Skip to content

Software and environment

jupyter.vse.cz — the environment used in the seminars

jupyter.vse.cz is a JupyterHub with JupyterLab 4, Python 3.12 and the whole data stack pre-installed. Log in with your school account.

  • Seminar notebooks are in the read-only exchange folder for the course; copy them to your home directory before you run them.
  • Assignments are fetched, validated and submitted through the Assignments tab (nbgrader). Choose the course 4iz566-582_ZS2627 there. Validate runs the visible checks only; passing it is not a guarantee of full marks.
  • Before submitting anything: Kernel → Restart Kernel and Run All Cells must run clean.

Language-model access — LiteLLM

Seminars 3 and 4 use only the school LiteLLM proxy, with API base URL https://litellm.vse.cz/v1/. No external account or credit card is needed. The openai Python package serves as a compatible client for this proxy.

The course API key is shared by all students of the course and is already set in the setup cell of the seminar-3 and seminar-4 notebooks; no further configuration is needed. Do not pass the key on outside the course. LLM_API_KEY in the kernel environment overrides it.

The default models are qwen3.6-35b for chat and qwen3-embedding:4b for embeddings. Use LLM_CHAT_MODEL and LLM_EMBED_MODEL to select other models available to your key. If the proxy cannot be reached, the notebooks demonstrate the workflow using simulated or replayed responses; this does not measure live model quality.

AI coding assistant — IBM Bob

The course comes with free tokens for IBM Bob, IBM's AI coding assistant. Access details are given out in the seminars. Use it under the AI-tools policy: allowed for the term project and the home assignments, switched off during all tests.

Working on your own computer

Python 3.12 and these packages:

numpy pandas scipy scikit-learn matplotlib jupyterlab openai

Either install Anaconda or create a virtual environment and pip install the list above. For the seminar-3 extra notebook add llm-feature-gen; for the seminar-4 deployment demo add anvil-uplink. PyTorch and the deep-learning tooling are introduced by ÚTIA in their block, with their instructions.

Editors: JupyterLab (what the seminars use), Visual Studio Code with the Python extension, or PyCharm.

Documentation you will keep open