Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
A sleep persists a wait and schedules a delayed queue message to wake it. Under `vercel dev` that message lives in an in-memory queue, so restarting the dev server mid-run loses the wake-up and strands the run in `running`. On dev worker startup, LocalWorld now re-enqueues every non-terminal run, retrying via RetryLoop until the dev queue is reachable; workflow_handler turns any elapsed wait into a wait_completed and continues. Replay is idempotent, so this should be fine?
Member
|
Yeah I think this is good; I'm working on implementing the local queue out of |
fantix
added a commit
that referenced
this pull request
Jul 28, 2026
The port to vercel-queue routed workflow topics through `vqs.sanitize_name()`, which doubles underscores, so `__wkf_workflow_` became `____wkf__workflow__`. That is neither what the TypeScript SDK publishes nor what the builder's queue trigger subscribes to. Subscribers also registered under a derived consumer group, while VQS pushes arrive with the `default` group the trigger declares, so nothing matched them. Both worlds now share `get_physical_topic()` and `QUEUE_CONSUMER_GROUP` in `world.py`, mirroring `@workflow/world-vercel` and `createWorkflowQueueTrigger`. The Vercel push callback hands the delivery to the world's own queue client, so acks and lease renewals inherit the proxy URL and token instead of falling back to an unconfigured default. The local world now re-enqueues non-terminal runs the first time it touches the queue. Its queue is in memory, so a sleeping run's wake-up was lost whenever the dev server restarted. This supersedes #131, which did the same thing before the queue moved in-process: back then recovery had to run on a throwaway thread and retry until `vercel dev`'s queue came up, and it predated queue namespaces.
fantix
added a commit
that referenced
this pull request
Jul 28, 2026
The port to vercel-queue routed workflow topics through `vqs.sanitize_name()`, whose reversible encoding doubles underscores, so `__wkf_workflow_` went out as `____wkf__workflow__` — outside the `__wkf_*` prefix that the builder's trigger pattern and platform service resolution key on, and out of step with what `@workflow/world-vercel` publishes. It also left `consumer_group` unset on `VercelWorld`, so subscribers registered under a name derived from the handler's qualname, which moves whenever that function is renamed. Dispatch is keyed on `(consumer_group, topic)`, so both matter. Both worlds now share `get_physical_topic()` and `QUEUE_CONSUMER_GROUP` in `world.py`. The Vercel push callback hands the delivery to the world's own queue client, so acks and lease renewals inherit the proxy URL and token instead of falling back to an unconfigured default. The local world now re-enqueues non-terminal runs the first time it touches the queue, and logs how many. Its queue is in memory, so a sleeping run's wake-up was lost whenever the dev server restarted. This supersedes #131, which did the same thing before the queue moved in-process: back then recovery had to run on a throwaway thread and retry until `vercel dev`'s queue came up, and it predated queue namespaces.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
A sleep persists a wait and schedules a delayed queue message to wake it. Under
vercel devthat message lives in an in-memory queue, so restarting the dev server mid-run loses the wake-up and strands the run inrunning.On dev worker startup, LocalWorld now re-enqueues every non-terminal run, retrying via RetryLoop until the dev queue is reachable; workflow_handler turns any elapsed wait into a wait_completed and continues. Replay is idempotent, so this should be fine?