Implementation:Astronomer Astronomer cosmos DbtDag Init

Overview

Concrete tool for rendering a dbt project as an Airflow DAG provided by the astronomer-cosmos library. DbtDag is a subclass of both Airflow's DAG and Cosmos's DbtToAirflowConverter, combining native DAG behavior with automatic dbt graph loading and task generation.

Source Location

Repository: astronomer-cosmos
DbtDag class: cosmos/airflow/dag.py Lines L14-26
DbtToAirflowConverter (base logic): cosmos/converter.py Lines L250-459

Signature

class DbtDag(DAG, DbtToAirflowConverter):
    def __init__(self, *args: Any, **kwargs: Any) -> None:

Key Keyword Arguments (from DbtToAirflowConverter.init)

Parameter	Type	Default	Description
`project_config`	`ProjectConfig`	required	Configuration for the dbt project (path, manifest path, project name).
`profile_config`	None	`None`	dbt profile and target configuration. Required for most execution modes.
`execution_config`	None	`None`	Execution mode and related settings (local, docker, kubernetes, etc.).
`render_config`	None	`None`	Controls graph rendering: node selection, exclusion, test behavior, load mode.
`operator_args`	None	`None`	Keyword arguments broadcast to every generated operator.
`on_warning_callback`	None	`None`	Callback invoked when dbt emits warnings during test execution.

Standard Airflow DAG Keyword Arguments

All standard Airflow DAG kwargs are also accepted, including:

dag_id (str) -- unique identifier for the DAG.
schedule (str | timedelta | Timetable | None) -- scheduling interval.
start_date (datetime) -- earliest date for DAG runs.
catchup (bool) -- whether to backfill missed runs.
default_args (dict) -- default arguments for all tasks.
tags (list[str]) -- tags for filtering in the Airflow UI.
max_active_runs (int) -- concurrency limit for DAG runs.

Import

from cosmos import DbtDag

I/O Contract

Inputs

Config objects: ProjectConfig, ProfileConfig, ExecutionConfig, RenderConfig define what to render and how.
Airflow DAG kwargs: dag_id, schedule, start_date, etc. define the DAG's scheduling behavior.
operator_args: Optional dict of kwargs broadcast to every generated operator.

Output

A fully rendered Airflow DAG containing:
- One operator per dbt node (model, test, seed, snapshot) that passes the render configuration filters.
- Task dependencies wired to match the dbt project's dependency graph.
- The DAG is immediately registered with Airflow when instantiated at module scope.

Lifecycle

DbtDag.__init__() calls DAG.__init__() to set up the Airflow DAG.
It then calls DbtToAirflowConverter.__init__(), which:
1. Creates a DbtGraph and calls load() to discover dbt nodes.
2. Calls build_airflow_graph() to create operators and wire dependencies.
The result is a ready-to-schedule DAG with all tasks in place.

Usage Example

from datetime import datetime
from pathlib import Path

from cosmos import DbtDag, ProjectConfig, ProfileConfig, ExecutionConfig, RenderConfig
from cosmos.constants import ExecutionMode, LoadMode

jaffle_shop = DbtDag(
    # Airflow DAG parameters
    dag_id="jaffle_shop",
    schedule="@daily",
    start_date=datetime(2024, 1, 1),
    catchup=False,
    tags=["dbt", "jaffle_shop"],

    # Cosmos configuration
    project_config=ProjectConfig(
        dbt_project_path=Path("/usr/local/airflow/dags/dbt/jaffle_shop"),
    ),
    profile_config=ProfileConfig(
        profile_name="jaffle_shop",
        target_name="dev",
        profiles_yml_filepath=Path("/usr/local/airflow/dags/dbt/jaffle_shop/profiles.yml"),
    ),
    execution_config=ExecutionConfig(
        execution_mode=ExecutionMode.LOCAL,
    ),
    render_config=RenderConfig(
        load_method=LoadMode.DBT_MANIFEST,
    ),
    operator_args={
        "install_deps": True,
    },
)

This single module-level instantiation produces a complete Airflow DAG with one task per dbt model/test/seed, scheduled to run daily.

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment