sql-correctness

To evaluate Databricks SQL correctness in traces, provide a trace containing execute_sql tool calls or raw SQL snippets to validate syntax, Unity Catalog patterns, and Delta-specific features against a structured rubric.

Unity Catalog requires 3-level namespace usage (catalog.schema.table) for Databricks trace analysis. The evaluation checks that SQL statements enforce this syntax along with modern DDL patterns and safe parameterization.

Yes, this SQL evaluation checks both DDL and DML statements in Databricks environments. It validates modern DDL syntax, safe data access patterns, and Delta-specific features while discouraging anti-patterns and unsafe practices.

You validate safe parameterization and prevent anti-patterns in Databricks SQL by applying the evaluation rubric to traces with execute_sql calls, checking for modern syntax and discouraging unsafe data access practices.

You can use this to check both raw SQL snippets and execute_sql tool calls. The evaluation workflow triggers whenever the trace contains Databricks SQL statements, covering syntax validity and data access patterns.