jupyter-live-kernel

To keep Python variables persistent across executions, use a stateful Python REPL backed by a live Jupyter kernel. This approach maintains your workspace, imports, and intermediate results across multiple steps without needing to re-run everything.

You can iteratively inspect variables and edit notebook cells by running code within a specific notebook through a live Jupyter kernel. This supports repeated execution, on-the-fly inspection, and incremental refinement of your feature engineering functions.

Yes, you need JupyterLab and uv to run this stateful Python REPL. The skill uses the hamelnb script against a running local Jupyter server, requiring these specific dependencies for server discovery and JSON-structured command outputs.

The best way to prototype machine learning code without losing intermediate results is using a live Jupyter kernel. It provides a stateful environment where your dataset, variables, and imported libraries survive across multiple executions for continuous experimentation.

Yes, you can discover running Jupyter servers and create a scratch notebook for REST-based sessions. The skill handles server and session orchestration, starting a local server if needed to support your notebook-aware operations.