jupyter-notebook

Use a live Jupyter kernel through the hamelnb script's execute command, which keeps variables, imports, and objects alive across calls. Point it at a notebook path and pass code with the --code flag; state persists until the kernel restarts.

Use a Jupyter kernel for iterative exploration, data science work, and tasks where you build up state incrementally. Use one-shot execution for stateless scripts, and the terminal for shell commands, builds, and installs.

You need uv installed, JupyterLab installed via uv tool install, and a running Jupyter server. You also must create a kernel session through the Jupyter REST API before executing code against a notebook.

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

Force zmq transport by adding --transport zmq to the execute command. This uses jupyter_client directly and bypasses broken websocket reply channels where the kernel actually ran but the response never arrived.

POST requests to /api/sessions fail with '_xsrf argument missing' when XSRF checks are enabled. Start the server with --ServerApp.disable_check_xsrf=True for REST-only local agent access.