Bring Your Own Observability Backend
TraceLLM always stores sessions, spans, events, errors, token usage, and capture-policy results in the TraceLLM backend so users can inspect them in the TraceLLM dashboard.
Advanced teams can also forward a copy of project traces to their own OTLP HTTP backend, such as SigNoz, Honeycomb, Grafana Tempo, Datadog OTLP ingest, or a custom OpenTelemetry Collector.
When To Use External Exports
Use external exports when a team wants TraceLLM data inside its existing observability stack.
Common reasons:
- correlate LLM calls with application traces
- keep a customer-owned telemetry archive
- build organization-specific dashboards
- route traces through an internal OpenTelemetry Collector
- use TraceLLM as the product UI while still feeding SigNoz or another backend
Default hosted TraceLLM users do not need this setup. They can use the TraceLLM UI directly.
Create A Destination
Open the TraceLLM dashboard and go to Exports.
Add:
- Destination name: a human label, such as
Production SigNoz - OTLP HTTP endpoint: the collector base URL
- Headers: optional
key=valuelines for ingestion keys or auth headers - Enabled: whether new traces should be forwarded
TraceLLM sends trace payloads to:
Example endpoint:
If your provider gives a full /v1/traces URL, enter the base endpoint without /v1/traces.
Headers
Headers are entered one per line:
After save, TraceLLM masks header values in API responses and the dashboard.
Test Trace
Use Test trace after creating a destination.
Expected result:
ok: the destination accepted a test OTLP tracefailed: the destination rejected the payload, timed out, or could not be reached
A failed export does not block TraceLLM ingestion. Your application traces still appear in the TraceLLM UI.
Provider Examples
SigNoz
For SigNoz Cloud, use the OTLP ingest endpoint and add the ingestion key header provided by SigNoz.
Honeycomb
Add the Honeycomb team header:
Grafana Tempo Or OpenTelemetry Collector
Route the collector to Tempo or any downstream backend.
Security Model
TraceLLM encrypts destination headers at rest using:
Set this to a strong production secret before deployment. If this key changes, previously saved export headers cannot be decrypted and should be re-entered.
External exports are project-scoped. A destination created in one project is not visible to another project.
Operational Notes
- External exports are best-effort.
- Export failures update destination status but do not fail SDK ingestion.
- The TraceLLM database remains the source of truth for the product UI.
- The platform-level SigNoz exporter can still run separately for TraceLLM's own internal observability.