06/06—Open source · infrastructure
- Role
- Author & maintainer
- Context
- @aplanka/reliable-queue · MIT
- When
- 2025
- Links
- Live demo
01Problem
Agentic and LLM workloads are made of flaky, rate-limited async calls. Teams keep hand-rolling retry and concurrency logic around them, and getting it subtly wrong.
02What I built
A zero-dependency task queue for JavaScript and TypeScript with retries, concurrency limits, priorities, persistence, an event stream and React hooks.
03Architecture
- 01Enqueuepriority · metadata
- 02Schedulerconcurrency limit
- 03Workerretry · backoff
- 04Persistencesurvives reloads
- 05EventsReact hooks
04Technical decisions
- 01
Zero dependencies
A reliability primitive should not drag a dependency tree behind it. Small surface, easy to audit, safe to embed.
- 02
Events as the observability seam
Every state transition emits an event, so logging, tracing and UI progress are layered on from the outside instead of baked in.
05Impact
Published to npm under an MIT licence, with a public demo.
Stack
TypeScript / React / npm / Browser APIs