Environment Setup¶

This is a one-time setup. After this you can run every notebook in the course locally, with no paid API keys required (we use open-source models).

1. Prerequisites¶

Python 3.10–3.12 (python --version)
git
~3–5 GB free disk space (embedding models download on first use)
A CPU is fine for everything in the course. A GPU just makes it faster.

2. Create a virtual environment¶

cd "Search Semantically"
python -m venv .venv

# Activate it:
source .venv/bin/activate        # macOS / Linux
# .venv\Scripts\activate         # Windows (PowerShell)

3. Install dependencies¶

pip install --upgrade pip
pip install -r requirements.txt

First run note: the first time you load an embedding model (e.g. sentence-transformers/all-MiniLM-L6-v2), it downloads (~90 MB) and caches locally.

4. Launch Jupyter¶

jupyter lab        # or: jupyter notebook

Open any chapter's notebooks/ folder and run the cells top to bottom.

5. Verify your install¶

Run this in a Python shell or a notebook cell:

from sentence_transformers import SentenceTransformer
model = SentenceTransformer("all-MiniLM-L6-v2")
print(model.encode("hello world").shape)   # -> (384,)

If that prints (384,), you're ready.

Optional: the LLM for the RAG chapters¶

Chapters 10–11 use a text-generation model. We default to a small open-source model so it runs locally. If you'd rather use a hosted API, each RAG notebook notes where to swap it in — but it's never required to complete the course.

Troubleshooting¶

pip is slow or fails on torch — install the CPU build first: pip install torch --index-url https://download.pytorch.org/whl/cpu, then re-run pip install -r requirements.txt.
Out of memory — close other notebooks; the small models here need < 2 GB RAM.
Behind a proxy / offline — pre-download models on a connected machine; Hugging Face caches under ~/.cache/huggingface.