An asynchronous Python Jupyter kernel built on AnyIO:

• runs in a separate process or in-process.
• supports top-level await.

For a standalone, out-of-process kernel, install the subprocess extra:

pip install "akernel[subprocess]"

This will give you a Python 3 (akernel) in JupyterLab.

akernel can also run in-process with Jupyverse. Install it with:

pip install "fps-akernel-task"

This will give you both Python 3 (akernel-task) and Python 3 (akernel-thread) kernels in JupyterLab.

To refresh the kernelspecs in an existing installation, run:

akernel install --mode process
akernel install --mode task
akernel install --mode thread

The default mode is process.

akernel-task runs in Jupyverse's process, so running blocking user code in the kernel will also block Jupyverse.

akernel-thread is an in-process kernel that runs user code in a separate thread, which won't block Jupyverse.

• Cell output redirection currently uses the kernel's print function rather than capturing all writes to stdout and stderr.
• In-process kernels cannot be interrupted while running blocking code.