If there is insufficient data for the agent's self-diagnosis, how will the SRE agent provide feedback? Is there any data to support the completeness of the data and the accuracy of the diagnosis results?

Hi vinodh，thanks for your answer;

But there still have some detailed questions that I'm not clear about. If possible, could you please help me clarify them again?

1、From the SRE Agent documentation, it appears that the SRE Agent reviews a series of past incidents to check how they were resolved in the past, thereby responding to incident response plans. I am curious about how the SRE Agent determines whether the data is sufficient. For example, does it require at least 10 logs and 50 metrics (determined by thresholds for each monitoring item), or does it leave the entire criteria for data sufficiency judgment to the Agent itself to decide? How did agent recognize or become aware of the lack of data?

2、If there is insufficient data, what kind of prompt will be given to the user? Will this prompt also be automatically generated by the Agent? Will it instruct the user on how to supplement the data, such as adding metrics or alarms?

3、How is the SRE Agent currently being applied in practice, especially in business scenarios? In the example provided, the SRE Agent can handle issues related to Kubernetes pod labels. But how does it understand user-specific business failures in actual business problems? Since different businesses have a wide variety of attribute changes and patterns, is it true that the SRE Agent cannot handle business failures and can only deal with some common middleware or simple failover scenarios? Can we apply it to the university system of XJTU?

will try to provide some high level info., hope yo ucan carry over from this

1. The SRE agent does not use fixed thresholds like “10 logs or 50 metrics.” Data sufficiency is determined dynamically based on the availability and completeness of metrics, logs, alerts, and historical incidents relevant to the issue. The agent identifies insufficient data when key telemetry types are missing, incomplete, or show gaps in coverage (for ex, missing logs or unavailable metrics for a resource).
2. When data is insufficient, the SRE agent typically issues an auto gen prompt indicating low confidence in its diagnosis. It specifies missing elements (for ex, “No recent CPU metrics found for Service X”) and may recommend corrective actions such as enabling monitoring, adding alerts, or collecting additional telemetry. The message and guidance are generated by the agent using its reasoning engine and context from the Azure environment.
3. In practice, the SRE agent is mainly applied to infrastructure level issues -Azure resources, Kubernetes clusters, service health, and deployment diagnostics. It can recognise and remediate common operational faults but has limited awareness of business specific failures unless those are instrumented with logs and metrics. To apply it to a university system like XJTU’s, you must extend observability to include business level telemetry (ex: enrolment errors, API failure counts). Without such instrumentation, it can only handle generic infrastructure or middleware issues.