Interactive Spooled demo
Hit Forge Sprite and Spooled runs a real workflow — parallel frame jobs, workers with leases, automatic retries, and live events — not a mocked animation.
Hit Forge Sprite
Watch Spooled queue → workers → assemble
Jobs · events · scheduled sprite
Concepts
Not technical? Think of a coffee shop. Each idea below is an everyday analogy, then how it shows up when you hit Forge Sprite.
Coffee shop A job is your order slip: “latte, oat milk.” It waits until a barista picks it up.
In this demo Forging creates jobs such as “generate frame 3” and one later “assemble” job. Each is a unit of work with a JSON payload.
{
"queue": "spriteforge-frames",
"payload": {
"kind": "frame",
"frameIndex": 3,
"seed": "pixel-hero",
"paletteName": "neon"
}
}Coffee shop The queue is the line of slips above the machine. Work is claimed in order by available workers.
In this demo
Three queues: spriteforge-frames, spriteforge-assemble, and
spriteforge-public (scheduled sprites). Separate queues keep fast parallel work off heavier assembly.
Coffee shop Workers are the baristas. They pull an order, do the work, and mark it done. More workers usually means more throughput.
In this demo Defaults: 8 frame workers, 2 assemble workers, and 1 public-sprite worker (overridable via env). They claim jobs with leases and heartbeats.
Coffee shop A combo order: brew the coffee and warm the muffin in parallel, then box them only when both are ready.
In this demo
N frame jobs run in parallel; the assemble job dependsOn every frame. When all frames complete, assemble runs once.
Coffee shop Dropped a drink? Make it again. The order is not thrown away — it goes back so someone can finish it.
In this demo
The Chaos slider injects random failures on frame jobs. Spooled retries automatically (this forge sets
maxRetries: 5 on workflow jobs). Turn Chaos up and watch red → retry → green.
Coffee shop The board that says “Order #42 ready!” — the shop tells you; you do not keep asking.
In this demo Spooled streams job events over WebSocket to this server, which forwards them to your browser over SSE. A light reconcile poll fills gaps if an event is missed — so the UI stays honest under imperfect networks.
Coffee shop Coffee brewed at 6 AM every day without a customer asking — work on a timetable.
In this demo
Sprite of the Minute uses cron 0 * * * * * (every minute, UTC) on the
spriteforge-public queue. Open the Minute tab to see it update.
Coffee shop Staff hold the back-room keys. Customers never need them to place an order.
In this demo Your browser talks only to the SpriteForge server. The Spooled API key stays on the server and is never sent to the client. Job payloads are ordinary JSON stored by Spooled — this demo does not claim field-level encryption or compliance certifications.
Context
Same demo needs jobs, a DAG, workers, retries, a schedule, and live visibility. Honest sketches of common stacks — and where each is genuinely stronger. Full matrices live on spooled.cloud/compare.
Trade-offs: Excellent if you are already deep in AWS. You wire several managed services yourself; Step Functions charges per state transition; browser streaming is not built-in. Great fit for Lambda-native shops.
Trade-offs: Outstanding Node DX and low latency. You operate Redis and own durability, multi-tenant platform surface, and polyglot APIs. Prefer BullMQ when the app is one Node codebase and you want a library, not a hosted queue product. Compare →
Trade-offs: Best-in-class for complex, long-running business processes and saga-style logic. Heavier programming model and ops than a simple job queue. Prefer Temporal when durable execution is the product; prefer a job queue when most work is “enqueue → claim → complete.” Compare →
step.run for parallel workTrade-offs: Excellent DX for Next.js / serverless teams. Different mental model than pull-based workers with leases. Prefer Inngest when functions-as-steps are natural; prefer worker-first queues when long-running processes pull jobs continuously. Compare →
A job queue that is simple to start with, still useful at scale: durable jobs, workflows without a heavy DSL, schedules, and real-time visibility — cloud or self-host, REST + gRPC, polyglot workers.
dependsOn, not a new language.
Being honest: Spooled is not always the right tool. Sub-millisecond in-process work is happier in Redis/memory. Deep AWS shops may prefer Step Functions + Lambda integration. Complex multi-week sagas are Temporal’s home turf. Serverless-only teams often prefer Inngest’s function model. If you want a queryable job queue with workers, schedules, and live status you can demo in an afternoon — try Spooled.