This page helps you take the first step, even if you have never used AI in your work before. It is short on purpose.
Match your situation to a starting point:
- New to using AI at work: start with the prompt library. Try the logs summary or the PromQL prompt on a real task today.
- Already using AI to read logs and write config: read the monitoring guide, then look at Stage 3 in the main roadmap (AI in CI/CD and GitOps).
- Ready to build your own tools: jump to Stage 5 and Stage 6 in the main README.
You have two main choices, and the right one depends on your data.
- Hosted models (an API from a provider). Easy to start, strong results. The catch: your input leaves your machine, so never send secrets or customer data.
- Local models (run on your own hardware, for example with Ollama). Your data stays with you. Good for sensitive work. The catch: setup takes more effort and results can be weaker for hard tasks.
A common pattern: use a local model for anything with sensitive data, and a hosted model for general help where the input is safe to share.
Try these on real work this week. Each one is low risk.
- Paste a confusing block of logs and ask for a summary. Compare it to what you already know.
- Describe a metric you want and ask for the PromQL. Check the query before you run it.
- Take a pull request and ask for a plain-language summary and a risk rating. Review it yourself as usual.
If AI helps on these three, you have found where it fits. Grow from there, one trust level at a time.
Before you paste anything, ask one question: is there a secret or customer detail in here? If yes, redact it or use a local model.
Before you act on any answer, ask one more: have I confirmed this against the real system? If no, check first.
Those two habits cover most of the risk.