Geely Auto

Solution

Geely built an agentic-operations system on Alibaba Cloud around a four-layer model of data, modeling, diagnosis, and knowledge capture. Cloud Monitor 2.0 unifies metrics, logs, traces, events, and change records from scattered environments so operations intelligence starts from one data base instead of many fragmented sources. UModel then creates an operations digital twin by associating cloud resources, service topology, and Geely's internal business models in one context, which lets the intelligent agent follow impact paths and reason about likely root causes. On top of that model, STAROps automates evidence collection, filters large numbers of low-value alerts in seconds, and returns diagnosis paths or recovery ideas. Geely is also turning operational experience into reusable Skills, Runbooks, and healing commands so incident response becomes an asset the AI system can reuse instead of a one-time human effort.

Data flow

Observability and change data from Geely's systems is unified in Cloud Monitor 2.0, associated into a digital twin through UModel, and then analyzed by STAROps for evidence gathering, alert filtering, diagnosis, and runbook-based healing suggestions. The outputs are fed back into operations teams and reusable skills for future incidents.