dbt + Dagster#

Using Dagster Cloud? Automatically load your dbt models as Dagster assets by importing an existing dbt project in Dagster Cloud

Dagster orchestrates dbt alongside other technologies, so you can schedule dbt with Spark, Python, etc. in a single data pipeline.

Dagster's Software-defined Asset approach allows Dagster to understand dbt at the level of individual dbt models. This means that you can:

  • Use Dagster's UI or APIs to run subsets of your dbt models, seeds, and snapshots.
  • Track failures, logs, and run history for individual dbt models, seeds, and snapshots.
  • Define dependencies between individual dbt models and other data assets. For example, put dbt models after the Fivetran-ingested table that they read from, or put a machine learning after the dbt models that it's trained from.

An asset graph like this:

Dagster graph with dbt, Fivetran, and TensorFlow

Can be produced from code like this:

from pathlib import Path

from dagster_dbt import DbtCliResource, dbt_assets, get_asset_key_for_model
from dagster_fivetran import build_fivetran_assets

from dagster import AssetExecutionContext, asset

fivetran_assets = build_fivetran_assets(
    connector_id="postgres",
    destination_tables=["users", "orders"],
)


@dbt_assets(manifest=Path("manifest.json"))
def dbt_project_assets(context: AssetExecutionContext, dbt: DbtCliResource):
    yield from dbt.cli(["build"], context=context).stream()


@asset(
    compute_kind="tensorflow",
    deps=[get_asset_key_for_model([dbt_project_assets], "daily_order_summary")],
)
def predicted_orders():
    ...

Getting started#

There are a few ways to get started with Dagster and dbt:


Understanding how dbt models relate to Dagster Software-defined assets#

Dagster’s software-defined assets (SDAs) bear several similarities to dbt models. A software-defined asset contains an asset key, a set of upstream asset keys, and an operation that is responsible for computing the asset from its upstream dependencies. Models defined in a dbt project can be interpreted as Dagster SDAs:

  • The asset key for a dbt model is (by default) the name of the model.
  • The upstream dependencies of a dbt model are defined with ref or source calls within the model's definition.
  • The computation required to compute the asset from its upstream dependencies is the SQL within the model's definition.

These similarities make it natural to interact with dbt models as SDAs. Let’s take a look at a dbt model and an SDA, in code:

Comparison of a dbt model and Dagster asset in code

Here's what's happening in this example:

  • The first code block is a dbt model
    • As dbt models are named using file names, this model is named orders
    • The data for this model comes from a dependency named raw_orders
  • The second code block is a Dagster asset
    • The asset key corresponds to the name of the dbt model, orders
    • raw_orders is provided as an argument to the asset, defining it as a dependency

To learn how to load dbt models into Dagster as assets, check out the tutorial or the quick version in the dagster-dbt reference.