In a notebook, and commands of your own

Without an app

import leeghwater

setup = leeghwater.prepare(secret_scopes=["ingest"], catalog="main", warehouse="<id>")
print("\n".join(setup.lines()))

p = leeghwater.create_pipeline("github", schema="github")
leeghwater.run(p, github_source())

prepare() is everything the app does before a run, and returns what it decided; create_pipeline and run go by the last prepare(), and create_pipeline by a setup= when you pass it one. A later prepare() starts from nothing. What an earlier call set in the environment is put back, and the scopes and vaults it registered with dlt are taken out. A prepare() that is refused leaves no setup: call it again before you make a pipeline. Importing leeghwater changes nothing and does not import dlt.

Asking where it runs

here = leeghwater.where()
if here.on_databricks:
    ...

where() reads DATABRICKS_RUNTIME_VERSION and nothing else. here.on_databricks is true when the variable is set, here.runtime is what it holds, and here.serverless is true when that starts with client..

Commands of your own

@app.command()
def backfill(day: str):
    """Load one day again."""
    ...

ingest backfill 2024-06-01 is prepared as a pipeline is, with the app's own settings, and finds the same secrets.