yeah @Hui Zheng, the real interesting thing if you go down this path is that you realize how much dbt handles for you in the development process for you, since it just looks at the SQL files. But you can absolutely create a similar system, where your Dagster job could look at a set of PRQL files, or whatever, and compiles them at runtime, and you rarely have to change the actual Dagster Python, and instead just work with the YAML.
Honestly it blows my mind all the cool pseudo-products you could build with Dagster-as-a-platform. I am personally waiting for someone to port all the Datahub scrapers into Dagster.