dbt

The dbt adapter needs no extra and does not import dbt. It reads the target/manifest.json that dbt compile, dbt run and dbt build write, and runs all SQL through a DuckDB or Postgres adapter connected to the database dbt built into.

For a dbt project, point the adapter at the manifest that dbt run (or dbt compile) writes, and at the warehouse dbt built into:

import duckdb
from dataveil.adapters.dbt import DbtAdapter
from dataveil.adapters.duckdb import DuckDBAdapter
from dataveil.core.profile import profile_table

adapter = DbtAdapter("target/manifest.json", DuckDBAdapter(duckdb.connect("shop.duckdb")))
adapter.list_tables()  # ['main_analytics.customers', ...]
profile_table(adapter, "main_analytics.customers")

Models are named schema.alias, the relation dbt builds. Ephemeral models are skipped, because they have no relation. Every classifier works, because the warehouse adapter registers the checksum functions.

Cleansing is not durable yet

An operation changes the relation dbt built, outside dbt. The next dbt run rebuilds a table model and undoes it. Views cannot be cleansed at all. Durable cleansing, written into the model, is planned for 0.5.