dbt v2 Goes GA: Quicker Parsing, Unified Adapter Layer, and Docs Without the Massive manifest.json

dbt Labs has released dbt v2 to general availability on most adapters, closing a gap that has split its tooling across two codebases for the past year. Product manager Hope Watson compared the situation to 1800s railways built on different track gauges: nothing was broken, but every transfer cost time and money. The fix is architectural, with dbt open-sourcing the overlapping pieces of v1 and v2 so the ecosystem converges on one engine and one binary over time, though the migration will take years.

To make upgrading practical now, dbt ships a v2 parser inside v1.12 so teams can check compatibility before switching, plus a dbt autofix deprecations command that resolved 22 flagged issues in a live demo in about five minutes.

Adapters, which Watson called dbt's secret foundation, now ship inside the same binary as the engine on a shared layer built on the ADBC driver standard. In v1, each adapter had its own repository and dependency chain, so installing a second adapter could quietly downgrade dbt core or pull in conflicting SDKs. A live demo showed Snowflake and Databricks adapters installing and running together without version conflicts.

Parsing gains were shown on a synthetic 1,800-model, 72-YAML-file project: v1 parsed in roughly 10 to 13 seconds, v2 in about 3. Catching a single SQL typo took about 1.5 seconds on the OSS distribution versus 0.06 seconds on dbt's full distribution. The new linter was described as 50 to 100 times faster than other linters.

Docs v2 replaces the single manifest.json, nearly half a million lines of JSON on that project, with small, queryable Parquet files. Watson used Claude to run ad hoc queries against them, including column-level lineage and a blast radius query showing every downstream model affected by renaming an upstream column. The redesigned docs UI also added DAG grouping views.

Source: dbt Labs Blog