Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.uselemma.ai/llms.txt

Use this file to discover all available pages before exploring further.

​
What tracing gives you

Every agent execution becomes a trace — a complete record of what happened, how long it took, and where it went wrong. Traces contain spans organized as a tree, with provider/framework spans (for example gen_ai.chat) and operation spans (for example tools, retrieval, and app logic).
application request
root span
trace ID · timing · metadata
gen_ai.chat
prompt · completion · tokens
tool.lookup-order
input · output
gen_ai.chat
second LLM call
The gen_ai.chat spans can come from Langfuse-supported framework/provider instrumentation or any other OpenTelemetry-compatible instrumentation that follows GenAI conventions, including OpenInference, Arize, Braintrust, or provider SDK instrumentation. Tool spans and app spans come from your framework, instrumentation library, or custom OpenTelemetry spans.

​
Setup guides

Langfuse integration (recommended for greenfield)

Add instrumentation from scratch with Langfuse, then export spans to Lemma.

Using a supported framework

Start from your framework page, then apply the shared exporter setup.

OpenTelemetry export

Add Lemma to an existing OpenTelemetry pipeline.
For greenfield implementations:
  1. Instrument agent execution with Langfuse or Langfuse-supported OTel integrations.
  2. Use LangfuseSpanProcessor as the processing layer.
  3. Export resulting spans to Lemma with an OpenTelemetry exporter and Lemma headers.
For existing implementations:
  1. Keep your current OpenTelemetry-compatible instrumentation (OpenInference, Arize, Braintrust, provider SDK instrumentation, or another OTel stack).
  2. Add Lemma as an OTLP trace export destination.
  3. Verify spans arrive in Lemma with expected trace IDs, parent/child relationships, and metadata.

​
Next Steps