The starting time, as milliseconds since epoch (inclusive).
to
integer
required
The ending time, as milliseconds since epoch (exclusive). May not be more than 100 days after from.
env
string
required
An environment key. Only metrics from this environment will be included.
contextKind
string
A context kind. Only metrics from events that include a context of this kind are included. Required if contextKey is provided.
contextKey
string
A context key. Only metrics from events whose context of the contextKind kind has this key are included. Requires contextKind.
Response
Metrics computed
inputTokens
integer
Number of input tokens
outputTokens
integer
Number of output tokens
totalTokens
integer
Total number of tokens
generationCount
integer
deprecated
Number of attempted generations
generationSuccessCount
integer
Number of successful generations
generationErrorCount
integer
Number of generations with errors
thumbsUp
integer
Number of positive feedback ratings
thumbsDown
integer
Number of negative feedback ratings
durationMs
integer
Total generation duration, in milliseconds
timeToFirstTokenMs
integer
Time to the first token, in milliseconds
satisfactionRating
number<float>
A value between 0 and 1 representing satisfaction rating
Required range: 0 <= x <= 1
inputCost
number<double>
Cost of input tokens in USD
outputCost
number<double>
Cost of output tokens in USD
judgeAccuracy
number<float>
Average accuracy judge score (0.0-1.0)
Required range: 0 <= x <= 1
judgeRelevance
number<float>
Average relevance judge score (0.0-1.0)
Required range: 0 <= x <= 1
judgeToxicity
number<float>
Average toxicity judge score (0.0-1.0)
Required range: 0 <= x <= 1
judgeScores
object[]
Average score and sample count for every judge metric recorded in the time range, keyed by the judge's evaluation metric key. Includes custom judges and the built-in accuracy, relevance, and toxicity judges. Omitted when no judge scores were recorded.