Would it be possible to define some N cells that would be automatically inserted into a newly created notebook? For example, one currently needs to know how to import pandas, pyodide, and reference the local database. Would be great if that was configured by default. Or to have markdown introduction that would mention any instructions/gotcha/etc that apply to the environment.

Very intrigued by this extension. I am wondering if it could be hacked so that it could be populated with notebooks stored in the database (with or without saved cell output). Almost like a self-hosted Colab notebook with analysis + data. My potential use cases:

• I do an analysis with my ridiculously heavy (ie out of scope for Pyodide) Python environment and store the input + output cells into a notebook for read-only consumption.
• Have a pre-populated notebook with input + optional output cells that allow the reader to execute/modify queries to poke at some data that might be trickier to handle in SQL. Or to provide markdown context surrounding what a query means.

Thinking out loud, it feels like there is a potential tie in with the Executable Book format to distribute an entire class in a file.

-Gushing fan boy who loves this project.

Currently if you are inside JupyterLite and you want to load data from the host Datasette instance into a DataFrame or similar you need to do this:

from js import fetch
from io import StringIO
import pandas

df = pandas.read_csv(
  StringIO(await (await fetch("https://latest-with-plugins.datasette.io/github/stars.csv")).text())
)

It would be great if there was a less boilerplate heavy way of doing this.