Airflow - AssetAlias

In Airflow, AssetAlias is a way to define a logical alias (group name) for one or more assets, so that producers and consumers can be loosely coupled without hard-coding exact asset URIs.

Think of it as a label or tag for assets that can be resolved dynamically at runtime.

---

Why AssetAlias exists

Assets normally require an exact identity match:

Asset("s3://bucket/orders")

That’s very strict. AssetAlias lets you say:

“I don’t care which exact asset(s) — just give me whatever assets are known as orders.”

This is especially useful when:

• Asset URIs change

• Multiple producers exist

• You want reusable / generic consumer DAGs

• You want late binding of assets

---

What AssetAlias does (conceptually)

AssetAlias resolves to the set of assets currently registered under that alias.

It is not a new asset itself.
It does not store data.
It does not create copies.

It’s a dynamic reference.

---

Basic example

Producer side

from airflow.sdk import Asset, AssetAlias, asset

orders_alias = AssetAlias("orders")

@asset(
    uri="s3://bucket/orders/2025/",
    aliases=[orders_alias],
)
def produce_orders():
    pass

This means:

• The asset s3://bucket/orders/2025/

• Is also known by alias "orders"

---

Consumer side

@dag(schedule=AssetAlias("orders"))
def consume_orders():
    @task
    def process():
        print("Orders updated!")

This DAG runs:

• When any asset with alias "orders" updates

No URI coupling required.

---

Multiple assets → one alias

@asset(uri="s3://bucket/orders/us/", aliases=[AssetAlias("orders")])
def orders_us():
    pass

@asset(uri="s3://bucket/orders/eu/", aliases=[AssetAlias("orders")])
def orders_eu():
    pass

Consumer DAG:

@dag(schedule=AssetAlias("orders"))
def reporting():
    ...

Result:

• Reporting DAG runs when either US or EU orders update

---

One asset → multiple aliases

@asset(
    uri="warehouse.orders",
    aliases=[
        AssetAlias("orders"),
        AssetAlias("finance_inputs"),
    ],
)
def orders():
    pass

---

AssetAlias with AND / OR

schedule = AssetAlias("orders") & AssetAlias("customers")

Meaning:

Run when any asset tagged orders
AND
any asset tagged customers
have both been updated.

---

How it appears in the UI

In the Assets view:

• Assets show their aliases

• Aliases show which assets belong to them

• Lineage graph groups assets via aliases

This gives a semantic layer over raw URIs.

---

Important constraints & behavior

1️⃣ Aliases are global

• Same alias name = same logical group

• No namespace by DAG

2️⃣ Aliases resolve at runtime

• Adding a new asset with alias "orders" automatically affects consumers

• No DAG code changes required

3️⃣ Aliases must exist

• If no asset has that alias, the consumer DAG will not run

---

Common use cases

✅ Stable interfaces between teams
✅ Schema / domain-based grouping
✅ Environment abstraction (orders_dev, orders_prod)
✅ Multi-region data pipelines
✅ Gradual migrations

---

When NOT to use AssetAlias

❌ When you need a specific dataset
❌ When asset identity must be precise
❌ For passing data between tasks

---

Mental model

AssetAlias("orders")
       │
       ├── s3://bucket/orders/us/
       ├── s3://bucket/orders/eu/
       └── warehouse.orders

Consumers depend on the concept, not the location.

---

One-sentence summary

AssetAlias lets you depend on a semantic group of assets instead of hard-coded asset URIs, enabling loose coupling and flexible, data-driven DAGs.