Hi I have a question regarding graph-backed SDAs. Would it be possible to generate

AssetMaterialization

events within the graph only? There is a redundant built-in materialization that I wish to avoid. Specific use-case is using dagster-dbt and generating the materializations based on the dbt output. Alternatively, is there a way to influence the built-in materializations? TIA

Hey Dimitris - a couple questions for you: • Do you want to set the metadata on the materialization, or the asset key? • Do you want to set values at definition time, or at run time? I.e. is it known what the metadata / asset key should be before you kick off the run

Hi Sandy. • I want to set the metadata (I can set the asset key using keys_by_output_name, right?) • I need to set values at run time (as in metadata from the parsed dbt output) Preferably I’d be able to define where exactly in the graph the materialization takes place, and not have redundant materializations. But if the only option is to use the materialization at the graph’s output I’ll have to work with that.

In that case, this is definitely doable - using an IO manager that just logs output metadata. I'll send a code snippet tomorrow

ah right, that should do it. I’ll try sth out based on the SnowflakeIOManager in the hackernews example, if you had something else in mind please share the snippet 🙂 Thanks anyway!