jupyter-live-kernel

Use the jupyter_live_kernel.py script from hamelnb with the execute subcommand, passing --path to a notebook and --code with your Python code. State persists across calls because the same live kernel handles every execution.

Use a live kernel when you need state to persist across steps, such as iterative data exploration, DataFrame inspection, or building up complex code incrementally. Use one-shot execution for stateless scripts that finish in a single run.

You need uv installed, JupyterLab installed via uv tool install jupyterlab, and a running Jupyter server. You also need a notebook file with an active kernel session created through the Jupyter REST API.

The kernel needs a moment to initialize after the server starts, so the first execution may exceed the 30-second default timeout. Retry the execution, or pass a larger value such as --timeout 120 for setup or heavy computation.

Use the variables subcommand with --path pointing to your notebook, followed by list to see all live variables or preview --name to inspect a specific one. Add --compact to reduce verbose JSON output.