Automatically provide notebook context in magic commands

[JasonWeill]

Problem

When a user asks a question, only the prompt provided, plus any other data provided by the user, gets sent to the model. For example, the user can select data and send it to the chat UI's language model with their message. In addition, the /ask command augments a question using data from previous /learn commands. Data from the workbook itself is not included.

Proposed Solution

If the user so desires, augment users' prompts with contextual data from the current notebook. For example, if a user asks, "What is in column 2 of the dataframe?", and the preceding cell contains a Pandas dataframe, augment the prompt with information about the dataframe.

Additional context

As with all Jupyter AI features that send data to a large language model, users should explicitly opt in to having additional context from their notebook sent to language models with their prompts.

[antonstor]

++ having jupyternaut have context of current notebook in both chat and %%ai magic would make it 100x more useful.
Would bring a lot of what VS Code copilot excels at to jupyter

[rbiswasfc]

++ Sending the entire (active) notebook context should be the default IMO. Users can use %ai reset at a desired location, if they don't want all the context above the reset cell.