SAP Automation Pilot Executions API Management

This skill manages command executions in SAP Automation Pilot - triggering, monitoring, controlling, and troubleshooting executions.

Prerequisites

1. Set the following required environment variables:

export AUTOPI_HOSTNAME="emea.autopilot.cloud.sap"
export AUTOPI_USERNAME="your-username"
export AUTOPI_PASSWORD="your-password"
export AUTOPI_DEFAULT_CATALOG="mycommands-<<<TENANT_ID>>>"

1. Ensure curl is available in your environment.

Execution Lifecycle

Statuses

Terminal statuses (FINISHED, ABORTED, SUSPENDED) mean the execution is complete and no further actions can be applied. Completed executions can be retriggered and deleted.

Active statuses (RUNNING, PAUSED, INPUT_REQUIRED, FAILED) mean the execution is still live and can be acted upon.

Actions

Each action is only valid from specific statuses — applying an action from the wrong status returns HTTP 409.

EXEC_ID="execution-uuid-here"

# Pause
curl -s -X POST -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  -H "Content-Type: application/json" \
  -d '{"type": "PAUSE", "reason": "Blocking on manual verification"}' \
  "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/actions"

# Resume
curl -s -X POST -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  -H "Content-Type: application/json" \
  -d '{"type": "RESUME", "reason": "Resuming after verification"}' \
  "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/actions"

# Retry
curl -s -X POST -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  -H "Content-Type: application/json" \
  -d '{"type": "RETRY", "reason": "Retrying after transient failure"}' \
  "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/actions"

# Abort
curl -s -X POST -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  -H "Content-Type: application/json" \
  -d '{"type": "ABORT", "reason": "Superseded by a newer run"}' \
  "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/actions"

Triggering Executions

Note: To retrieve logs after execution, include feature:logs in the tags of the trigger request. This tag is opt-in and cannot be added retroactively — an execution triggered without it will return no logs from GET /executions/{id}/logs.

Trigger a Command

Add "feature:dryRun": "" to the tags to execute without making actual changes.

# Basic trigger (opt-in feature:logs tag enables log retrieval)
curl -s -X POST \
  -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  -H "Content-Type: application/json" \
  -d '{
    "commandId": "my-catalog:MyCommand:1",
    "inputValues": {
      "param1": "value1",
      "param2": "value2"
    },
    "tags": {"feature:logs": ""}
  }' \
  "https://$AUTOPI_HOSTNAME/api/v1/executions"

# With dry run mode (adds feature:dryRun tag)
curl -s -X POST \
  -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  -H "Content-Type: application/json" \
  -d '{
    "commandId": "my-catalog:MyCommand:1",
    "inputValues": {"param1": "value1"},
    "tags": {"feature:logs": "", "feature:dryRun": ""}
  }' \
  "https://$AUTOPI_HOSTNAME/api/v1/executions"

All inputValues entries must be strings on the wire, regardless of the declared parameter type. The runtime converts each value to its declared type inside the executor:

"inputValues": {
  "appName": "my-app",
  "instances": "3",
  "enabled": "true",
  "regions": "[\"cf-eu10\",\"cf-us10\"]",
  "config": "{\"maxRetries\":5}"
}

Monitoring Executions

List Executions

# List all (up to limit)
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions?limit=100"
# Filter by status
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions?status=FAILED&limit=50"
# Filter by command
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions?commandId=my-catalog:MyCommand:1"
# Filter by time (epoch milliseconds)
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions?startedAfter=1707000000000"

Get Execution Details

EXEC_ID="execution-uuid-here"
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID"

Get Execution Summary

# Get counts by status
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/summary"

Get Input/Output

# Get execution input
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/input"
# Get execution output (only for FINISHED)
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/output"

Controlling Executions

All actions use the same endpoint: POST /api/v1/executions/$EXEC_ID/actions

See the Actions table in the Execution Lifecycle section for valid statuses and expected outcomes.

