Run a scientific experiment on your own data. Single-pass SQL vs 2-pass Reflection. An LLM Judge scores both. Measure the difference.
One LLM call generates SQL from your schema and question. Fast, simple, the baseline every team uses today.
Generate SQL, execute it, let an LLM critique the result, rewrite if needed, execute again. The Reflection agentic pattern applied to SQL.
A third LLM scores accuracy and completeness 1–10 for each approach. See which wins and why — in structured JSON you can analyse.
Link your Snowflake or BigQuery database. The tool reads your real schema so the generated SQL targets your actual tables and columns.
Type a business question: "show me the top 10 customers by revenue in the last 30 days". The experiment runs both tracks automatically.
Control SQL, Reflection SQL, execution results, and a scored verdict appear on canvas. Export to BigQuery for further analysis.
Run 10 questions through both tracks, store results in BigQuery, and have data to back up your architecture decisions.
Try free — no signup →