Skip to main content
TruLayer speaks native OpenTelemetry. You can push spans in via POST /v1/otlp/traces using the standard OTLP/HTTP wire format, and pull them back out as OTLP TracesData JSON via GET /v1/otlp/export to hand off to any downstream OTel-compatible backend.

​
Ingest — POST /v1/otlp/traces

Accepts an ExportTraceServiceRequest payload in either protobuf (application/x-protobuf) or JSON. TruLayer maps OpenTelemetry GenAI semantic-convention attributes (gen_ai.system, gen_ai.request.model, gen_ai.usage.input_tokens, …) onto the native span model and routes each span through the standard Kafka → processor → TimescaleDB pipeline.

​
OTel Collector exporter config

The exporter appends /v1/traces to the endpoint, which matches our ingest path exactly.

​
Size and status codes

  • Request bodies over 20 MiB return 413.
  • Unsupported Content-Type returns 415.
  • Partial failures (some spans accepted, some rejected) return 200 with partial_success populated in the response body.

​
Export — GET /v1/otlp/export

Streams traces as OTLP TracesData JSON so downstream tooling can ingest without a TruLayer-specific adapter. Dashboard-scoped (Clerk session auth) and capped at 10,000 spans per response; paginate via the X-Next-Cursor response header.
Point any OTel-JSON-capable ingester at the resulting payload — Datadog, Honeycomb, Grafana Tempo, Jaeger, and Zipkin all accept it.

​
See also