Does Langfuse support OpenTelemetry logs, metrics, or span links?
Langfuse ingests traces only, at /api/public/otel/v1/traces over OTLP/HTTP with protobuf or JSON payloads. gRPC is not supported.
| Signal | Supported |
|---|---|
| Traces (spans, attributes, events) | Yes |
| Metrics | No. /api/public/otel/v1/metrics accepts requests so exporters do not fail, but the data is discarded. |
| Logs | No. There is no logs endpoint. |
| Span links | No. Links on spans are ignored during ingestion and are not shown in the UI. |
Any language with an OpenTelemetry SDK, for example Java, Go, or .NET, can send traces to Langfuse with an OTLP/HTTP exporter. To show token usage, cost, and model information, set the attributes described in OpenTelemetry. Langfuse calculates metrics such as cost and latency from the ingested traces; see Metrics.
Send logs and metrics to ClickStack
For OpenTelemetry logs and metrics, use ClickStack, the open-source observability stack built on ClickHouse. ClickStack ingests logs, metrics, and traces for operational monitoring, while Langfuse focuses on inspecting, evaluating, and improving your AI application.
Both accept OpenTelemetry, so you can export the same traces to Langfuse and ClickStack and keep sending logs and metrics to ClickStack only. See Langfuse and ClickStack for the setup.
Last updated on
