FEATURES

Everything you need for cross-engine data transformation, built on the tools you already know.

CROSS-ENGINE SQL MODELS

Write SQL models that reference sources on ANY database. DVT detects cross-engine references and automatically handles data movement.

  • ▸Join MySQL + Snowflake + PostgreSQL in one SELECT statement
  • ▸No manual ETL pipelines — DVT routes data automatically
  • ▸Automatic execution path selection based on source locations
  • ▸Two dialects coexist: target SQL for pushdown, DuckDB SQL for extraction

THREE EXECUTION PATHS

DVT intelligently selects the optimal execution strategy for each model in your DAG.

  • ▸Adapter Pushdown — all sources on same engine? SQL runs directly on the database. Zero data movement.
  • ▸Extraction via DuckDB — cross-engine? Sling extracts sources → DuckDB joins → Sling loads to target.
  • ▸Cross-Target — model targets a different database or cloud bucket? Sling streams the result.

DBT COMPATIBILITY

100% backward compatible with dbt-core. Your existing dbt project works unchanged — DVT extends it.

  • ▸Same ref(), source(), config() functions
  • ▸Same materializations: table, view, incremental, ephemeral
  • ▸Same testing framework, documentation, packages
  • ▸Same CLI patterns: run, build, test, compile, seed, docs

INCREMENTAL EXTRACTION

Full incremental support for cross-engine models with persistent DuckDB cache.

  • ▸Watermark-based delta extraction — only new/changed rows move
  • ▸Three strategies: append, delete+insert, merge
  • ▸Persistent .dvt/cache.duckdb between runs
  • ▸Fast is_incremental() checks — local, no network

HIGH-PERFORMANCE SEEDS

Sling-based CSV loading — 10-100x faster than dbt's native agate loader.

  • ▸Bulk loading via COPY, bcp, sqlldr
  • ▸Automatic schema inference
  • ▸Handles dirty data gracefully (encoding, mixed types, nulls)
  • ▸Works across all 19 supported engines

SELF-HEALING ENVIRONMENT

dvt sync automatically verifies and installs everything your project needs.

  • ▸Database drivers (psycopg2, snowflake-connector, etc.)
  • ▸DuckDB extensions (postgres_scanner, mysql_scanner)
  • ▸Cloud SDKs (boto3, google-cloud-storage)
  • ▸Sling binary for data movement

DATA CATALOG

dvt docs generate enriches the catalog with column metadata from ALL engines.

  • ▸Queries remote databases for column types and descriptions
  • ▸Engine-colored lineage graph nodes
  • ▸Futuristic dark theme UI
  • ▸Cross-engine source and model documentation

FEDERATION OPTIMIZER

Intelligent query decomposition for cross-engine models via SQLGlot.

  • ▸Predicate pushdown to source engines
  • ▸Column pruning — only extract needed columns
  • ▸LIMIT propagation
  • ▸Source-specific query decomposition