dbt-af 0.14.6

Distibuted dbt runs on Apache Airflow

dbt-af: distributed run of dbt models using Airflow

Overview

dbt-af is a tool that allows you to run dbt models in a distributed manner using Airflow. It acts as a wrapper around the Airflow DAG, allowing you to run the models independently while preserving their dependencies.

Why?

1. dbt-af is domain-driven. It is designed to separate models from different domains into different DAGs. This allows you to run models from different domains in parallel.
2. dbt-af is dbt-first solution. It is designed to make analytics' life easier. End-users could even not know that Airflow is used to schedule their models. dbt-model's config is an entry point for all your settings and customizations.
3. dbt-af brings scheduling to dbt. From @monthly to @hourly and even more.
4. dbt-af is an ETL-driven tool. You can separate your models into tiers or ETL stages and build graphs showing the dependencies between models within each tier or stage.
5. dbt-af brings additional features to use different dbt targets simultaneously, different tests scenarios, and maintenance tasks.

Installation

To install dbt-af run pip install dbt-af.

To contribute we recommend to use uv to install package dependencies. Run uv sync --all-packages --all-groups --all-extras to install all dependencies.

dbt-af by Example

All tutorials and examples are located in the examples folder.

To get basic Airflow DAGs for your dbt project, you need to put the following code into your dags folder:

# LABELS: dag, airflow (it's required for airflow dag-processor)
from dbt_af.dags import compile_dbt_af_dags
from dbt_af.conf import Config, DbtDefaultTargetsConfig, DbtProjectConfig

# specify here all settings for your dbt project
config = Config(
    dbt_project=DbtProjectConfig(
        dbt_project_name='my_dbt_project',
        dbt_project_path='/path/to/my_dbt_project',
        dbt_models_path='/path/to/my_dbt_project/models',
        dbt_profiles_path='/path/to/my_dbt_project',
        dbt_target_path='/path/to/my_dbt_project/target',
        dbt_log_path='/path/to/my_dbt_project/logs',
        dbt_schema='my_dbt_schema',
    ),
    dbt_default_targets=DbtDefaultTargetsConfig(default_target='dev'),
    dry_run=False,  # set to True if you want to turn on dry-run mode
)

dags = compile_dbt_af_dags(
    manifest_path='/path/to/my_dbt_project/target/manifest.json',
    config=config,
)
for dag_name, dag in dags.items():
    globals()[dag_name] = dag

In dbt_project.yml you need to set up default targets for all nodes in your project (see example):

sql_cluster: "dev"
daily_sql_cluster: "dev"
py_cluster: "dev"
bf_cluster: "dev"

This will create Airflow DAGs for your dbt project.

Check out the documentation for more details here.

Features

1. dbt-af is essentially designed to work with large projects (1000+ models). When dealing with a significant number of dbt objects across different domains, it becomes crucial to have all DAGs auto-generated. dbt-af takes care of this by generating all the necessary DAGs for your dbt project and structuring them by domains.
2. Each dbt run is separated into a different Airflow task. All tasks receive a date interval from the Airflow DAG context. By using the passed date interval in your dbt models, you ensure the idempotency of your dbt runs.
3. dbt-af lowers the entry threshold for non-infrastructure team members. This means that analytics professionals, data scientists, and data engineers can focus on their dbt models and important business logic rather than spending time on Airflow DAGs.

Requirements

dbt-af is tested with:

Airflow version	Python versions	dbt-core versions
2.6.3	>=3.10,<3.12	>=1.7,<=1.10
2.7.3	>=3.10,<3.12	>=1.7,<=1.10
2.8.4	>=3.10,<3.12	>=1.7,<=1.10
2.9.3	>=3.10,<3.13	>=1.7,<=1.10
2.10.5	>=3.10,<3.13	>=1.7,<=1.10
2.11.0	>=3.10,<3.13	>=1.7,<=1.10

Project Information

• Docs
• PyPI
• Contributing