Prompts and Response Schemas

Important: Changes to prompts and schemas will take effect immediately for new evaluations. Existing evaluations will not be affected.

These are the instructions given to the AI model for evaluating calls.
This JSON structure defines what metrics will be tracked. Numeric fields will be used for scoring and analytics.
JSON Template Guidelines
Basic Rules:
  • All numeric fields (integers and floats) are automatically tracked as metrics
  • Text fields are not included in statistical calculations
  • Adding a new field will make it appear in all ratings UI tables
  • Every field should have a sensible default value
Example Fields:
  • score_total: Overall score (0-10)
  • english_level: Language proficiency (0-10)
  • greeting: Quality of greeting (0-10)
  • problem_solving: Problem-solving ability (0-10)
  • sentiment: Call sentiment (0-10)
  • toxicity: Detected toxicity (0-10)
  • improvement_advice: Text feedback for improvement