// the find
dbt-labs/dbt-core
dbt enables data analysts and engineers to transform their data using the same practices that software engineers use to build applications.
dbt Core v2 is a ground-up rewrite of the popular SQL transformation tool in Rust, replacing the Python implementation. It targets data engineers who need faster parse/compile times and want a single binary without Python dependency management. The main branch is alpha-quality v2; stable v1 lives on the `1.latest` branch.
- Single self-contained binary with no Python runtime dependency is a significant operational improvement over v1, which notoriously suffered from pip dependency hell and slow startup times.
- Parquet artifacts replacing JSON manifests is a genuinely good call — querying a large manifest.json with jq is miserable; being able to SQL-query project metadata is immediately useful.
- The codebase shows good modular decomposition into per-adapter crates (bigquery, snowflake, databricks, redshift, etc.) with per-relation-type config structs, making the adapter boundary explicit rather than hidden in Python mixins.
- Snapshot-based fixture testing for the xdbc catalog connection logic provides a concrete regression safety net for the most error-prone config surface area.
- Still alpha with an explicit 'APIs and on-disk formats may change' warning — anyone adopting this in production pipelines today is signing up for breaking changes with no stable migration path yet.
- The split between 'dbt Core' (Apache 2.0, open) and 'Fusion engine' (free but not fully open) is confusing and feels like a bait-and-switch; the README steers you toward Fusion for anything serious, and it's unclear which features will stay in Core vs. become Fusion-only.
- The existing Python adapter ecosystem (hundreds of community adapters) doesn't carry over — the Rust rewrite requires adapters to be rewritten, so most of the long tail of community adapters simply won't work with v2 initially.
- No migration guide or compatibility matrix between v1 and v2 project configs is visible in the repo yet, which makes evaluating the actual upgrade cost impossible for existing users.