Add slow-query detection and request budgets to Node without swapping your DB client

This is a practical integration post, not a benchmark flex.

The problem

Two different failures look the same in user land:

1. One query is genuinely slow.
2. Every query is "fine", but a single HTTP request executes dozens of them.

Traditional "log slow queries" tooling often optimizes for (1). It is easy to miss (2) until p95 latency jumps.

What queryd does

@olegkoval/queryd is a small library that wraps the boundary where your app talks to the database:

• Slow query warnings via a configurable threshold.
• Optional per-request budgets (query count + total time) when you run work inside runWithDbContext.
• Structured events you can forward to your logger or APM adapter.

It is driver-agnostic for common patterns (wrapTaggedTemplate, wrapQueryFn) and has an optional Prisma entry: @olegkoval/queryd/prisma.

Minimal example (postgres.js-style tagged template)

The README quick start is the source of truth; the idea is:

1. Create a SlowQueryDetector with thresholds + optional requestBudget.
2. Wrap your sql tagged template with wrapTaggedTemplate.
3. Wrap request work with runWithDbContext({ requestId, ... }, async () => { ... }).

Full copy-paste code lives here:

https://github.com/oleg-koval/slow-query-detector#60-second-quick-start

When not to use it

• If you already have end-to-end tracing + query instrumentation you trust, you may not need another layer.
• If you want a hosted UI and retention analytics, this is not that — it emits events; you choose the sink.

Call to action

Try the quick start on a branch and tell me where wrapping gets awkward:

https://github.com/oleg-koval/slow-query-detector/issues/new?template=integration-feedback.md