StackJack 0.1.0.10729+6b0eaac
StackJack 0.1.0.10729+6b0eaac
This release changes how automation runs are recorded when a decision gate stops them before the AI session starts. These runs now show as Skipped instead of Failed, so frequently scheduled gated automations no longer trigger a stream of failure alerts.
Bug fixes
- When an automation's decision step has a required rule that isn't met and the step is set to abort, the run now ends as Skipped instead of Failed. Previously, an automation on a frequent schedule that was usually gated recorded a failure on almost every run. On a 10-minute schedule that meant roughly 24 failure emails a day, and those emails used up the daily alert allowance and showed failure cards. Gate-stopped runs now send no failure email and don't count against the alert allowance.
- A gate-stopped run's summary in run history begins with "Skipped by its gate:" followed by the rule that wasn't met, so you can see why it stopped. The summary names only the rule. It does not list any earlier steps that ran before the decision step.
- Only a real "not met" result counts as a gate stop. If the decision couldn't be evaluated, the run still ends as Failed and sends its usual failure email. This covers cases such as a missing or unreadable answer, no confidence value, no option selected, or an empty rule. A change in a vendor's response format therefore can't silently turn your gated automations into Skipped runs with no alert.
- Gate-stopped runs no longer have verification steps added, because no work was done.
- Credit usage for gate-stopped runs is unchanged. They are the only Skipped runs that use credits.
- When you delete a gate-stopped run's transcript, the run stays recorded as a gate stop instead of changing to another status.
Improvements
- Webhook-triggered automations that are stopped by their gate now respond with 202 (Accepted) instead of 503. Because 503 signals a temporary error, callers may have been retrying these requests. The webhook receipt log now records these runs as "Skipped by its gate" so you can tell them apart from runs that started.
- Scheduled automations that are stopped by their gate are no longer retried.
UI & UX changes
- In the automation builder's live test panel, a test run that ends before the AI session starts now shows the run's summary, such as the gate reason. Previously it showed a misleading "Test run stream failed" error.