Async Work
Everything that shouldn't happen inside the request — queues, workers, schedules, and the retry that ran your job twice.
7 topics
When Work Leaves the Request
Deciding what to do inside the request and what to hand to a background worker — and the contract you owe the client either way.
core20 minQueues & Workers
What a job queue actually guarantees, how a worker should be written, and why your database is often the right broker.
core25 minRetries, Backoff & Dead Letters
Retrying only what's worth retrying, spreading the attempts so you don't amplify an outage, and giving permanent failures somewhere to go.
core25 minIdempotent Consumers
Writing a worker that produces the same result whether a message arrives once or five times, and doing it without a distributed transaction.
core25 minScheduled & Recurring Jobs
Cron in a world of many instances — running a job exactly once per tick, catching up after downtime, and not designing a nightly stampede.
core20 minLong-Running AI Jobs
The backend shape of a model call that takes minutes — partial results, cancellation, cost accounting, and retries that don't pay twice.
core25 minIdempotency
Making an operation safe to repeat, which is the only way a retry over an unreliable network can ever be correct.
core25 minshared