cleandataengineering.aiAgent Harness, drawn simply
You are here: Testing › Load tests

Agent Harness Tutorial · Testing

Performance and load

No load test exists yet. The plan: Azure Load Testing at three times the expected storm peak, passing when the 95th-percentile reply time and the error rate stay inside their targets.

Today the real ceiling is the model quota, not our servers: Groq's free tier allows 8,000 tokens a minute, which is 3–5 agent-loop turns a minute in total (see budgets). Stateless servers on Container Apps scale out with no code change; k6 and Locust are the open-source alternatives.