The adapter moved into the tool

dbt v2, which runs on the Rust-based Fusion engine, is the first dbt release to ship with a DuckDB adapter included rather than installed separately. For a large part of the dbt user base this closes a gap that had been open since Fusion was announced.

The history explains why it matters. The community adapter dbt-duckdb received its first pull request in August 2021 and has since accumulated 1,400 stars. Its appeal was that it removed the warehouse from the equation entirely: install one Python package, point it at a local file, and have a working project without signing up for or paying for a server.

When dbt Labs announced the Fusion engine in May 2025, DuckDB was not supported out of the box. Two days later a user opened an issue asking for an adapter, and it went on to collect 146 heart reactions and 21 thumbs-up — an unusually loud signal for a GitHub issue, and a reasonable proxy for how many projects depended on the setup.

In June 2026 dbt Labs released the first alpha of dbt Core 2.0 built on the same foundations, and open-sourced a large part of the Fusion code under Apache 2.0 into the dbt-core repository. The separate dbt-fusion repository was archived.

The v2 release also covers DuckLake and Iceberg catalogs, and the ability to query dbt's own Parquet metadata with DuckDB — which means the transformation tool's internal state becomes queryable with the same engine that runs the transformations.