You can trace LangGraph applications either by using the native LangSmith tracing or by integrating OpenTelemetry. Both methods provide valuable insights into application executions, token usage, latency, and costs.
Select a method based on your use case:
•Choose the native LangSmith Tracing if you want a quick and simple setup focused solely on LangSmith’s observability features, without dealing with additional dependencies or code changes.
•Choose OpenTelemetry Integration if your environment already uses OpenTelemetry for monitoring or you want a standardized, extensible approach to telemetry that can integrate with multiple tools and services beyond LangSmith.
For more information, see the LangSmith and OpenTelemetry documentation.