Airflow Hooks¶

Hooks are the right place for source loading, connection handling, and reusable Airflow-side data access patterns. They are especially important when the execution path is not a pure local-file zero-config workflow.

DataQualityHook¶

DataQualityHook is the general-purpose Airflow hook for first-party quality execution.

from truthound_airflow import DataQualityHook

hook = DataQualityHook(connection_id="warehouse")

Use it when:

the DAG already models source access through Airflow connections
you want to reuse connection-aware logic across operators
you need a hook boundary that can normalize sources before execution

TruthoundHook¶

TruthoundHook is the explicit Truthound-first hook path for teams that want the engine choice to be visible in code.

DataLoader And DataWriter¶

DataLoader and DataWriter are useful when you want source access or result persistence patterns to stay hook-level and reusable instead of being embedded into individual tasks.

Common Parameters¶

Parameter	Type	Description
`connection_id`	str	Airflow connection ID for SQL or credential-backed sources
`engine_name`	str	explicit engine override when not using the default Truthound path
`timeout`	int	hook-level timeout for source access or execution
`retries`	int	hook-level retry policy

ConnectionConfig¶

Use ConnectionConfig when you want to centralize connection metadata instead of scattering connection details across DAG code.

How Hooks Fit With Operators¶

from truthound_airflow import DataQualityCheckOperator, DataQualityHook

hook = DataQualityHook(connection_id="warehouse")

check = DataQualityCheckOperator(
    task_id="quality_check",
    sql="select * from analytics.users",
    hook=hook,
    rules=[{"column": "id", "type": "not_null"}],
)

Use hooks when the source contract belongs to Airflow. Use the operator alone when a simple local-path or already-normalized source is enough.

Connection Setup Guidance¶

create or reuse a standard Airflow connection
keep credentials in Airflow-managed storage
let the hook mediate data access rather than embedding credentials into DAG code