{
  "context": "Evaluates how the solution leverages JupyterLab notebook and services APIs to build, execute, and persist notebooks for the requested workflow. Checks focus on constructing notebook models, starting kernel-backed sessions, sequencing execution, capturing results, and saving the executed document.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Notebook model",
      "description": "Creates the notebook using @jupyterlab/notebook primitives (e.g., NotebookModel or NotebookPanel) bound to the requested kernel name instead of manually crafting nbformat JSON.",
      "max_score": 20
    },
    {
      "name": "Cell insertion",
      "description": "Populates markdown and code cells in order through notebook APIs such as NotebookModel.sharedModel.insertCell or NotebookActions.insertBelow/insertAbove rather than editing raw arrays.",
      "max_score": 20
    },
    {
      "name": "Session setup",
      "description": "Starts and manages the kernel session via @jupyterlab/services (ServiceManager.sessions.startNew or SessionContext.initialize/ready) using the provided kernelName and reuses the session for all cell executions.",
      "max_score": 20
    },
    {
      "name": "Sequential execution",
      "description": "Runs code cells sequentially through the kernel connection (kernel.requestExecute or sessionContext.session?.kernel?.requestExecute), awaits future.done, collects IOPub output messages, and stops or continues based on allowErrors.",
      "max_score": 25
    },
    {
      "name": "Saving outputs",
      "description": "Persists the executed notebook to outputPath with captured outputs using notebook-aware saving (ContentsManager.save or context.save()) and surfaces the saved path in the returned result.",
      "max_score": 15
    }
  ]
}