I chaos-tested my AI agent's LLM provider. 66% of runs failed, and the fix was one argument.

Evals, load tests and injected provider faults for AI agents, offline and with no API key.

Read Original

Related