Evaluate your AI agents faster and more effectively

The DigitalOcean Gradient™ AI Platform has introduced updates to its agent evaluations feature, aimed at enhancing the speed and effectiveness of AI agent assessments. The redesigned evaluation experience addresses previous challenges by introducing goal-oriented metric grouping, example datasets, and clear, persistent error messaging, which simplifies the debugging process. Metrics are organized into intuitive groups like Safety & Security and Correctness, with the former preselected for quick startup. Deep integration with observability tools allows developers to trace low scores back to the source for precise debugging and improvements. These evaluations help developers systematically test and optimize AI agents, providing insights into performance and enabling faster, more reliable deployment. The platform offers a step-by-step tutorial for new users to create test cases, select metrics, and interpret results, facilitating the development of safer and more efficient AI systems.