Source code
Cost
Performance
Stability
Releases
Cost controls
The code patterns behind a bill that doubles without a traffic spike.
Overview
Cloud spend is mostly written into the code, not the pricing page. lumioguard looks for the loops, fetches, and retry paths that quietly multiply invocations, egress, and log volume as usage grows.
What it looks for
- Polling a provider that already offers a webhook
- SELECT * and over-fetching on hot paths
- Retries and logging with no upper bound
- Caches with no TTL, and no cache where reads repeat
- Fan-out with no concurrency cap, and per-invocation client construction in serverless
Why it matters
A cron that re-scans everything each minute and a retry loop with no ceiling cost nothing on your laptop and a lot at 10,000 users. These are the findings that show up on an invoice before they show up in an error budget.
Rules in this check
| What it reports | Fix path |
|---|---|
| Polling where webhook exists | Guide |
| SELECT * / over-fetching | Pull request |
| Unbounded retries / log storm | Pull request |
| Missing cache TTL / unbounded cache | Pull request |
| Bundle bloat (serverless) | Guide |
| PrismaClient instantiated on every serverless invocation | Pull request |
| Cron runs every minute on full scan | Guide |
| Unbounded fan-out / no concurrency cap | Guide |
How the fixes arrive
Pull requestLow-risk changes arrive as a pull request on a lumioguard branch, with a breakage analysis attached. You review and merge it.
GuideChanges only you can make safely arrive as a step-by-step guide you can follow yourself or hand to your coding agent.
Run them all on your app
Connect your repo and your live services with read-only scopes. The first scan is free, and nothing changes without your approval.