This is probably more applicable for the dbt Slack community but I'm going to try here anyways. I'm confused about the models not needing to be declared in the

.yml

with schemas etc. Seems that if I just drop my

.sql

files in the

models

folder the views are created. So far so good. I want to integrate it with Metabase, and it seems

dbt-metabase

needs the models to be properly declared in the YaML files... But I wonder; why introduce this source of inconsistency? My `SELECT`s may very well not have the same schema as whatever I declare in

models.yml

. To me the query should be enough declaration in itself, shouldn't it?

It sounds like you are hardcoding the schema in the model SQL files, rather than using the

ref

macro.

why introduce this source of inconsistency? My `SELECT`s may very well not have the same schema as whatever I declare in

models.yml

.

If you use the

ref

macro in your SQL models, there’s no inconsistency. You can define the schema just once (e.g., in your

dbt_project.yml

file). This is a far better approach because it supports workflows like a developer using a differenct schema for local testing than for production - easy to configure that when it’s defined in just one place rather than being hardcoded in many.

The set of selected columns in my

.sql

file SHOULD be the same as in

models.yml

but as far as I can tell that's not enforced