How can I include custom asset observations i.e. row count into DBT-backed assets? Are there any configuration options I am overlooking?

You should be able to use the

runtime_metadata_fn

parameter in

load_assets_from_dbt_*

. I’m not sure if that generates an AssetObservation, but it would add to the metadata returned.

Some documentation example on how to use

runtime_metadata_fn

for dbt projects would go a long way.

https://dagster.slack.com/archives/C04CW71AGBW/p1670540244487949 this thread might help

One option is in your IO manager's

handle_output

method, you could do

context.log_event(AssetObservation(...))

Is there any exaple out there? It would be cool of the default docs of dagster could showcase such an example.

hmm... our api docs do have a simple example, but it currently isn't displaying correctly: https://docs.dagster.io/_apidocs/io-managers#dagster.OutputContext.log_event Here it is copy pasted:

from dagster import IOManager, AssetMaterialization

            class MyIOManager(IOManager):
                def handle_output(self, context, obj):
                    context.log_event(AssetMaterialization("foo"))

I agree that this pattern would be helpful to document in the docs though. We also have a similar utility that attaches observations via the

load_input

method of an IO manager. I'd recommend filing an issue so that we can track adding this to our documentation

https://github.com/dagster-io/dagster/issues/12612 here you go