Monitor and Control Executions

Use execution routes as the authoritative view of workflow progress. Provider state can lag or contain unrelated drift; Alfred's execution and stage records describe what the workflow has accepted and completed.

List Executions

curl --fail-with-body --silent --show-error \
  "$ALFRED_API_URL/v1/executions?limit=20" \
  -H "Authorization: Bearer $ALFRED_API_KEY" | jq

Filter by workflow, status, or principal:

curl --fail-with-body --silent --show-error \
  "$ALFRED_API_URL/v1/executions?workflow=add-certificate&status=failed" \
  -H "Authorization: Bearer $ALFRED_API_KEY" | jq

limit defaults to 50 and must be between 1 and 100.

Inspect One Execution

curl --fail-with-body --silent --show-error \
  "$ALFRED_API_URL/v1/executions/$EXECUTION_ID" \
  -H "Authorization: Bearer $ALFRED_API_KEY" | jq

Check these fields in order:

1. status for the overall outcome.
2. currentStage for the active or failed boundary.
3. stages[].status for completed work.
4. stages[].outputJson for durable results and mutation receipts.
5. stages[].errorJson for failure context.

Stream Progress

curl -N --fail-with-body \
  "$ALFRED_API_URL/v1/executions/$EXECUTION_ID/events" \
  -H "Accept: text/event-stream" \
  -H "Authorization: Bearer $ALFRED_API_KEY"

Reconnect with the last received event ID to request replay:

curl -N --fail-with-body \
  "$ALFRED_API_URL/v1/executions/$EXECUTION_ID/events" \
  -H "Accept: text/event-stream" \
  -H "Last-Event-ID: $LAST_EVENT_ID" \
  -H "Authorization: Bearer $ALFRED_API_KEY"

Cancel

Cancel an eligible non-terminal execution:

curl --fail-with-body --silent --show-error \
  -X POST "$ALFRED_API_URL/v1/executions/$EXECUTION_ID/cancel" \
  -H "Authorization: Bearer $ALFRED_API_KEY" | jq

Queued commands are canceled before starting. When a Workflow SDK run exists, Alfred sends cancellation to that run and updates the execution projection. An add-subdomain execution stops being cancelable after Route53 accepts its change. Terminal and otherwise ineligible executions return 409 execution_not_cancelable.

Resume or retry

Use the generic resume control for a paused run that still owns its active command lane, or for a failed or canceled update-lambda-version execution:

curl --fail-with-body --silent --show-error \
  -X POST "$ALFRED_API_URL/v1/executions/$EXECUTION_ID/resume" \
  -H "Authorization: Bearer $ALFRED_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{}' | jq

Alfred does not expose manual pause. Workflow-owned approval waits remain durable and normally continue through their specific approval route. For example, approve an add-subdomain replacement through POST /v1/executions/{id}/subdomain-plan/approve, not the generic resume route.

For a failed or canceled update-lambda-version execution, the resume endpoint validates the persisted input and queues unfinished work on the existing execution. A newly queued retry returns status: retry_queued; repeating the request while it is pending returns status: retry_already_queued. Other failed or canceled workflows return 409 execution_not_resumable; recover them through Slack or, where supported, by re-submitting the same logical workflow request.

Rolled-back executions and otherwise ineligible states return 409. See Recover Workflows before retrying or rolling back provider work.

Handle Control Responses