Sign in
All library
Library· Ops

Evaluating agents in production

Golden sets, LLM-as-judge, trace-based debugging, and the tooling that makes it repeatable.

#evals#observability
7 min read

"It worked on the demo" is not an evaluation. Production agents need continuous, automated evals across correctness, cost, and latency.

What to measure

  • Task success rate on a golden set of representative inputs.
  • Trajectory quality — did the agent take a reasonable number of steps, or loop forever?
  • Tool-call accuracy — right tool, right arguments.
  • Cost per resolution — tokens × price + tool spend.
  • Latency P50 / P95.

Tooling

Workflow

  1. Log every trace from day one.
  2. Curate 50–200 real examples into a golden set.
  3. Write LLM-as-judge or code-based graders per example.
  4. Run evals on every prompt or model change; block deploys on regressions.

See Hamel Husain's Your AI product needs evals for the canonical playbook.