I want an asset to depend on and load in a dbt model after its materialized

[dagsir[bot]]

Hi all - I’m trying to use a dbt-generated asset to create a Python one. I’m following <https://github.com/dagster-io/dagster/blob/8978f184b8b041e7f37fbdd886a1800d27fc87a2/examples/assets_dbt_python/assets_dbt_python/assets/forecasting/__init__.py#L20|this example>, and here is a snippet:

    ins={"my_dbt_model": AssetIn(key=get_asset_key_for_model([my_dbt_assets], model_name="my_dbt_model"))},
    compute_kind="python"
)
def my_python_asset(my_dbt_model: pd.DataFrame):
    df = my_dbt_model
    return df```
However, when I materialize the python asset in local Dagster (via dagster dev), that leads to the error: `FileNotFoundError: [Errno 2] No such file or directory: '/GitHub/my-repo/tmpnlfo8dgi/storage/my_dbt_model'` . What am I missing for referencing the dbt asset?

---

*The question was originally asked in [Dagster Slack](https://dagster.slack.com/archives/C04CW71AGBW/p1699890703735749?thread_ts=1699890703.735749&cid=C04CW71AGBW).*

[rexledesma]

You are missing an IO manager for your dbt assets:

dagster/examples/assets_dbt_python/assets_dbt_python/__init__.py

Lines 36 to 37 in 8978f18

	# this io_manager allows us to load dbt models as pandas dataframes
	"io_manager": DuckDBPandasIOManager(database=os.path.join(DBT_PROJECT_DIR, "example.duckdb")),

IO managers allow you to load assets as in-memory representations for downstream computations. See the docs for more details.

[skwon615]

Ah gotcha, thank you for that! Is the Duck DB Pandas IO manager the default one that you'd recommend for this use case?

[rexledesma]

In general, you should choose (or create) an IO manager that takes your storage backend and maps it to your desired Python object.

If your dbt assets are stored in DuckDB and you want to manipulate your dbt models as pandas dataframes, then you should use the DuckDBPandasIOManager.

[skwon615]

Gotcha, that makes sense - thanks! And one more question - in the GitHub example project, where does it specify that the dbt assets use the DuckDB io_manager?

[rexledesma]

By default, if the io_manager is not specified, assets will use the io_manager defined by the resource key "io_manager".

So in the snippet that I provided above, we define the default IO manager for all assets. So the dbt assets definition uses the DuckDBPandasIOManager.

[skwon615]

Ah cool, so that resource key is the default. Awesome, will give this a shot - thank you!