How to run evals for Conversational Analytics agents

Prism is an open-source evaluation framework for testing Conversational Analytics agents.

• •Standardized test suites with assertions to measure accuracy and catch performance regressions
• •Text/query checks, data validation, latency limits, and AI judge for evaluation
• •Trace view shows execution paths, SQL, and reasoning for debugging
• •Comparison dashboard tracks performance changes across evaluation runs
• •Available for BigQuery, Conversational Analytics API, and Looker APIs

This summary was automatically generated by AI based on the original article and may not be fully accurate.