Is This Metric Normal? On-Call Baseline Answer Bot

When an on-call engineer posts a question like "is checkout p99 latency normal right now?" in a Slack channel, this bot pulls the current metric value from Datadog, compares it against the same metric's historical baseline (last 7 and 28 days, same time-of-day window), and answers in plain English with a verdict: normal, elevated, or anomalous. Every answer cites the Confluence runbook passage that defines the expected range so the engineer can trust and trace the call.

Use it during incidents or routine on-call shifts when someone sees a number on a dashboard and isn't sure whether it warrants action. It replaces the "let me scroll through three weeks of graphs" reflex with a cited, two-second answer.

1. 1A Slack mention or slash command triggers the flow with the engineer's question.
2. 2An OpenAI step parses the question into a Datadog metric query and time window.
3. 3Datadog returns the current value plus historical series for matching time-of-day windows.
4. 4Confluence is searched for the runbook section defining that metric's healthy range.
5. 5OpenAI synthesizes a verdict from the live value, the computed baseline, and the runbook text.
6. 6The bot replies in-thread with the verdict, the baseline comparison, and a runbook citation link.