Course Notebooks¶

Twenty complete, executed Jupyter notebooks — one per major course topic. Each notebook contains markdown theory, learning objectives, explanation before code, beginner → intermediate examples, exercises, a challenge exercise, a recap, and review questions. Every code cell was executed and verified end-to-end (see How these were verified below).

#	Notebook	Module	Key tools
01	Python for Data Science	A	core Python
02	NumPy	A	numpy
03	Pandas	A	pandas
04	Data Cleaning	A	pandas
05	Data Aggregation	A	pandas (groupby/merge)
06	Data Visualization	A	matplotlib, seaborn
07	EDA	A	pandas, seaborn
08	Git/GitHub Workflow	A	git (run inside the notebook)
09	APIs and Data Acquisition	A	requests, JSON
10	Introduction to ML	B	scikit-learn
11	Regression	B	scikit-learn
12	Classification	B	scikit-learn
13	Clustering	B	scikit-learn
14	PandasAI	C	pandasai (guarded)
15	LLM Fundamentals	C	prompt patterns (guarded)
16	Ollama with Python	C	ollama (guarded)
17	Tool/Function Calling	C	ollama + schemas (guarded)
18	AI Agents	C	agent loop (guarded)
19	n8n Workflows	C	workflow-in-Python pattern
20	End-to-End Data Science Project	A+B+C	full pipeline

How to use¶

python -m venv .venv
.venv/Scripts/pip install -r requirements.txt     # Windows
# or: source .venv/bin/pip install -r requirements.txt   (macOS/Linux)
.venv/Scripts/jupyter lab course-notebooks/01-python-for-data-science.ipynb

Run notebooks in order — later notebooks reuse patterns from earlier ones.

Guards and degraded modes¶

Notebooks 14–18 depend on optional, LLM-related software that may not be installed or running on a given machine:

pandasai (14): installed separately; on Python 3.14 it does not build. Cells import it inside a guard and print a message instead of failing.
Ollama (16–18): requires the Ollama app running with a model pulled (ollama pull llama3.2). Guarded cells degrade gracefully, and the mock/simulation paths always run, so the notebooks teach the full workflow with or without a model.
Notebook 09 & 19 call the keyless Open-Meteo weather API; cells cache results in datasets/ so reruns work offline.

Anything a notebook saves (CSVs, figures, JSON) goes into a datasets/ folder created next to the notebook on first run. That folder is git-ignored.

How these were verified¶

scripts/build_course_notebooks.py generates the .ipynb files from readable source modules in scripts/notebook_src/. scripts/execute_notebooks.py executes every notebook with a real kernel and writes outputs back into the files. Current status: 20/20 notebooks execute without errors on Python 3.14 with the pinned requirements.txt stack.

To rebuild or re-verify after editing a source module:

.venv/Scripts/python scripts/build_course_notebooks.py
.venv/Scripts/python scripts/execute_notebooks.py

Continuous integration & read-only checking¶

GitHub Actions runs the same validation on every push and pull request that touches the notebooks (.github/workflows/notebooks.yml). CI executes all 20 notebooks with a real kernel and fails if any cell raises an error.

Two properties matter:

Notebooks are never modified by CI. scripts/check_notebooks.py executes each notebook in memory and discards the outputs, so the .ipynb files on disk are untouched (the workflow double-checks this with git diff).
No optional services are required. The Ollama daemon is not installed in CI; notebooks 14–18 detect that and run their mock/simulation paths. No API keys or paid services are used anywhere.

To run the same check locally before pushing (nothing is modified):

.venv/Scripts/python -m pip install -r requirements.txt nbclient nbformat
.venv/Scripts/python scripts/check_notebooks.py          # all notebooks
.venv/Scripts/python scripts/check_notebooks.py 03 12    # a subset

Use scripts/execute_notebooks.py instead only when you intentionally want to re-execute and save outputs back into the delivered files.

Instructor note: exercise cells ship with # your code here placeholders followed by worked solutions — delete the solution cells before distributing if you want the exercises unsolved.