Prometheus

Infrastructure

Query metrics, investigate alerts, and understand your system's health — Neotask speaks PromQL through OpenClaw.

What You Can Do

Query Metrics Without Writing PromQL

Describe what you want to measure — "show me CPU usage per pod in the production namespace over the last hour" — and Neotask writes and runs the PromQL expression, returning results in a readable format with context about what the numbers mean.

Investigate Active Alerts

When an alert fires, ask Neotask to pull the underlying metrics, show you the trend that triggered it, and explain whether it looks like a spike or a sustained degradation. Get from "alert fired" to "root cause" faster.

Build Alerting Rules

Describe your alerting intent in plain English — "alert if any pod restarts more than 5 times in 10 minutes" — and Neotask generates the correct PromQL expression, sets sensible labels and annotations, and formats the rule in the Prometheus YAML format.

Explore Available Metrics

Ask what metrics are being scraped from a specific target, which jobs are currently down, or what labels are available on a particular metric. Neotask queries the Prometheus API to explore your metric catalog.

Analyze Trends and Capacity

Ask for a summary of how a key metric (like request rate, error rate, or disk usage) has trended over the past week. Get plain-language interpretations: "your error rate doubled on Tuesday evening, correlating with deployment v2.3."

Try Asking

  • "What is the current CPU usage for each node in my cluster?"
  • "Show me the 95th percentile request latency for the payment-service"
  • "Write a PromQL query to track memory usage per namespace"
  • "Which Prometheus targets are currently down?"
  • "Generate an alerting rule that fires when disk usage exceeds 85% for 5 minutes"
  • "Why did the HighMemoryUsage alert fire this morning?"
  • "Show me error rate trends for the past 7 days by service"
  • "What metrics are being scraped from the node-exporter job?"

    • Pro Tips

  • Ask for rate() vs irate() guidance when building queries — Neotask explains which is appropriate for counters in your use case.
  • Include the time range in your prompts: "over the last 30 minutes" or "since yesterday at 9am" for relevant results rather than the Prometheus default range.
  • Ask Neotask to explain a PromQL expression you inherited — paste the query and ask "what does this measure and are there any issues with it?"
  • Use label matchers in your prompts: "only for the frontend service in production" — Neotask will add the correct label selectors to the query.
  • When creating alerting rules, ask Neotask to include a runbook_url annotation pointing to your internal wiki for faster on-call response.

    • Works Well With