Cost optimization insights
Idle, memory-pressure, overprovisioned, deploy-regression
How it works, with exact details
Percher has two suggestion surfaces. The dashboard overview shows a few immediate hints from data already loaded on that page, such as a high error rate or memory close to the account pool. The deeper analysis is available through percher_app_insights or GET /apps/:app/insights. It computes four rule kinds:
- idle — no counted traffic in recorded data over 30 UTC dates (app at least 14 days old). Review missing dates; no data produces no idle suggestion. Confirm an app is unused before deleting.
- memory-pressure — 2+ OOM kills in 14 days. Upgrade plan or reduce memory use.
- overprovisioned — on maker/max with <100 recorded req/day with all 14 UTC dates present and no recorded OOMs. Check account resource use and lower-plan limits before downgrading.
- deploy-regression — 3+ failed deploys in 24h. Something changed; review the latest build log.
The API/MCP result includes an actionHint when there is a safe follow-up tool, such as billing or crash diagnostics. The dashboard overview does not currently show this four-rule result or these action hints.
Suggestions are advisory. They never change a plan or delete an app automatically. Your AI assistant should explain the evidence and ask before following an action that changes billing or removes data.