dataset-curation

What problem does it solve?

Create and manage evaluation datasets for Cortex Agents to enable robust testing and benchmarking of agent behavior.

Core Features & Use Cases

• Workflow design: Define dataset schemas, source questions, and expected answers to ensure consistent evaluation formats.
• Format standardization: Produce datasets in the Snowflake Agent Evaluations format, including ground_truth structures and tool invocations.
• Versioned delivery: Maintain dataset versions (v1, v2, ...) and re-register with clear change logs for reproducibility.

Quick Start

Create a new evaluation dataset by outlining the source questions, expected answers, and the steps to register it for evaluation.