PROJECT RESOURCES
A dbt project is more than models — and all of it works in DVT. Each resource type has its own page, at full depth, with examples.
The short story: tests, snapshots, analyses and macros pass through to dbt unchanged, and seeds get a serious upgrade — Sling bulk loading and four file formats (csv, parquet, json, jsonl) instead of one. The long story lives on the pages below.
Version-controlled data files loaded as tables — same semantics as dbt seed, but with COPY-class bulk loading through Sling, four file formats, and any target including cloud buckets.
Read the docs →
Type-2 slowly changing dimensions with dbt snapshot semantics — timestamp and check strategies, full history columns, and snapshots that survive engine switches.
Read the docs →
Assert what must be true about your data: generic yml tests, singular SQL tests, custom generics, severity and where configs — and honest handling of tests on federated sources.
Read the docs →
Version-controlled SQL that compiles but never runs — full Jinja, DAG-aware, kept in analyses/. A DVT target flip transpiles analyses too.
Read the docs →
dbt's own Jinja layer, unchanged: ref, source, var, env_var, custom macros and Hub packages — plus the one DVT rule about which dialect a macro must speak.
Read the docs →
Looking for models themselves? Those are on .sql models and Python models. The files that configure the project live under project files.