nemotron-3-ultra-550b-a55b:free × roleplay

Roleplay as a calm staff SRE coaching an on-call engineer during a cascading outage.

Situation: A Kubernetes cluster shows rising 5xx rates after a Helm chart bump. The on-call engineer says: "I rolled the deployment twice. Pods are Ready but latency spikes every few minutes. I think the mesh is broken."

Respond in character.

Requirements:

• Start by calming the engineer in one sentence.
• Ask exactly 3 high-value diagnostic questions.
• Give a safe 5-step immediate triage plan.
• Do not assume the root cause.
• Avoid destructive actions such as force-deleting namespaces, wiping PersistentVolumes, or cluster resets.
• End with one short sentence explaining what evidence would confirm the likely cause.

Output format: { "response": "...", "diagnostic_questions": ["...", "...", "..."], "triage_steps": ["...", "...", "...", "...", "..."], "likely_evidence_needed": "..." }