Beginner Tutorial: AI Evaluation Tool with Gemini & Google Sheets

This workflow serves as an accessible tutorial for n8n beginners, showcasing the Evaluation tool to automatically assess AI-generated answers against predefined correct responses (ground truth) stored in Google Sheets. It integrates Google Gemini as the AI model within an AI Agent node equipped with a Calculator tool for precise computations, enabling reliable handling of math-based queries like 'How much is 8 * 3?'.

The workflow supports two modes: Manual Test Mode, triggered by clicking 'Execute workflow', for quick ad-hoc testing with fixed inputs; and Evaluation Mode, triggered by fetching dataset rows from Google Sheets, where each row provides an input question and expected output. In Evaluation Mode, it processes multiple rows scalably, sending questions to the AI Agent, generating outputs, and comparing them for factual accuracy.

Key benefits include its simplicity for learning AI evaluation, flexibility with input sources, automated scoring and result tracking back to Sheets, and demonstration of tool usage in AI agents. Ideal use cases are structured AI model testing, quality assurance for text-based AI tasks, and batch evaluation of LLM performance in production environments, saving significant manual review time.