Supported Data Warehouses
Tuva 1.0 supports Snowflake, Databricks, BigQuery, Amazon Redshift, Microsoft Fabric, and DuckDB. It also adds SQL Server and AWS Athena compatibility paths. The table below distinguishes the established targets from the newer paths and their validation limits.
| Platform | dbt Core adapter | Tuva 1.0 coverage |
|---|---|---|
| Snowflake | dbt-snowflake | Routine pull-request CI and the complete package release matrix. |
| Databricks | dbt-databricks | Complete package release matrix. See Databricks setup. |
| Google BigQuery | dbt-bigquery | Complete package release matrix. |
| Amazon Redshift | dbt-redshift | Complete package release matrix. Asset loading uses IAM_ROLE default; configure the cluster's default IAM role to read Tuva's public S3 assets. |
| Microsoft Fabric | dbt-fabric | Complete package release matrix. Use the Fabric Warehouse target. |
| DuckDB | dbt-duckdb | Local package and runtime compatibility validation, including dbt Core 2 and Fusion. Use one thread for Tuva asset loading. |
| Microsoft SQL Server | dbt-sqlserver | New Core SQL and asset-loading implementation, tested on SQL Server 2022. Optional-package validation is narrower than the release matrix. |
| AWS Athena | dbt-athena | New Core SQL and asset-loading implementation, tested on Athena engine version 3. Optional-package validation is narrower than the release matrix. |
dbt versions and runtimes
Core and all eight standalone 1.0 packages declare >=1.10.5,<3.0.0. The
1.10.5 minimum is required for the nested generic-test argument syntax used by
the packages. The 1.0 compatibility work also validated dbt Core 2 and Fusion
on DuckDB. Those runtime checks used dbt Core 2.0.0-beta.2 and Fusion
2.0.0-preview.218; they do not establish every runtime-and-warehouse
combination. See the runtime compatibility change
and the released package requirements.
The five hosted warehouse release builds used dbt Core 1.11.14. Install the
adapter for your target when using dbt Core. Fusion has its own connector and
authentication setup; check the connection requirements for the runtime and
target you select.
What release validation covers
Routine Core pull requests run Snowflake CI. The release matrix builds Core, the integration project, and all eight standalone packages on Snowflake, Databricks, BigQuery, Fabric, and Redshift. It loads the small synthetic dataset and runs model builds, unit tests, data tests, and the optional Data Quality relations, including failure keys. The 1.0 release matrix passed on all five targets. DuckDB is validated separately; it is not a hosted matrix job.
SQL Server and Athena received targeted execution and portability checks in the compatibility implementation. They are not additional legs of that five-warehouse matrix. The SQL Server Core checks did not cover all large provider seeds, and the optional marts have different levels of validation. For example, the released CMS HCC compatibility notes do not claim complete Athena validation. Check each selected mart's README before adopting either newer target.
SQL Server requirements
Use SQL Server 2022 or later with dbt-sqlserver; the Tuva checks used adapter
1.11.1. Tuva's CSV loader needs the ordinal form of STRING_SPLIT, which
Microsoft introduced for SQL Server 2022. The database must have compatibility
level 130 or higher. The adapter's own older SQL Server minimum is therefore
not sufficient for Tuva's loader. See Microsoft's function requirements
and the adapter setup guide
for the ODBC driver and connection profile.
As documented in Core's SQL Server requirements,
create the database with a case-sensitive collation, such as
SQL_Latin1_General_CP1_CS_AS. Logical Data Quality includes exact lowercase
value checks; a case-insensitive collation can incorrectly accept uppercase
values. Changing the database default later does not change existing columns'
collations.
The native loader reads gzip assets from Tuva's public Azure mirror using
OPENROWSET, DECOMPRESS, and CSV parsing. It creates an external data source
for the public container. Give the dbt identity the required bulk-read and
external-data-source permissions, and allow the SQL Server host to reach Azure
Blob Storage. This loader does not require PolyBase or a separate converted
asset format. The released loader
also includes the concurrent external-data-source creation fix.
Athena requirements
Use an Athena workgroup running engine version 3 and the dbt-athena
adapter. Configure an AWS identity, region, Glue catalog/database, and writable
S3 locations for query results and dbt table data. Keep those output locations
separate from Tuva's public asset bucket. See the Athena connection guide
and table-location configuration.
Tuva's loader reads the published S3 CSV assets through temporary external tables, then creates typed seed tables. The identity needs Athena query and Glue catalog permissions, read access to the public assets, and write access to its own output locations. Athena's nonadjustable 262,144-byte query limit still applies; large custom queries or expanded Data Quality checks must fit within it.
For a reproducible platform issue, open a Core issue or use the owning mart's issue tracker. Include the Tuva tag, dbt runtime, adapter version, warehouse version, and failing command.