Templates reference

Variables, macros and filters can be used in templates (see the Jinja Templating section)

The following come for free out of the box with Airflow. Additional custom macros can be added globally through Plugins, or at a DAG level through the DAG.user_defined_macros argument.


The Airflow engine passes a few variables by default that are accessible in all templates




{{ data_interval_start }}


Start of the data interval. Added in version 2.2.

{{ data_interval_end }}


End of the data interval. Added in version 2.2.

{{ logical_date }}


A date-time that logically identifies the current DAG run. This value does not contain any semantics, but is simply a value for identification.
Use data_interval_start and data_interval_end instead if you want a value that has real-world semantics,
such as to get a slice of rows from the database based on timestamps.

{{ ds }}


The DAG run’s logical date as YYYY-MM-DD.
Same as {{ logical_date | ds }}.

{{ ds_nodash }}


Same as {{ logical_date | ds_nodash }}.

{{ exception }}

None | str | Exception KeyboardInterrupt

Error occurred while running task instance.

{{ ts }}


Same as {{ logical_date | ts }}.
Example: 2018-01-01T00:00:00+00:00.

{{ ts_nodash_with_tz }}


Same as {{ logical_date | ts_nodash_with_tz }}.
Example: 20180101T000000+0000.

{{ ts_nodash }}


Same as {{ logical_date | ts_nodash }}.
Example: 20180101T000000.

{{ prev_data_interval_start_success }}

pendulum.DateTime | None

Start of the data interval of the prior successful DagRun.
Added in version 2.2.

{{ prev_data_interval_end_success }}

pendulum.DateTime | None

End of the data interval of the prior successful DagRun.
Added in version 2.2.

{{ prev_start_date_success }}

pendulum.DateTime | None

Start date from prior successful DagRun (if available).

{{ prev_end_date_success }}

pendulum.DateTime | None

End date from prior successful DagRun (if available).

{{ inlets }}


List of inlets declared on the task.

{{ outlets }}


List of outlets declared on the task.

{{ dag }}


The currently running DAG. You can read more about DAGs in DAGs.

{{ task }}


The currently running BaseOperator. You can read more about Tasks in Operators

{{ macros }}

A reference to the macros package. See Macros below.

{{ task_instance }}


The currently running TaskInstance.

{{ ti }}


Same as {{ task_instance }}.

{{ params }}

dict[str, Any]

The user-defined params. This can be overridden by the mapping
passed to trigger_dag -c if dag_run_conf_overrides_params
is enabled in airflow.cfg.

{{ var.value }}

Airflow variables. See Airflow Variables in Templates below.

{{ var.json }}

Airflow variables. See Airflow Variables in Templates below.

{{ conn }}

Airflow connections. See Airflow Connections in Templates below.

{{ task_instance_key_str }}


A unique, human-readable key to the task instance. The format is

{{ conf }}


The full configuration object representing the content of your
airflow.cfg. See airflow.configuration.conf.

{{ run_id }}


The currently running DagRun run ID.

{{ dag_run }}


The currently running DagRun.

{{ test_mode }}


Whether the task instance was run by the airflow test CLI.

{{ map_index_template }}

None | str

Template used to render the expanded task instance of a mapped task. Setting this value will be reflected in the rendered result.

{{ expanded_ti_count }}

int | None

Number of task instances that a mapped task was expanded into. If
the current task is not mapped, this should be None.
Added in version 2.5.

{{ triggering_dataset_events }}

dict[str, list[DatasetEvent]]

If in a Dataset Scheduled DAG, a map of Dataset URI to a list of triggering DatasetEvent
(there may be more than one, if there are multiple Datasets with different frequencies).
Read more here Datasets.
Added in version 2.4.


The DAG run’s logical date, and values derived from it, such as ds and ts, should not be considered unique in a DAG. Use run_id instead.

Accessing Airflow context variables from TaskFlow tasks

While @task decorated tasks don’t support rendering jinja templates passed as arguments, all of the variables listed above can be accessed directly from tasks. The following code block is an example of accessing a task_instance object from its task:

from airflow.models.taskinstance import TaskInstance
from airflow.models.dagrun import DagRun

def print_ti_info(task_instance: TaskInstance | None = None, dag_run: DagRun | None = None):
    print(f"Run ID: {task_instance.run_id}")  # Run ID: scheduled__2023-08-09T00:00:00+00:00
    print(f"Duration: {task_instance.duration}")  # Duration: 0.972019
    print(f"DAG Run queued at: {dag_run.queued_at}")  # 2023-08-10 00:00:01+02:20

Deprecated variables

The following variables are deprecated. They are kept for backward compatibility, but you should convert existing code to use other variables instead.

