uv-jupyter-kernel

Script to configure Jupyter kernels that use the uv environment and dependency manager. It allows you to create isolated kernels for different Python versions, making it easier to use Jupyter with controlled and reproducible environments.

Requirements

• Some Jupyter client (e.g., VSCode extension)
• uv installed and available in the PATH

Usage

uvx uv-jupyter-kernel --versions 3.13 3.12

By default, the script configures kernels for Python versions 3.13 and 3.12. You can specify other versions by passing them as arguments.

This will create (or update) kernel files at ~/.local/share/jupyter/kernels/uv-<version>/kernel.json, allowing you to select the corresponding kernel for the desired Python version within Jupyter.

What does the script do?

• Locates the uv executable on the system.
• For each specified Python version, creates a Jupyter kernel that:
  • Ephemerally installs ipykernel and starts the executor itself.
  • Ensures the uv PATH is available in the kernel environment.
  • Allows each notebook to run in an independent, ephemeral, and isolated environment.

Advantages

• The only thing that needs to be available initially is uv.
• Ephemeral environments: each notebook has its own and can install new dependencies with uv pip install; when Jupyter is restarted, the environment is reset.
• No need to install Jupyter in the environment: less chance of dependency conflicts.
• Shared cache: each thing can be downloaded only once and then reused.

Disadvantages

• Higher disk usage if your system does not support hardlinks (e.g., Termux on Android)