Let's say you have two Capos DAGs. How would you create a dependency so that DAG A must finish before DAG B starts?

In Airflow you can enforce that DAG B waits for DAG A to finish by using an ExternalTaskSensor in DAG B. The sensor references the upstream DAG ID and the task ID (or the DAG run) you want to wait for. Example:

from airflow import DAG
from airflow.sensors.external_task_sensor import ExternalTaskSensor
from datetime import datetime

with DAG('dag_b', start_date=datetime(2024,1,1), schedule_interval='@daily') as dag:
    wait_for_a = ExternalTaskSensor(
        task_id='wait_for_dag_a',
        external_dag_id='dag_a',
        external_task_id='final_task_in_a',
        mode='poke',
        timeout=600,
        poke_interval=30
    )
    # downstream tasks

Alternatively, you can use trigger_dag to programmatically start DAG B from DAG A’s final task, or set depends_on_past=True and use a shared XCom flag. The ExternalTaskSensor is the simplest declarative way to enforce the finish‑before‑start rule without modifying the upstream DAG.