Where are queues managed?

[borisghidaglia]

I digged into the code (especially world-local and core), but I'm struggling to pinpoint where exactly queues are managed.
For example I can see where world-local adds stuff to the queue, bot not where things get dequeued and how.

Could you point me in the right direction?

Thanks a lot in advance!

Boris

[pranaygp]

So each world can implement "queues" differently. For example, an AWS world can use SQS for the queue. On vercel, we use Vercel Queues as the queue

for the local world, it's actually implemented in-memory. When the runtime calls world.queue to run a step/workflow, it instantly tries to run the route in the same process.

workflow/packages/world-local/src/queue.ts

Lines 50 to 62 in bd81aea

	const response = await fetch(
	`http://localhost:${port}/.well-known/workflow/v1/${pathname}`,
	{
	method: 'POST',
	duplex: 'half',
	headers: {
	'x-vqs-queue-name': queueName,
	'x-vqs-message-id': messageId,
	'x-vqs-message-attempt': String(attempt + 1),
	},
	body,
	}
	);

---

Note that "queue" is technically not the best word for this because there's no actual ordering in some of the worlds. For example, with local world, It's really just invoking a callback immediately to allow the workflow to execute the step and vice versa

We wil be adding more features around concurrency/throttling/etc. that will make use of priority queues, but currently - the moment a step is invoked, we just trigger the step right away with retrying and persistence.