Huey

Requires Huey 2.4+ and <3, Python 3.11+. See the compatibility matrix for the full pin string.

Install

pip install z4j-huey

What it captures

Huey exposes signals:

Huey signal	z4j event
`SIGNAL_ENQUEUED`	`task.received`
`SIGNAL_EXECUTING`	`task.started`
`SIGNAL_COMPLETE`	`task.succeeded`
`SIGNAL_RETRYING`	`task.retried`
exhausted `SIGNAL_ERROR`	`task.failed`
terminal `SIGNAL_CANCELED` / `SIGNAL_EXPIRED`	`task.revoked`

Actions

Verb	How
`submit`	instantiate the registered Huey Task and call `huey.enqueue(...)`
`retry`	re-enqueue only with both complete operator replacement collections
`cancel`	call `huey.revoke_by_id(task_id)` to prevent pending work from starting; running work is not interrupted

Cancel is advertised with that pending-only contract, and its command result reports pending_only: true. Purge, bulk retry, and dead-letter replay are not advertised. Huey has no primitive here that meets those z4j contracts.

Periodic tasks

Huey's @huey.periodic_task(crontab(...)) decorators are code. Once the application has imported them, the companion adapter inventories Huey's periodic-task registry at agent start and surfaces those definitions as read-only schedules in the dashboard. It does not scan source files for decorated functions.

The companion adapter inventories these definitions but does not advertise pause, enable, disable, create, update, delete, or trigger-now.

Sqlite / in-memory huey

Single-process setups (SQLite huey, MemoryHuey) work but each process has its own queue. The dashboard reflects this.

Caveats

Huey's result expiry defaults to 7 days. If you look up an older task, you may get ResultTimeout - z4j renders this as "result expired" rather than lost.
No chord/group primitives.

See scheduler: huey-periodic.