YanpengQi7/ai-reliability-copilot: Incident-to-Postmortem LLM Pipeline Shipping a Regression Suite and LLM-as-Judge Evals

GitHub·medium signal

ai-reliability-copilot (TypeScript, 103 stars) turns a production incident into a structured 9-section LLM response covering severity, root cause, mitigation, and postmortem. The standout is engineering rigor for an early repo: it ships a 5-scenario regression suite plus an LLM-as-judge eval pipeline, modeling how to test non-deterministic agent output. Useful as a reference pattern for builders who want their incident/SRE agents to be evaluated, not just vibes-checked.