A run that ends in the Failed state is final — it won’t move again. But the task isn’t finished: you can start another attempt. This page covers reading what went wrong and starting that fresh run. For starting a run in general, see Execute a task.
Notice the failure
When a run fails, its task moves to the Failed state, and the task’s own page shows a Last Run Failed card. The card tells you which run failed — Run #2, say — and offers two next steps: View Run Details to see what happened, and the Re-run button to try again.
Read why it failed first
Before you re-run, find out why the last attempt stopped — a re-run that hits the same wall doesn’t help. Open the failed run’s run detail page (the View Run Details link on the card takes you there).
The run detail page shows a Failure Details card. It names the reason the run failed in plain terms and, where one is available, shows a longer explanation of the error. Read it alongside the run’s step history, which shows how far the attempt got before it stopped.
Use that to decide what to do:
- If the failure looks like a transient problem, a re-run on its own may well succeed.
- If the failure points to something in the task itself — an unclear request, a missing detail — fix that first. You might edit the underlying issue in your PM tool, or be ready to steer the next run when it asks you a question or shows you its plan.
Start the fresh run
When you’re ready, select Re-run — on the Last Run Failed card, or the Re-run button in the task’s header, or the task’s row on the Tasks page. There’s no confirmation dialog; the run starts as soon as the usual checks pass.
A re-run is a fresh run from the beginning — it does not resume the failed run from where it stopped. The new attempt reads the task, plans, and works through your pipeline from the start, exactly as a first run would. It’s recorded as the next run in sequence — Run #3 after a failed Run #2 — and the failed run stays in the task’s run history as its own record.
The same checks that gate any new run apply here: a minimum balance of $10.00, your organization’s daily spending limit, and the one-run-at-a-time rule for the project. If a check blocks the re-run, DIJJI.ai tells you which one — clear it and select Re-run again. See Execute a task for the detail on those checks.
Re-running isn’t admin-only — anyone on the project’s team who can open the task can do it.
Follow the new run
Once the re-run starts, the task returns to Queued, then Running, and follows the same path any run does — planning, your approval of its plan, then through the pipeline. Follow it from the task page’s Current run link. For every state the new run can pass through, see Run states.