EXEC_ID="execution-uuid-here"

# Reset — resets to a previous child command (provide the executor path via "value")
curl -s -X POST -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  -H "Content-Type: application/json" \
  -d '{"type": "RESET", "value": "stepAlias", "reason": "Retry after fixing the input"}' \
  "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/actions"

# Comment — add a note to the Action Log (max 10 per execution)
curl -s -X POST -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  -H "Content-Type: application/json" \
  -d '{"type": "COMMENT", "reason": "Escalated to L2 on-call"}' \
  "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/actions"

# List action history
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/actions"

Viewing Logs

Get Execution Logs

EXEC_ID="execution-uuid-here"

# Get paginated logs
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/logs?page=0&maxPageSize=20"
# Get logs for specific executor
EXECUTOR_PATH="step1"
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/logs/$EXECUTOR_PATH"

Troubleshooting

Quick Troubleshooting

EXEC_ID="your-execution-id"

# 1. Get status and error
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID"

# 2. Check input used
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/input"

# 3. Check recent logs
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/logs?page=0&maxPageSize=10"

# 4. Check output (if FINISHED)
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/output"

Common Issues

Execution Stuck in RUNNING

1. Check logs for the current step
2. Look for external dependencies (HTTP timeouts, etc.)
3. Consider aborting and retrying

curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/logs?page=0&maxPageSize=20"

curl -s -X POST -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  -H "Content-Type: application/json" \
  -d '{"type": "ABORT", "reason": "Aborting stuck execution"}' \
  "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/actions"

Execution Failed

1. Get execution details to see error
2. Check logs for the failed step
3. Review input parameters

curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID"
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID/input"

Input Required

The execution is waiting for manual input:

curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID"
# Inspect the response — check the suspendedStep field

No Logs Available

Cause: The execution was triggered without the feature:logs tag, so logging was not enabled. Solution: Always include "feature:logs": "" in the tags object when triggering. Logs cannot be retrieved retroactively for executions that were triggered without this tag. Re-trigger the execution with the tag included.

Filtering & Pagination

Query Parameters for List Executions

Examples

# Last 24 hours, failed only
YESTERDAY=$(($(date +%s) * 1000 - 86400000))
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  "https://$AUTOPI_HOSTNAME/api/v1/executions?status=FAILED&startedAfter=$YESTERDAY&limit=50"
# Multiple statuses
curl -s -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  "https://$AUTOPI_HOSTNAME/api/v1/executions?status=RUNNING&status=PAUSED"

Delete Executions

Only completed executions can be deleted (not RUNNING).

EXEC_ID="execution-uuid-here"
curl -s -X DELETE -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" "https://$AUTOPI_HOSTNAME/api/v1/executions/$EXEC_ID"

Event Triggers

Trigger via Generic Event

curl -s -X POST \
  -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  -H "Content-Type: application/json" \
  -d '{
    "commandReference": "my-catalog:MyCommand:1",
    "event": {
      "type": "alert",
      "message": "High CPU usage detected"
    }
  }' \
  "https://$AUTOPI_HOSTNAME/api/v1/triggers/generic-event"

Trigger via ANS Event

curl -s -X POST \
  -u "$AUTOPI_USERNAME:$AUTOPI_PASSWORD" \
  -H "Content-Type: application/json" \
  -d '{
    "commandReference": "my-catalog:MyCommand:1",
    "event": {...ANS event payload...}
  }' \
  "https://$AUTOPI_HOSTNAME/api/v1/triggers/ans-event"

Error Handling

HTTP Status Codes

Permissions Required

Examples

Example 1: Trigger and monitor a command

User: "Run my health check command"

1. POST to /api/v1/executions with commandId, input values, and "tags": {"feature:logs": ""}
2. Capture the execution ID from the response
3. GET /api/v1/executions/$EXEC_ID and poll until status is no longer RUNNING
4. Report the final status and output

Example 2: Troubleshoot a failed execution

User: "My execution failed, what went wrong?"

1. GET /api/v1/executions/$EXEC_ID — extract status, error, progressMessage
2. GET /api/v1/executions/$EXEC_ID/input — verify the parameters that were used
3. GET /api/v1/executions/$EXEC_ID/logs?page=0&maxPageSize=20 — find the failed step
4. Report the root cause from the error message and log output

Example 3: Find and abort stuck executions

User: "Are there any stuck executions? Abort them."

1. GET /api/v1/executions?status=RUNNING to list all currently running executions
2. For each one, check startedAt to identify those running unexpectedly long
3. For each stuck execution, POST {"type": "ABORT", "reason": "Stuck execution"} to /api/v1/executions/$EXEC_ID/actions
4. Confirm each abort returned HTTP 202