Sravan Vidiyala

You can't load-test an LLM agent with a dumb mock

The first thing I tried was the obvious thing: stand up a fake OpenAI endpoint that returns a hardcoded response, point the agent at it, and ramp up concurrent users. Mock the expensive dependency, is