Agent Analytics API
Usage, evaluation, and conversation analytics for an agent.
Get all sections (legacy)
Deprecated legacy endpoint that returns every analytics section in one response. Prefer the per-section endpoints (summary, trends, durations, evaluations, unanswered).
Get node and reasoning execution counts
Returns exact execution counts grouped by workflow node and evaluation criterion for the requested window.
Get durations
Returns `executionDurations` (one point per execution), `nodeDurations` (one point per node) and `errorHotspotExecutions` (recent error executions, not limited to the scatter sample) for the requested window.
Get evaluations
Returns `evaluationDurations`, one point per individual evaluation with its score and duration, grouped by evaluation criteria, for the requested window.
Get insight snapshots per execution
Returns a map keyed by execution id with the latest conversation insights snapshot produced by up to 500 recent executions in the requested window. Pass executionIds (up to 50) to load tooltips only for visible chart points.
Get conversation insight chart data
Returns `null` when the agent has Insights disabled or no data points configured; otherwise an object with `analyzedSnapshotCount` and one aggregate per configured data point. When the agent groups conversations by a context field, charts count closed groups by default; pass `unit=conversation` for per-conversation counts.
Get summary
Returns `summary` (execution KPIs for the window), `statusDistribution` and `executionTypeDistribution`. Filters follow the shared analytics query semantics.
Get trends
Returns `dailyExecutions` (execution counts per day) and `dailyDuration` (duration percentiles per day) for the requested window.
Get unanswered-threads
Pass `forceRefresh=true` to bypass the cached getters and recompute the section.
Poll document insights
Returns the latest evaluation results with completion percentage and pagination metadata. `sort` and `dir` follow the insights table sort params; `q` filters rows.