cyberneticlibrary

Trace and evaluate LLM applications

langsmith-observabilityskillsetup L29,423
Orchestra-Research/AI-Research-SKILLs
What it does

Trace and debug multi-step LLM agents with latency breakdown and error telemetry

Best for

Multi-step agent debugging when you need visibility into every LLM/tool call and latency bottleneck.

Inputs
  • · agent execution trace
  • · LLM calls
  • · tool calls
Outputs
  • · trace DAG
  • · latency analysis
  • · error telemetry
Requires
  • · langsmith-sdk
  • · LangSmith cloud account
Preconditions

Agent instrumented with LangSmith decorators; API key configured; traces flushed

Failure modes

High trace volume causes quota exhaustion; PII in traces exposes sensitive data; clock skew breaks latency

Trust signals
  • · LangChain ecosystem
  • · OpenTelemetry compatible