Overview Quickstart Install & Platforms dbt Compatibility The Style Guide dvt_project.yml dbt_project.yml profiles.yml

DATABASE ADAPTERS

DVT supports eighteen engines from one environment. Every adapter below works as a federation source, a federation target, and a seed target — with full dbt compatibility on your default target, while Sling moves the data during federation.

All drivers are installed automatically by dvt sync based on what your profiles.yml actually uses.

Click any adapter for its setup guide, supported features, and federation notes.

TRANSACTIONAL ENGINES & LOCAL FILES

POSTGRESQL

Open-source relational database. The most popular choice for on-premises analytics and operational data stores.

SQL SERVER

Microsoft's enterprise relational database. Dominant in corporate environments, especially with .NET ecosystems.

ORACLE

Enterprise-grade relational database. Industry standard for mission-critical financial, ERP, and healthcare systems.

MYSQL

The world's most widely deployed open-source relational database. The operational store behind countless web applications and OLTP systems.

MYSQL 5.7

Legacy MySQL 5.7 for the installed base that hasn't moved yet. Same connection setup as MySQL 8 — with the 5.7 SQL surface honestly enforced.

MARIADB

Community-developed MySQL fork with its own storage engines and release cadence. Default database on many Linux distributions.

DUCKDB

In-process analytical database. DVT uses DuckDB internally for federation compute and as the extraction cache engine. MotherDuck is served by the same adapter — point path at md:your_database.

SQLITE

Serverless, file-based relational database. Ideal for local development, testing, and embedded analytics.

LOCAL FILESYSTEM

Local directories as sources and targets — CSV, Parquet, JSON, JSONL; no cloud account needed.

CLOUD BUCKETS & SFTP

AMAZON S3

Materialize models and seeds as Parquet/CSV objects; read lake files as sources. MinIO-compatible.

GOOGLE CLOUD STORAGE

GCS buckets as federation sources and targets for your GCP-side lake.

AZURE BLOB STORAGE

Blob containers as sources and targets. Azurite-compatible for local development.

SFTP

Any SSH-reachable server as a bucket-style endpoint — exchange folders, partner drop zones, legacy hosts. Files move through the same Sling lane as the cloud buckets.

CLOUD WAREHOUSES

SNOWFLAKE

Cloud-native data warehouse with automatic scaling. Popular for analytics workloads and as a central data platform.

BIGQUERY

Google's serverless, highly scalable data warehouse. Pay-per-query pricing makes it cost-effective for intermittent analytics.

REDSHIFT

Amazon's cloud data warehouse based on PostgreSQL. Designed for petabyte-scale analytics with familiar SQL syntax.

DATABRICKS

Unified analytics platform combining data engineering, data science, and business analytics on a lakehouse architecture.

MICROSOFT FABRIC

Microsoft's unified analytics platform combining data engineering, data science, and business intelligence in one SaaS offering.

CLICKHOUSE

Column-oriented OLAP database for real-time analytics. Designed for sub-second queries over billions of rows.

STARROCKS

High-performance MPP OLAP database for real-time analytics. Speaks the MySQL wire protocol, so it connects like MySQL — through the frontend (FE) query port.

TRINO

Distributed SQL query engine for federated analytics across heterogeneous data sources. Formerly known as Presto.

EXASOL

In-memory MPP analytics database built for consistently fast BI and reporting workloads. Connects through a single dsn (host:port) endpoint.

AMAZON ATHENA

AWS serverless query service for analyzing data in S3 using standard SQL. Pay-per-query with no infrastructure to manage.

BEYOND ENGINES: PYTHON MODEL SOURCES

Python models run locally and can pull from anything Python reaches — the result lands in your DAG like any other model and materializes to any target above. Like everything in DVT, they're free for individuals. That makes systems without a native adapter first-class sources anyway. Python models docs →

REST APIS

Stripe, HubSpot, internal services — anything with an HTTP endpoint

MONGODB

documents → DataFrames

ELASTICSEARCH

search indexes and aggregations

EXCEL / LOCAL FILES

spreadsheets, exports, anything pandas reads

ML INFERENCE

score, classify, embed — inline in the DAG

Not supported:Vertica, SingleStore, and self-hosted Spark Thrift have no native adapter. MongoDB and Elasticsearch are best reached through Python models (above). Adapters marked SUPPORTED haven't been run in our live verification suite yet; VERIFIED adapters are exercised end-to-end (federation source + target, seeds, debug) on every release.