opensre

Why we included this project

When production breaks, the evidence is usually split across logs, metrics, traces, runbooks, and Slack threads, and OpenSRE is a framework that gives you a programmable way to sort through it. You build your own AI site-reliability agents that connect to the 60+ tools you already run, then let them work an alert in a tool-calling loop until they produce a root cause report backed by evidence and a suggested next step. If your team is worn out by alert fatigue, this is a concrete starting point for agents that begin triaging before anyone gets paged, all running on infrastructure you control. You bring your own LLM and define the workflows, so you decide how much autonomy the agents get. The bundled synthetic incident suites and end-to-end tests are also handy for checking whether an agent genuinely improves root-cause accuracy before you trust it in production.