Snowflake is suky's primary data connection. A read-only warehouse link lets suky discover events, compute quality metrics, and validate contracts — all executed in your environment. Raw event data never leaves your infrastructure.
Setup
- Go to Integrations → Snowflake.
- Enter your account identifier, warehouse, database, and schema.
- Authenticate with key-pair authentication — upload the RSA private key for a read-only service account. Credentials are stored per-workspace and used only by the Snowflake edge functions.
- suky runs a discovery query to find candidate event tables and extract column metadata, sample values, and row counts.
- Review discovered events and activate them as monitored contracts.
What suky reads
- Schema metadata — table names, column names, types, and row counts.
- Quality metrics — freshness, volume baselines, null rates, and schema compliance via incremental SQL queries that use high-water marks to avoid reprocessing.
- Violations — per-event diff between observed data and the active contract: missing required properties, type mismatches, unexpected fields.
CDP-aware parsing
Discovery understands the common warehouse schemas produced by Segment, RudderStack, and Snowplow (nested properties / context columns, per-event tables, self-describing events). If your warehouse lands raw events directly, suky also supports custom table layouts.
Least-privilege role
Grant the suky role USAGE on the warehouse/database/schema and SELECT on the event tables you want to govern. No write grants are required. suky never runs INSERT, UPDATE, DELETE, or DDL.