Hello!

I am working on a project that requires executing Jupyter Notebooks from a Python script. nbclient would be ideal to do this, unfortunately the notebooks frequently request user input with input or getpass. Per the Jupyter Contribution Workflow, I am opening a proposal here. If accepted, I'm happy to implement this myself.

I propose a new callable named on_cell_input (or on_cell_input_request) is created. It is called with kwargs cell, cell_index and cell_messages. When this is set, the client will be updated to keep a log of cell messages and call the on_cell_input callable whenever an std message is requested.

This is the code in question.

It would be updated roughly as:

if self.on_cell_input:
   run_cell = self.kc.execute(...)
   # handle messages
   parent_msg_id = await ensure_async(run_cell)
else:
   # Continue as usual

As this path will only be followed followed when on_cell_input is set, it should have minimal impact on anyone not using it (exactly one if-statement to check whether