dbt 37,942 's migration guide sets three conditions: the project must run on the latest v1 release without deprecation warnings, use a supported adapter, and keep its Semantic Layer YAML on the latest spec. Packages must declare a require-dbt-version range that includes 2.0.0, and anything that parses v1's log lines, or SQLFluff, needs replacing (dbt lint is the substitute). BookNest's project passed the first test in Transforming Data with dbt, so it ran unchanged in a separate virtual environment:
pip install dbt # a 125 MB wheel fetched from dbt's CDN
dbt build --target duck
dbt debug --target pg Succeeded model main.fct_sales (incremental) [9 of 21 in 1.21s]
Passed test assert_fct_sales_reconciles [20 of 21 in 0.22s]
==================== Execution Summary =====================
Finished 'build' successfully for target 'duck' [6.1s]
Summary: 21 total | 21 success
...
[error] [InvalidConfig (dbt1005)]: Configuration Error: The 'postgres' adapter is not yet
supported by dbt.
Supported adapters: snowflake, bigquery, databricks, redshift, duckdb, salesforce, clickhouse.
...The DuckDB 61,228 target built all 21 nodes, but the PostgreSQL 1,289 target is not supported yet (an environment variable enables experimental adapters). The first connection also downloaded an ADBC driver, libadbc_driver_duckdb-1.5.4.so (70 MB), into ~/.cache/com.getdbt/. Timed cold, with no target/ folder, dbt build --target duck took 21-23 s on dbt Core 1.12.5 and 6.6-6.8 s on dbt 2.0.6, about 3x faster on this shared 4-CPU host; most of v1's time is Python parsing the project. Plan a migration per adapter: projects on Snowflake, BigQuery 1 , Databricks 2,717 , Redshift 24 or DuckDB can try v2 now, the rest stay on dbt Core.