Skip to main content
This topic describes metrics that LaunchDarkly autogenerates from AgentControl config events. An AgentControl config is a resource that you create in LaunchDarkly and then use to customize, test, and roll out new large language models (LLMs) within your generative AI applications. As soon as you start using AgentControl in your application, your AI SDKs begin sending events to LaunchDarkly and you can track how your AI model generation is performing. AI SDK events are prefixed with $ld:ai and LaunchDarkly automatically generates metrics from these events. Some events generate multiple metrics that measure different aspects of the same event. For example, the $ld:ai:feedback:user:positive event generates a metric that measures the average number of positive feedback events per user, and a metric that measures the percentage of users that generated positive feedback. The following expandable sections explain the metrics that LaunchDarkly autogenerates from AI SDK events:
Metric kind: Custom conversion countSuggested analysis unit: UserDefinition:
  • Measurement method: Count
  • Unit aggregation method: Sum
  • Analysis method: Average
  • Success criterion: higher is better
  • Units without events: Include units and set the value to 0
Description: Average number of positive feedback events per contextExample usage: Running an experiment to find out which variation causes more users to click “thumbs up”
Metric kind: Custom conversion binarySuggested analysis unit: RequestDefinition:
  • Measurement method: Occurrence
  • Unit aggregation method: Average
  • Analysis method: Average
  • Success criterion: higher is better
  • Units without events: Include units and set the value to 0
Description: Percentage of contexts that generated positive AI feedbackExample usage: Running a guarded rollout to make sure there is a positive feedback ratio throughout the rollout
Metric kind: Custom conversion countSuggested analysis unit: UserDefinition:
  • Measurement method: Count
  • Unit aggregation method: Sum
  • Analysis method: Average
  • Success criterion: lower is better
  • Units without events: Include units and set the value to 0
Description: Average number of negative feedback events per contextExample usage: Running an experiment to find out which variation causes more users to click “thumbs down”
Metric kind: Custom conversion binarySuggested analysis unit: UserDefinition:
  • Measurement method: Occurrence
  • Unit aggregation method: Average
  • Analysis method: Average
  • Success criterion: lower is better
  • Units without events: Include units and set the value to 0
Description: Percentage of contexts that generated negative AI feedbackExample usage: Running an experiment to find out which variation causes more users to click “thumbs down”
Metric kind: Custom numericSuggested analysis unit: RequestDefinition:
  • Measurement method: Value/size
  • Unit aggregation method: Average
  • Analysis method: Average
  • Success criterion: higher is better
  • Units without events: Exclude units that generate no events
Description: For example, for a chatbot, this might indicate user engagementExample usage: Running an experiment to find out which variation generates more input tokens, indicated better engagement
Metric kind: Custom numericSuggested analysis unit: RequestDefinition:
  • Measurement method: Value/size
  • Unit aggregation method: Average
  • Analysis method: Average
  • Success criterion: lower is better
  • Units without events: Exclude units that generate no events
Description: Indicator of cost, when charged by token usageExample usage: Running an experiment to find out which variation results in fewer output tokens, reducing cost
Metric kind: Custom numericSuggested analysis unit: RequestDefinition:
  • Measurement method: Value/size
  • Unit aggregation method: Average
  • Analysis method: Average
  • Success criterion: lower is better
  • Units without events: Exclude units that generate no events
Description: Indicator of cost, when charged by token usageExample usage: Running an experiment to find out which variation results in fewer total tokens, reducing cost
Metric kind: Custom numericSuggested analysis unit: RequestDefinition:
  • Measurement method: Value/size
  • Unit aggregation method: Average
  • Analysis method: Average
  • Success criterion: lower is better
  • Units without events: Exclude units that generate no events
Description: Time required for LLM to finish a completionExample usage: Running an experiment to find out which variation results in faster user completion, improving engagement
Metric kind: Custom conversion countSuggested analysis unit: UserDefinition:
  • Measurement method: Count
  • Unit aggregation method: Sum
  • Analysis method: Average
  • Success criterion: higher is better
  • Units without events: Include units and set the value to 0
Description: Counter for successful LLM completion requestsExample usage: Running an experiment to find out which variation results in more user completion requests (“chattiness”), improving engagement
Metric kind: Custom conversion countSuggested analysis unit: UserDefinition:
  • Measurement method: Count
  • Unit aggregation method: Sum
  • Analysis method: Average
  • Success criterion: lower is better
  • Units without events: Include units and set the value to 0
Description: Counter for erroneous LLM completion requestsExample usage: Running a guarded rollout to make sure the change doesn’t result in a higher number of errors
Metric kind: Custom NumericSuggested analysis unit: UserDefinition:
  • Measurement method: Value/size
  • Unit aggregation method: Average
  • Analysis method: Average
  • Success criterion: lower is better
  • Units without events: Exclude units that generate no events
Description: Time required for LLM to generate first tokenExample usage: Running a guarded rollout to make sure the change doesn’t result in longer token generation times
As an example, the autogenerated metric in the first expandable section above tracks the average number of positive feedback ratings per user. Here is what the metric setup looks like in the LaunchDarkly user interface:
An autogenerated metric.

An autogenerated metric.