Versions and logging

Every save is a version, and every run is a record. Between the two, you can see what the task was at any point and what happened each time it ran.

Versions

Each time you save a task, Aisle records a new version. Open the Versions tab on the task page to see them, newest first — version number, name, change summary, author, and when it was saved. The version that matches the live task is flagged Latest.

Open a version to see its script. When a previous version exists, a This version / Changes from vN toggle switches between the version's code and a line-by-line diff against the one before it.

Rollback

To go back to an earlier version, open it and click Restore this version. Aisle confirms, then makes that version the live task; your current draft is saved as a new version the next time you save, so nothing is lost. Restoring requires edit access to the task.

Run logs

Every run creates an execution record with its status, inputs, output, errors, and attempts. Open the Logs tab on the task page for the run history, and open a run for its detail. The run log shows each recorded step — prompt calls, integration calls, output, and errors — as it happens.

A run moves through these statuses:

Status	Meaning
Queued	Waiting for capacity or a concurrency slot
Running	The task is executing
Retrying soon	A retry is scheduled after a failed attempt
Cancelling	Cancellation has been requested
Cancelled	The run was stopped
Success	The task completed
Error	The task failed
Timeout	The task exceeded its timeout

Attempts and retries

When retries are enabled, one run can contain several attempts. Aisle records each attempt's number, status, duration, and error, and an Attempts selector on the run lets you switch between them.

For tasks that process batches, use checkpoints so a retry skips work that already succeeded:

for customer in customers:
    key = f"customer:{customer['id']}"
    if aisle.checkpoint.done(key):
        continue

    process_customer(customer)
    aisle.checkpoint.mark(key)

Checkpoints are run-scoped and survive retries, so a re-run resumes instead of starting over.

Cancellation

A run that is queued, running, or waiting to retry can be cancelled from its detail page. A running task first enters Cancelling while the runner acknowledges the request, then settles as Cancelled.

Where to go next

• Deployment — triggers, entry points, and run limits.
• SDK Reference — aisle.checkpoint, aisle.run, and retry configuration.