ActionTriggered
Delivery record for a single action invocation after a condition fires.
Actions are best-effort — status='failed' does NOT roll back the decision. The full pipeline response returns HTTP 200 regardless of action outcomes.
payload_sent is populated for triggered actions (webhook / workflow payloads). error is populated only when status='failed'.
Delivery status of a single action after a decision fires.
triggered — action was invoked and delivery succeeded skipped — fire_on rule not met (e.g. fire_on='true' but decision=False) failed — action was attempted but delivery raised an error would_trigger — dry_run mode; action would have fired but was not invoked
Possible values: [triggered, skipped, failed, would_trigger]
payload_sent object
- object
- null
error object
- ErrorResponse
- null
error objectrequired
The inner error object serialised inside ErrorResponse.
This is the exact shape the API spec requires: { "type": "...", "message": "...", "location": null, "suggestion": null }
Callers must branch on type, never on message.
Canonical error type identifiers. The string value is the wire value.
Rules (from API spec):
- Callers must branch on .type, never on .message.
- Message text is for humans; type is for machines.
- The enum spelling is the ground truth for YAML, HTTP, and Python alike. In particular: z_score (not zscore), not_found (not 404), conflict (not 409).
Possible values: [syntax_error, type_error, semantic_error, reference_error, parameter_error, graph_error, execution_error, execution_timeout, auth_error, not_found, conflict, rate_limit_exceeded, bounds_exceeded, action_binding_failed]
location object
- string
- null
suggestion object
- string
- null
{
"action_id": "string",
"action_version": "string",
"status": "triggered",
"payload_sent": {},
"error": {
"error": {
"type": "syntax_error",
"message": "string",
"location": "string",
"suggestion": "string"
}
}
}