Practical Exercises

How-To for redoing the exercises on your own laptop after the training.

Download and unpack the notebooks¶

• Download sessions_download.zip (right-click > “Save Link As”) - see explaination for Pitfall below!

• Unpack it. You get a folder sessions_download/ with:

  • pyproject.toml

  • secrets

  • session1/

  • session2/

Create the virtual environment (venv)¶

• Open the terminal and navigate into the unpacked folder

cd /path/to/sessions_download

• Create the venv and install all packages

uv sync

• This creates a local .venv folder and a uv.lock file inside sessions_download/

Fill your secrets file¶

• Open the file secrets in a text editor and fill in your values:

  • ENDPOINT_URL - the LLM API endpoint URL (https://...)

  • OPENAI_API_KEY - your API key (sk-...)

  • ZENODO_SANDBOX_TOKEN - Personal access token for Zenodo Sandbox

• Leave SYSTEM_PROMPT_DATA_STEWARD as provided

NOTE: Treat the secrets file like a password - never commit or share it.

Start JupyterLab with your secrets¶

• From the sessions_download/ folder, start JupyterLab so it loads secrets as environment variables

uv run --env-file secrets jupyter lab

• Work through the notebooks under session1/notebooks/ and session2/notebooks/

• To stop JupyterLab, press Ctrl + C in the terminal

Pitfall: cloud-synced and network folders¶

IMPORTANT: Do not place sessions_download/ inside any synced folder (OneDrive, Dropbox, Google Drive, iCloud), or other remote network drives.

On many managed laptops these locations apply stricter execution restrictions and constant file syncing, which can:

• block running the Python interpreter/scripts installed in .venv (so uv sync, uv run, uv add fails or hangs), and

• corrupt the environment as files get locked or re-synced mid-install.

Use a plain local folder outside any synced or networked path instead, e.g. C:\Users\<you>\eosc-ssds-2026 (Windows) or ~/eosc-ssds-2026 (macOS/Linux).