Project Setup¶

By Sameer Alam · Updated 14 September 2026

What You'll Learn¶

• Why system Python and global pip install cause problems on servers
• How to install Python versions and manage projects with uv
• How pyproject.toml and a lock file make installs reproducible
• A project layout that scales from one script to a packaged tool

Why Not System Python?¶

Your operating system's Python belongs to the operating system. Package managers and system tools depend on specific library versions in it, and modern distributions mark it as externally managed so pip install fails:

error: externally-managed-environment
× This environment is externally managed

Don't work around this with --break-system-packages or sudo pip. Give every project its own isolated environment instead.

Install uv¶

uv installs Python versions, creates virtual environments, resolves and locks dependencies, and runs tools — replacing pyenv, venv, pip, pip-tools, and pipx with one fast binary.

curl -LsSf https://astral.sh/uv/install.sh | sh      # Linux and macOS
# or: brew install uv   /   pipx install uv

uv --version
uv python install 3.14
uv python list

Create a Project¶

uv init --package opsctl
cd opsctl
uv add httpx typer boto3 pyyaml jinja2 tenacity
uv add --dev pytest ruff mypy respx "moto[ec2]"

This creates:

opsctl/
├── .python-version          # the Python version for this project
├── pyproject.toml           # metadata, dependencies, tool settings
├── uv.lock                  # exact resolved versions of everything — commit it
├── README.md
└── src/
    └── opsctl/
        └── __init__.py

pyproject.toml¶

[project]
name = "opsctl"
version = "0.1.0"
description = "Operations helpers for the platform team"
readme = "README.md"
requires-python = ">=3.12"
dependencies = [
    "boto3>=1.40",
    "httpx>=0.28",
    "jinja2>=3.1",
    "pyyaml>=6.0",
    "tenacity>=9.0",
    "typer>=0.16",
]

[project.scripts]
opsctl = "opsctl.cli:app"          # installs an "opsctl" command

[dependency-groups]
dev = [
    "moto[ec2]>=5.1",
    "mypy>=1.17",
    "pytest>=8.4",
    "respx>=0.22",
    "ruff>=0.12",
]

[build-system]
requires = ["uv_build>=0.8,<1"]
build-backend = "uv_build"

• dependencies lists compatible ranges for what your code needs.
• uv.lock pins exact versions and hashes of every package, including transitive dependencies. Everyone and every CI run gets the same environment.

Daily Commands¶

uv sync                          # create .venv and install exactly what's in uv.lock
uv run opsctl --help             # run inside the project environment, no activation needed
uv run pytest
uv add rich                      # add a dependency and update the lock
uv remove rich
uv lock --upgrade-package httpx  # upgrade one dependency deliberately
uv tree                          # dependency tree

In CI, fail if the lock file is out of date instead of silently resolving new versions:

uv sync --locked

Running Standalone Tools¶

uvx ruff check .                 # run a tool in a temporary, cached environment
uv tool install ansible-core     # install a CLI tool globally, isolated from other tools
uv tool list

Single-File Scripts With Inline Dependencies¶

For a one-off script that doesn't need a whole project, declare dependencies at the top of the file (PEP 723):

# /// script
# requires-python = ">=3.12"
# dependencies = ["httpx>=0.28"]
# ///
import ssl, socket, sys
from datetime import datetime, timezone

def days_left(host: str, port: int = 443) -> int:
    ctx = ssl.create_default_context()
    with socket.create_connection((host, port), timeout=5) as sock:
        with ctx.wrap_socket(sock, server_hostname=host) as tls:
            not_after = tls.getpeercert()["notAfter"]
    expires = datetime.strptime(not_after, "%b %d %H:%M:%S %Y %Z").replace(tzinfo=timezone.utc)
    return (expires - datetime.now(timezone.utc)).days

for host in sys.argv[1:]:
    print(f"{host}: {days_left(host)} days")

uv run check_certs.py api.example.com shop.example.com

uv creates a cached environment with those dependencies automatically.

The Standard Library venv¶

When uv isn't available, the standard library works fine:

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -e ".[dev]"   # if dev dependencies are declared as an optional extra
deactivate

Always use python -m pip, so you know which interpreter you're installing into.

Project Layout¶

opsctl/
├── pyproject.toml
├── uv.lock
├── src/
│   └── opsctl/
│       ├── __init__.py
│       ├── cli.py              # command-line entry point — thin
│       ├── health.py           # logic, importable and testable
│       ├── aws_cleanup.py
│       └── config.py
└── tests/
    ├── test_health.py
    └── test_aws_cleanup.py

The src/ layout forces tests to import the installed package, catching packaging mistakes that a flat layout hides. Keep cli.py thin: parse arguments, call functions in other modules, and turn results into output and exit codes.

Deploying Automation¶

Common Mistakes¶

• sudo pip install into system Python, breaking OS tools or getting overwritten by package upgrades.
• Not committing the lock file, so production installs different versions than the ones you tested.
• Unpinned or overly broad dependencies like boto3 with no lower bound, silently pulling breaking changes.
• One giant script.py with logic, argument parsing, and AWS calls mixed together, impossible to test.
• Activating the wrong virtual environment and installing packages somewhere unexpected — uv run avoids the whole problem.

Interview Questions¶

• Why shouldn't you install packages into the system Python on a server?
• What's the difference between the dependency ranges in pyproject.toml and a lock file?
• How would you make a Python tool's dependencies reproducible in CI?
• Why use a src/ layout?

Next¶

Continue to CLI Tools and subprocess.