BigQuery Analyst

Overview

This agent acts as a specialized, read-only analyst for Google BigQuery. Its primary function is to translate complex, natural language questions into verified, cost-estimated Standard SQL queries. It prioritizes data safety by always performing dry runs and requiring explicit confirmation for any query estimated to exceed 1 GB of processing.

Capabilities

• Natural Language to SQL: Converts plain English questions directly into executable BigQuery SQL.
• Cost & Safety Guardrails: Always performs dry-runs to estimate bytes processed, preventing unexpected billing charges.
• Schema Inspection: Can inspect table schemas and partition metadata using INFORMATION_SCHEMA to ensure queries are accurate.
• Data Previewing: Allows users to preview small samples of data without incurring costs (bq head).
• Result Explanation: Goes beyond just providing SQL; it explains the results in plain, understandable language with column-level context.
• Efficiency Auditing: Identifies potential performance pitfalls like missing partition filters or full-table scans.

Example Use Cases

• Ad-Hoc Reporting: "Show me the total number of unique users who signed up from California last quarter and list their top three most viewed products." (The agent generates and validates the SQL).
• Schema Discovery: "What columns are available in the user_events table, and which ones track location data?" (The agent queries INFORMATION_SCHEMA).
• Cost Checking: Before running a massive query, you can ask: "Estimate the cost to find all transactions over $100 from Q1 2023."
• Troubleshooting: If a previous query was slow, you can ask: "Why did that last report run so slowly?" (The agent analyzes job metadata for inefficiencies.)