Appendix A — Tooling reference

This appendix is a one-stop reference for the tools the book uses, grouped by the job they do. For each, it gives what the tool is for, the command or entry point that gets you started, and the chapter that introduces it. You do not need all of them, and you certainly do not need them all at once — reach for each when the problem it solves actually appears in your work. Where two tools do a similar job, the more modern or widely used default is listed first.

A.1 Version control and collaboration

Tool	What it’s for	Where to start	Ch.
Git	Versioning code and recording the history of decisions	git init, git add, git commit	2
nbdime	Notebook-aware diffs and merges (readable .ipynb changes)	nbdiff notebook.ipynb	2
Jupytext	Pairing a notebook with a plain-text script that diffs cleanly	jupytext --set-formats ipynb,py:percent nb.ipynb	2
nbstripout	Stripping notebook outputs before commit	nbstripout --install	2
GitHub (or GitLab)	Hosting repositories, pull requests, and code review	open a pull request	17

A.2 Environments and packaging

Tool	What it’s for	Where to start	Ch.
venv	Isolating a project’s Python environment	python -m venv .venv	3
pip-tools	Compiling an abstract spec into a pinned lockfile	pip-compile requirements.in	3
uv	A fast, modern alternative for environments and locking	uv pip compile, uv venv	3
conda / mamba	Environments with heavy binary or GPU dependencies	conda create -n proj	3
pyproject.toml	Declaring an installable package and its dependencies	pip install -e .	6

A.3 Code quality

Tool	What it’s for	Where to start	Ch.
ruff	Fast linting and formatting in one tool	ruff check ., ruff format .	5
black	Opinionated code formatting	black .	5
mypy	Static type checking from your type hints	mypy src/	5

A.4 Testing, debugging, and profiling

Tool	What it’s for	Where to start	Ch.
pytest	Running tests; the standard test runner	pytest	7
Hypothesis	Property-based testing (generates inputs for you)	@given(...)	7
pdb	Interactive debugging at a breakpoint	breakpoint()	8
logging	Levelled, structured diagnostics that scale past print	logging.getLogger(__name__)	8
cProfile / line_profiler	Finding where the time actually goes	python -m cProfile -s cumtime script.py	8

A.5 Project structure and pipelines

Tool	What it’s for	Where to start	Ch.
cookiecutter	Scaffolding a standard project layout	cookiecutter <template>	9
make	A simple task runner and dependency graph	a Makefile with named targets	4, 10
Snakemake / Prefect / Dagster	Orchestrating complex, scheduled pipelines with retries	declare stages and dependencies	10
pandera / Great Expectations	Validating a DataFrame against a declared schema	schema.validate(df)	10

A.6 Configuration, secrets, and APIs

Tool	What it’s for	Where to start	Ch.
pydantic	Typed, validated configuration and data models	class Config(BaseModel): ...	11, 12
PyYAML	Reading and writing YAML configuration files	yaml.safe_load(...)	11
python-dotenv	Loading secrets from a local, untracked .env	load_dotenv()	11
Hydra	Composing and sweeping over experiment configurations	@hydra.main(...)	11
FastAPI	Serving a model behind a typed HTTP API	@app.post("/predict")	12
uvicorn	The server that runs a FastAPI app	uvicorn main:app	12
httpx	Making HTTP requests; backs FastAPI’s TestClient	httpx.post(...)	12

A.7 Operations: containers, CI/CD, deployment, monitoring

Tool	What it’s for	Where to start	Ch.
Docker	Packaging the whole environment as a portable image	a Dockerfile, docker build	14
docker-compose	Running several services together as one stack	compose.yaml, docker compose up	14
GitHub Actions	Running tests and checks automatically on every change	.github/workflows/ci.yml	13
pre-commit	Fast local checks before a commit is recorded	.pre-commit-config.yaml, pre-commit install	13
cron / Airflow	Scheduling batch jobs (simple to complex)	a cron entry, or an Airflow DAG	15
Prometheus / Grafana	Collecting and dashboarding service metrics	scrape /metrics, build a dashboard	16
Evidently	Off-the-shelf data and prediction drift reports	compare a reference to live data	16

A.8 Documentation, data, and model versioning

Tool	What it’s for	Where to start	Ch.
MkDocs / Sphinx / Quarto	Generating documentation from your code and prose	mkdocs serve / quarto render	18
DVC	Versioning large data and models alongside Git	dvc add data/raw/...	22
MLflow	Tracking experiments and registering model versions	mlflow.log_metric(...), the model registry	23
joblib	Serialising a trained model to a portable artefact	joblib.dump(model, path)	15, 21

A closing note in the spirit of the book: this list is deliberately a menu, not a checklist. A throwaway analysis needs almost none of it; a model that real users depend on eventually touches most of it. The skill is choosing the smallest set of tools that makes a given piece of work reliable enough for what it has to do — and reaching for the next one only when the problem it solves is the problem you actually have.