DEVELOPER TOOLS  ·  GRAPHQL API OPTIMIZATION  ·  BETA

dataloader-ai

dataloader-ai is a drop-in wrapper for dataloader in Node GraphQL servers.
Track batch size, cache hit rate, and latency right in your terminal — no account, no API key, no data leaves your machine.

Try it now — npx dataloader-ai demo Cloud dashboard — free GitHub
Early traction — npm downloads

Try it in 5 seconds

npx dataloader-ai demo

No install, no account, no API key. This simulation shows exactly what you'll see in your terminal.

dataloader-ai — live
 
▲ dataloader-ai  00:00:00
──────────────────────────────────────────────────────
user
  cache [████████░░░░░░░░░░░░░░░░░░] 34.2%
  avg=12.4ms p95=18.1ms batched=0 avoided=0 savings=$0.0000
  batch efficiency
  recommendation waiting for data...
 
product
  cache [████████░░░░░░░░░░░░░░░░░░] 22.1%
  avg=8.7ms p95=14.3ms batched=0 avoided=0 savings=$0.0000
  batch efficiency
  recommendation waiting for data...
 
──────────────────────────────────────────────────────
events streamed: 0

The problem

DataLoader fixes the N+1 problem, but most teams still have no visibility into whether their loaders are batching well, missing cache too often, or using the wrong batch size for production traffic. Generic APM tools rarely answer those questions cleanly.

The solution

dataloader-ai wraps your existing DataLoader usage, records batch, cache, and latency telemetry, and prints it to your terminal in real time. No account, no API key, no data leaves your server. Optionally, connect a free API key to get a cloud dashboard with historical trends and alerts.

Who it's for

Integration — 3 lines, zero config

// before
import DataLoader from 'dataloader'
const userLoader = new DataLoader(batchLoadUsers)

// after
import { DataLoaderAI } from 'dataloader-ai'
const userLoader = new DataLoaderAI(batchLoadUsers, {
name: 'user',
})

// same load/loadMany API — metrics appear in your terminal
const user = await userLoader.load(userId)

What you get

Your terminal becomes a live dashboard. Every 5 seconds, per-loader metrics print automatically:

Cache hit bar visual bar + percentage per loader — green/yellow/red at a glance
Latency stats avg and p95 per loader so you know where bottlenecks are
Batch sparkline rolling 12-batch efficiency visualization — spot trends instantly
Size recommendations increase, decrease, or hold — based on your latency target, not guesses

Sample benchmark

Internal test workload using repeated keys, parallel loads, and simulated backend latency. This is not a customer case study yet — it shows the kind of changes teams can inspect.

Before
  • Avg batch latency: 68 ms
  • Cache hit rate: 12%
  • DB round trips: 52
After instrumentation
  • Avg batch latency: 41 ms
  • Cache hit rate: 46%
  • DB round trips: 16

Works with

Apollo Server NestJS GraphQL Yoga Mercurius Node.js

Beta status

dataloader-ai is in open beta. Local mode (terminal metrics) works forever, for free. Cloud dashboard access is free during beta — get a key instantly, no approval needed.

Pricing

Local
Free forever
  • Terminal metrics & recommendations
  • No API key required
  • No data leaves your machine
  • Unlimited loaders
Cloud
$0 /mo
  • Everything in Local
  • Personal dashboard
  • Historical metrics & trends
  • Team dashboards
  • Alerting (webhook & email)
Team
$29 /mo
  • Everything in Cloud
  • Up to 10 team members
  • Shared dashboards & alerts
  • Priority support
  • Custom retention

Local mode is and always will be free. Cloud is free during beta. No auto-charges, no credit card required.

FAQ

Is this open source?

The SDK is MIT-licensed and on GitHub. The API and dashboard are hosted.

Does it work with my server?

Any Node.js GraphQL server using the dataloader package works — Apollo Server, NestJS, Yoga, Mercurius, or custom.

What data do you collect?

Batch size, cache hit/miss events, and batch latency per loader. No query content, no user data, no schema. See our privacy policy.

Will this slow down my resolvers?

Telemetry is sent asynchronously in a background flush. It adds <1ms per batch. If the API is unreachable, data is buffered locally and retried.

What happens if the API goes down?

Your DataLoader works normally. dataloader-ai wraps the existing API — if telemetry can't be sent, it queues and retries. Zero impact on your resolvers.

What happens after 90 days?

Beta keys expire. We'll contact you before that with renewal options. No auto-charges, no surprise billing.

Get your cloud dashboard — instant, free

Local mode works without an account. Want a hosted dashboard with historical trends? Enter your email and get an API key immediately. Free during beta.

Free during beta · No credit card · Key + setup instructions emailed to you

Your dashboard

Every approved beta user gets a personal dashboard with live loader metrics, cache hit rates, latency breakdowns, and batch-size recommendations. Log in with the credentials we email you.

Open dashboard

Privacy policy

What we collect: Batch size, cache hit/miss events, and batch latency per loader name. We also collect your email and stack description when you sign up.

What we don't collect: No query content, no user data, no schema, no resolver arguments, no HTTP headers.

How we use it: Telemetry is used to generate batch-size recommendations and dashboard metrics. Your email is used only for beta key delivery and important product updates.

Data retention: Telemetry is retained for the duration of your beta period (90 days). You can request deletion at any time by emailing support@dataloader-ai.com.

Third parties: We don't sell or share your data. Telemetry is sent over HTTPS.

Contact: support@dataloader-ai.com