Result

String id = flexiq.enqueue(add, payload);

flexiq.awaitJob(id, Duration.ofSeconds(30));         // block until terminal
Optional<Integer> sum = flexiq.getResult(id, Integer.class); // 5

Awaiting

awaitJob(id, timeout) polls the store until the job reaches a terminal state (COMPLETE, DEAD, or CANCELLED) and returns its Optional<Job> snapshot; it throws a FlexiQException if the timeout elapses first. Any process sharing the storage can await a job by id.

Reading the result

Both return Optional.empty() while the job is still pending or running — pair them with awaitJob (or an EventName.SUCCESS listener on the worker) when you need to block.

Empty is also what a job whose result TTL has expired looks like, and what a DEAD or CANCELLED job looks like: neither produced a result. So an empty Optional is "no value to read", never "the job succeeded with nothing". When the difference matters, unwrap the job snapshot and read its status — getJob(id).map(job -> job.status).

The raw overload is the escape hatch, not the normal path. Reach for it to re-encode a result under a different serializer, to forward bytes without decoding them, or to inspect a payload a codec chain wrapped. Everything else wants the typed overload, which applies the same serializer the worker used to write it — a mismatch there is what turns a working task into a decode error at the reader.

The Job snapshot

getJob(id) returns the job's current state without waiting:

flexiq.getJob(id).ifPresent(job -> {
    System.out.println(job.status);      // JobStatus.RUNNING
    System.out.println(job.progress);    // 0–100, if reported
    System.out.println(job.error);       // last error, if any
});

Job carries id, queue, taskName, status, priority, createdAt, scheduledAt, startedAt, completedAt, retryCount, maxRetries, timeoutMs, progress, error, uniqueKey, namespace, and metadata. All timestamps are Unix milliseconds.