Define custom kernels (notebook)¶

• Using a uv virtual environment

uv managed virtual environment¶

We can use uv to create a virtual environment for different versions of Python.

Let's create a new virtual environment in the folder named ".venv_py315" using uv.

Requires a recent version of uv to be installed.

In [1]:

%uv venv .venv_py315 --python 3.15 --clear
%uv pip install --directory .venv_py315 --python 3.15 async-kernel

%uv venv .venv_py315 --python 3.15 --clear %uv pip install --directory .venv_py315 --python 3.15 async-kernel

Downloading cpython-3.15.0a5-linux-x86_64-gnu (download) (35.1MiB)

Now we can write a kernel spec that uses uv to start the kernel from the virtual environment.

In [2]:

import pathlib

from async_kernel.kernelspec import write_kernel_spec

uv_path = pathlib.Path.cwd().joinpath(".venv_py315")
assert uv_path.exists()

write_kernel_spec(
    kernel_name="async_3.15",
    display_name="Python 3.15 (async)",
    env={"UV_PROJECT_ENVIRONMENT": str(uv_path)},
    executable=("uv", "run", "--no-sync", "python", "-m", "async_kernel"),
)

import pathlib from async_kernel.kernelspec import write_kernel_spec uv_path = pathlib.Path.cwd().joinpath(".venv_py315") assert uv_path.exists() write_kernel_spec( kernel_name="async_3.15", display_name="Python 3.15 (async)", env={"UV_PROJECT_ENVIRONMENT": str(uv_path)}, executable=("uv", "run", "--no-sync", "python", "-m", "async_kernel"), )

Out[2]:

PosixPath('/home/runner/work/async-kernel/async-kernel/.venv/share/jupyter/kernels/async_3.15')

The kernel spec with the display name "Python 3.15 (async kernel)" has been added. You will need to refresh the list of kernels for it to be available.