Background jobs,
batteries included.
OpenQueue is a TypeScript job framework built on BullMQ and Redis. Typed tasks, retries, cron, flows, and a dashboard you didn't have to build — one config file, one CLI, zero glue code.
MIT · built on BullMQ · runs on Bun & Node
exports
47 active · 1,204 waiting · last 24h
completed
184,392
+12.4%
failed
287
-3.1%
active
47
+8/min
waiting
1,204
stable
Throughput · jobs/min
■ completed ■ failed
Tasks
If you can write a function, you can ship a job.
A task is an id, a schema, and a function. OpenQueue wires up the queue, the worker, validation, retries, backoff, and concurrency — then hands you a typed .trigger() to call from anywhere in your app.
- →Payloads validated by Zod before they ever hit Redis.
- →Logs and progress stream straight into the dashboard.
- →Defaults that make sense: 3 attempts, exponential backoff.
import { task } from '@openqueue/sdk';
import { z } from 'zod';
export const exportCsv = task({
id: 'export-csv',
schema: z.object({
reportId: z.string(),
format: z.enum(['csv', 'xlsx']),
}),
attempts: 5,
backoff: { type: 'exponential', delay: 1000 },
run: async (payload, ctx) => {
ctx.logger.info('building report', { id: payload.reportId });
const file = await build(payload);
await ctx.progress({ step: 'uploading' });
return { url: file.url };
},
});
// anywhere in your app — typed, validated, durable
await exportCsv.trigger({ reportId: 'rep_812', format: 'csv' });
Runs
Inspect every run. Replay any failure.
Full payloads, attempt history, logs, and progress for every job. Filter by status, search by ID, retry from the UI — and keep history in Postgres long after Redis forgets.
Runs 184,679
| Job ID | Name | Duration | Attempts | Age | |
|---|---|---|---|---|---|
| r_3aa18 | export-csv active | — | 1/3 | now | |
| r_3aa17 | send-invoice active | — | 1/3 | 2s | |
| r_3aa16 | export-csv completed | 284ms | 1/3 | 8s | |
| r_3aa15 | sync-contacts completed | 1.2s | 1/3 | 14s | |
| r_3aa14 | export-csv failed | 1.8s | 3/3 | 22s | |
| r_3aa13 | ocr-document completed | 612ms | 1/3 | 31s | |
| r_3aa12 | send-invoice completed | 4.7s | 2/3 | 44s | |
| r_3aa11 | send-receipt delayed | — | 0/3 | in 2m | |
| r_3aa10 | export-csv completed | 198ms | 1/3 | 1m | |
| r_3aa09 | sync-contacts waiting | — | 0/3 | 1m |
Errors
Triage failures, not log files.
Failures are grouped by error class, ranked by frequency, and trended over time — so a regression shows up as a spike the moment it ships, not as a support ticket three days later.
⊘Top errors · last 24h
287 events
- ECONNRESET184
- TimeoutError63
- ValidationError28
- S3AccessDenied12
ECONNRESET · exports · spiking +240% vs yesterday · first seen 4h ago
Flows
Pipelines as parent/child tasks.
enqueueFlow turns task.child trees into real DAGs with per-node status and duration. Drill into any node and see exactly where the time went.
order-fulfillment
flow_4f12 · 7 jobs · 2.4s total
validate-cart
completed 42ms
charge-card
completed 612ms
reserve-stock
completed 118ms
email-receipt
completed 287ms
generate-pdf
completed 894ms
notify-warehouse
completed 71ms
finalize
completed 33ms
Schedules
Cron you can change without redeploying.
Declare cron on a task for static schedules, or create and pause them at runtime with task.schedules.create(). See what runs next, what ran last, and how long it took.
Schedules
5 active · next: in 2m
| Name | Cron | Next | Last run | |
|---|---|---|---|---|
| ▶ | weekly-digest | 0 9 * * MON | in 2d 4h | ok · 1.2s |
| ▶ | cleanup-stale | */15 * * * * | in 4m | ok · 87ms |
| ▶ | reindex-search | 0 */6 * * * | in 1h 12m | ok · 4.7s |
| ▶ | trial-reminders | 0 10 * * * | in 18h | ok · 612ms |
| ❚❚ | rotate-creds | 0 0 1 * * | in 6d | ok · 28ms |
| ▶ | warm-cache | */5 * * * * | in 2m | ok · 142ms |
What's inside
Everything a job framework should do. Nothing it shouldn't.
No proprietary runtime, no per-job pricing, no second cloud. Your Redis, your Postgres, your code — OpenQueue just makes them feel like a product.
Typed payloads
Give a task a Zod schema and payloads are validated at trigger time and fully typed inside run — no any, no guessing.
Retries with taste
Exponential backoff by default. Throw NonRetryableError to stop, RetryableError to keep going — control flow, not string matching.
Cron + live schedules
Declare cron on the task, or create, list, and pause schedules at runtime through the schedules API. No redeploy to change a cadence.
Flows & DAGs
Compose parent/child trees with enqueueFlow and task.child. Per-node status and duration, rendered as a real graph.
Workbench built in
Overview, runs, flows, schedules, errors, metrics, and a test console — served by your worker, or mounted in your own app.
Postgres run history
The Drizzle adapter persists runs, logs, and alerts past Redis retention. Your audit trail survives a FLUSHALL.
OpenTelemetry native
Trace context propagates from enqueue to process automatically. Plug in any exporter and see jobs inside your existing traces.
Hot-reload dev
openqueue dev watches your task files and restarts the worker in milliseconds. The dashboard reconnects on its own.
CLI
Four commands. That's the whole workflow.
init scaffolds a config and your first task. dev runs the worker with hot reload. build compiles a self-contained server artifact, and start runs it in production. Failed runs retry with backoff and land in the dashboard — not in your logs at 3am.
$ bunx @openqueue/cli init
created worker.config.ts
created worker/example.ts
OpenQueue initialized
$ openqueue dev
[openqueue] 3 tasks · 2 queues · 1 schedule
[openqueue] workbench on http://localhost:8090/workbench
[openqueue] watching ./worker for changes
✓ export-csv r_3aa16 · 284ms
✓ sync-contacts r_3aa15 · 1.2s
✗ export-csv r_3aa14 · attempt 2/3 · retrying in 2s
Run it
One worker. Three ways to run it.
Develop
Next to your app
openqueue dev discovers tasks from your worker/ directory, hot-reloads on change, and serves Workbench locally. One process, no Docker, no YAML.
$ openqueue dev
Embed
Inside your server
Mount the dashboard in the app you already run. h3 and Next.js App Router adapters — or a plain fetch handler anywhere else — one catch-all route and Workbench shares your auth.
$ import { workbench } from '…/next'
Ship
In a container
openqueue build compiles a self-contained Nitro server artifact; openqueue start runs it. Or point the prebuilt Docker worker at your Redis and go.
$ openqueue build && openqueue start
Proven in production
It already runs a business.
OpenQueue handles every background job at Expensicat — invoicing, inbox OCR, bank sync, transaction matching, notifications. Same tasks, same retries, same Workbench you see above, in production around the clock.
Expensicatexpensicat.com →Stop babysitting Redis.
You already run the infrastructure. OpenQueue turns it into a job platform — in about the time it takes your CI to go green.