Class TraceAggregateManager

Manages trace-level aggregate state for error propagation and span tracking.

Tracking strategy to prevent memory leaks:

All spans are tracked at onStart (we need the count for cleanup)
At onEnd, we check if the span is an AI SDK span and mark the trace accordingly
Error state is only aggregated for AI SDK traces
Non-AI SDK traces are cleaned up immediately when they complete

Index

Constructors

Methods

clear onEnd onStart

Constructors

constructor

new TraceAggregateManager(): TraceAggregateManager
Returns TraceAggregateManager

Methods

clear

clear(): void
Clear all tracked aggregates. Called during shutdown/forceFlush to prevent leaks.

Returns void
- Defined in packages/openinference-vercel/src/TraceAggregateManager.ts:179

onEnd

onEnd(span: ReadableSpan): void
Process a span ending: update error state, rename root span, set root status.
Parameters
- span: ReadableSpan
Returns void
- Defined in packages/openinference-vercel/src/TraceAggregateManager.ts:127

onStart

onStart(span: Span): void
Track a span starting. All spans are tracked to maintain accurate counts.
Parameters
- span: Span
Returns void
- Defined in packages/openinference-vercel/src/TraceAggregateManager.ts:110