Weekly Runbook Coverage Gap Report for Datadog Spikes

Closes the loop on the knowledge base itself. It looks back at the week's alerts and the retrieval confidence each one produced, finds the monitors that repeatedly returned weak or no runbook matches, and turns that into a ranked list of documentation gaps with suggested page titles.

Use it when on-call keeps hitting alerts the runbooks do not cover and you want a data-driven backlog of what to write. It makes 'we should document that' a concrete weekly task instead of a vague intention.

1. 1A weekly schedule starts the review.
2. 2The flow reads the past week's alert and retrieval-confidence logs from Postgres.
3. 3It aggregates low-confidence and no-match spikes by monitor and metric.
4. 4A logic step ranks gaps by alert frequency and absence of runbook coverage.
5. 5An LLM proposes runbook titles and outlines for the top gaps, checking Confluence to avoid duplicating existing pages.
6. 6The prioritized gap report is posted to the team's Slack channel.