dbt-core

How dbt models work

Analysts using dbt transform data by writing select statements, and dbt turns those statements into tables and views inside a data warehouse. These select statements, called models, form a dbt project. Models frequently build on top of one another, and dbt makes it easy to manage relationships between models and to visualize those relationships. Quality is assured through testing. The README notes the main branch now hosts dbt Core v2.0 in beta, a ground up rewrite in Rust that forms the foundation of the Fusion engine, while the Python implementation lives on the 1.latest branch.

What changed in v2.0

dbt Core v2.0 is described as engineered for performance at scale, with parse and compile times dramatically improved on the largest projects. It enforces a tightly defined language specification at parse time for stricter correctness, and it produces Parquet artifacts that can be queried and joined, alongside the JSON artifacts kept for backwards compatibility. Installation is easier because v2.0 ships as a single self contained binary with no Python runtime or dependency management. The local documentation experience, dbt docs, is rebuilt on top of the new artifacts so it scales to large projects.

Distributions and platforms

dbt Core is the baseline distribution and Fusion extends it with additional SQL comprehension, and both are free to install and can run locally. The project states they share one framework and one language specification, so business logic is portable in both directions. Supported operating systems for v2.0 and its drivers include macOS, Linux, and Windows on x86-64, with ARM support on macOS and Linux and Windows ARM still marked as not yet supported.

Editorial conclusion

dbt Core is licensed under the Apache License 2.0, and the v2.0 line is distributed as a single self contained binary that runs without requiring a Python runtime.