every deploy.

The one thing to remember: A lock file is a snapshot of every exact version your project needs — commit it, and deployments become predictable.