Deprecated Variable


{{ execution_date }}

the execution date (logical date), same as logical_date

{{ next_execution_date }}

the logical date of the next scheduled run (if applicable); you may be able to use data_interval_end instead

{{ next_ds }}

the next execution date as YYYY-MM-DD if exists, else None

{{ next_ds_nodash }}

the next execution date as YYYYMMDD if exists, else None

{{ prev_execution_date }}

the logical date of the previous scheduled run (if applicable)

{{ prev_ds }}

the previous execution date as YYYY-MM-DD if exists, else None

{{ prev_ds_nodash }}

the previous execution date as YYYYMMDD if exists, else None

{{ yesterday_ds }}

the day before the execution date as YYYY-MM-DD

{{ yesterday_ds_nodash }}

the day before the execution date as YYYYMMDD

{{ tomorrow_ds }}

the day after the execution date as YYYY-MM-DD

{{ tomorrow_ds_nodash }}

the day after the execution date as YYYYMMDD

{{ prev_execution_date_success }}

execution date from prior successful DAG run; you may be able to use prev_data_interval_start_success instead if the timetable/schedule you use for the DAG defines data_interval_start compatible with the legacy execution_date.

Note that you can access the object’s attributes and methods with simple dot notation. Here are some examples of what is possible: {{ task.owner }}, {{ task.task_id }}, {{ ti.hostname }}, … Refer to the models documentation for more information on the objects’ attributes and methods.

Airflow Variables in Templates

The var template variable allows you to access Airflow Variables. You can access them as either plain-text or JSON. If you use JSON, you are also able to walk nested structures, such as dictionaries like: {{ var.json.my_dict_var.key1 }}.

It is also possible to fetch a variable by string if needed (for example your variable key contains dots) with {{ var.value.get('my.var', 'fallback') }} or {{ var.json.get('my.dict.var', {'key1': 'val1'}) }}. Defaults can be supplied in case the variable does not exist.

Airflow Connections in Templates

Similarly, Airflow Connections data can be accessed via the conn template variable. For example, you could use expressions in your templates like {{ conn.my_conn_id.login }}, {{ conn.my_conn_id.password }}, etc.

Just like with var it’s possible to fetch a connection by string (e.g. {{ conn.get('my_conn_id_'+index).host }} ) or provide defaults (e.g {{ conn.get('my_conn_id', {"host": "host1", "login": "user1"}).host }}).

Additionally, the extras field of a connection can be fetched as a Python Dictionary with the extra_dejson field, e.g. conn.my_aws_conn_id.extra_dejson.region_name would fetch region_name out of extras. This way, defaults in extras can be provided as well (e.g. {{ conn.my_aws_conn_id.extra_dejson.get('region_name', 'Europe (Frankfurt)') }}).


Airflow defines some Jinja filters that can be used to format values.

For example, using {{ logical_date | ds }} will output the logical_date in the YYYY-MM-DD format.


Operates on




Format the datetime as YYYY-MM-DD



Format the datetime as YYYYMMDD



Same as .isoformat(), Example: 2018-01-01T00:00:00+00:00



Same as ts filter without -, : or TimeZone info. Example: 20180101T000000



As ts filter without - or :. Example 20180101T000000+0000


Macros are a way to expose objects to your templates and live under the macros namespace in your templates.

A few commonly used libraries and methods are made available.




The standard lib’s datetime.datetime


The standard lib’s datetime.timedelta


A reference to the dateutil package


The standard lib’s time


The standard lib’s uuid


The standard lib’s random.random

Some airflow specific macros are also defined:

airflow.macros.datetime_diff_for_humans(dt, since=None)[source]

Return a human-readable/approximate difference between datetimes.

When only one datetime is provided, the comparison will be based on now.

  • dt (Any) – The datetime to display the diff for

  • since (DateTime | None) – When to display the date from. If None then the diff is between dt and now.

airflow.macros.ds_add(ds, days)[source]

Add or subtract days from a YYYY-MM-DD.

  • ds (str) – anchor date in YYYY-MM-DD format to add to

  • days (int) – number of days to add to the ds, you can use negative values

>>> ds_add("2015-01-01", 5)
>>> ds_add("2015-01-06", -5)
airflow.macros.ds_format(ds, input_format, output_format)[source]

Output datetime string in a given format.

  • ds (str) – input string which contains a date

  • input_format (str) – input string format. E.g. %Y-%m-%d

  • output_format (str) – output string format E.g. %Y-%m-%d

>>> ds_format("2015-01-01", "%Y-%m-%d", "%m-%d-%y")
>>> ds_format("1/5/2015", "%m/%d/%Y", "%Y-%m-%d")
airflow.macros.random() x in the interval [0, 1).

Was this entry helpful?