How long does it take for traces to appear in Langfuse?
Traces typically appear in Langfuse within seconds of being flushed by the SDK. The Langfuse SDKs send trace data asynchronously in batches, so there is a short gap between an operation running in your application and the corresponding trace being visible in the UI. See queuing and batching for how this works and how to tune it.
Two factors determine the delay you see:
- Flushing: The SDK must flush its event queue before data is sent. Long-running applications flush automatically on an interval. In short-lived environments such as serverless functions, batch scripts, or notebooks, flush manually before the process exits, otherwise events can be lost or delayed.
- Ingestion path: On Langfuse Cloud with the Langfuse v4 preview enabled, data sent via Python SDK v4.7.0+, JS/TS SDK v5.4.0+, or direct OpenTelemetry ingestion with the
x-langfuse-ingestion-version: 4header appears in the unified tracing table in real time. Data sent via older SDK versions, or via direct OpenTelemetry ingestion without that header, can appear with a delay of up to 10 minutes. See Working with the observations table for upgrade steps.
If traces still do not appear after a few minutes, work through the missing traces troubleshooting guide. On self-hosted instances, see missing events after ingestion.
Was this page helpful?
Last edited
