Snowflake
Zingle connects to Snowflake to run SQL, introspect your schemas, preview data, validate models, and estimate cost. Snowflake is the supported data warehouse today.
Requirements
- A Snowflake account.
- A role provisioned for Zingle, scoped to least privilege (see below).
- Per-user credentials - a password or a key-pair private key.
Access needed
Zingle runs SQL as the connecting user, using a role you provide. Scope the role to least privilege:
- Read access to the databases/schemas Zingle should introspect and query.
- No privileges on your production tables to modify data.
- If you use Refactor validation, a scratch space where Zingle may create and drop disposable validation databases and temporary tables (kept separate from production).
See the Production safety model for the full shared-responsibility posture.
Connection model
- A connection holds the Snowflake account and connection-level configuration (created once by an admin, shared across the org).
- Targets are compute contexts within a connection - a database, schema, and warehouse to run against. A connection can have multiple targets.
- Credentials are per user and encrypted at rest.
What it enables
- Run SQL for validation and for previewing model output.
- Introspect schemas - list databases, schemas, tables, and columns.
- Validate models -
dbt compile/dbt testand result diffs. - Estimate cost and compute model/source statistics.
What it can't do
- Model materialization (
dbt run) is not part of the active toolset. - Query results are row-capped (currently 1,000 max) - Zingle does not bulk- export tables. See Data handling.
- Zingle can only do in the warehouse what the role you grant permits.
Common setup issues
- Connection test fails. Re-check the account identifier, credentials, and that the role/warehouse is usable by the connecting user.
- Queries can't see expected objects. The role likely lacks grants on those databases/schemas.
- Refactor validation fails to create validation objects. The role lacks the privilege to create the disposable validation database/tables - grant it in a scratch space separate from production.