Weirdly, if I use examples straight out of the docs at https://docs.dagster.io/_apidocs/libraries/dagster-dbt I get this error

Param "command" is not a str. Got ['build', '--vars', '{"key": "value"}'] which is type <class 'list'>.

dbt_build_invocation = dbt.cli(["build", "--vars", json.dumps(dbt_vars)], context=context)
    yield from dbt_build_invocation.stream()

Can you confirm that you're uisng

DbtCliResource

?

Well, it's set up like this

from dagster_dbt import (
    dbt_cli_resource,
    load_assets_from_dbt_project,
    dbt_assets,
    DbtCliResource,
    DagsterDbtTranslator

)
.........

def my_dbt_assets(context: OpExecutionContext, dbt: DbtCliResource):
    dbt_vars = {"key": "value"}

    <http://context.log.info|context.log.info>("Running my_dbt_assets")
    # dbt_build_invocation = dbt.cli(" ".join(["build", "--vars", json.dumps(dbt_vars)]), context=context)
    # yield from dbt_build_invocation.stream()
    run_command = " ".join(["run", "--vars", json.dumps(dbt_vars)])
    <http://context.log.info|context.log.info>(f"Running dbt command NO CONTEXT: {run_command}")
    # yield from dbt.cli(run_command, context=context).stream()

    yield from dbt.cli(run_command).stream()
    yield from dbt.cli("test").stream()

you're using

dbt_cli_resource

, not

DbtCliResource

.

dbt_cli_resource

is deprecated and is the old resource.

acch! You are correct. Thanks.

We can definitely make the experience better here in the future: I've raised https://github.com/dagster-io/dagster/issues/16035 so that we raise an error on this next time

well to be fair it's on me for not checking the resource definition

haha fair, but the framework should have warned you, since you annotated it as

: DbtCliResource

🙂

Well in that case it's on me for not submitting a PR

love to hear that

Would using this cause changes to dbt models not to be reflected if I reload the module? I.e. I update the code (I have it mounted inside a container) so

docker compose exec dagit cat /opt/dagster/dagster_home/modules/batch_puller/batch_puller/models/batch_product_data/stg_batch_data.sql
... this shows up to date sql!

But then in the dagit ui I go to deployment and reload the module, and when I check the definition for stg_batch_data it's still out of date... maybe a result of having previously used

load_assets_from_dbt_project

?

There's a lot of problems with

load_assets_from_dbt_project

such as: • Random file contention • Incurs a ton of latency when compiling the manifest for large projects • Incurs repeated latency since it must be done for each run/code location/etc So we don't really recommend using it. But we do give you an escape hatch if you want to use this behavior for local development: https://dagster.slack.com/archives/C04CW71AGBW/p1690395671541379?thread_ts=1690388053.337689&cid=C04CW71AGBW

thanks again - I am not using load_assets_from_dbt_project but I also am trying to figure out how to get dagster to see that the dbt models have changed (or even been removed entirely)

got it: if you're not using

load_assets_from_dbt_project

, then you should recompile your dbt manifest using your CLI. e.g.

dbt parse

Dagster uses this manifest to understand the state of your dbt project (which is why you pass it into

@dbt_assets

)