Graph Element Event Schema

The compacted graph.elements.events topic contains JSON lifecycle commands keyed by element_id. The producer schema is 2.0.

Envelope

Field	Type	Description
schema_version	string	"2.0"
event_id	string	Deterministic SHA-256 transition ID
event_type	string	"graph_element_state_changed"
operation	string	"upsert" or "delete"
element_id	string	Graph element ID and Kafka key
observed_at_unix_nano	integer	Observation or final expiry time
emitted_at_unix_ms	integer	Event emission wall-clock time
payload_hash	string or null	Hash of the complete upsert element
element	object or null	Complete node or edge on upsert

Node upsert

{
  "schema_version": "2.0",
  "event_id": "sha256",
  "event_type": "graph_element_state_changed",
  "operation": "upsert",
  "element_id": "service:checkout-api",
  "observed_at_unix_nano": 1784977200000000000,
  "emitted_at_unix_ms": 1784977200500,
  "payload_hash": "sha256",
  "element": {
    "kind": "node",
    "id": "service:checkout-api",
    "type": "service",
    "attributes": {
      "service.name": "checkout-api",
      "service.version": "2.4"
    }
  }
}

Edge upsert

{
  "schema_version": "2.0",
  "event_id": "sha256",
  "event_type": "graph_element_state_changed",
  "operation": "upsert",
  "element_id": "edge:sha256",
  "observed_at_unix_nano": 1784977200000000000,
  "emitted_at_unix_ms": 1784977200500,
  "payload_hash": "sha256",
  "element": {
    "kind": "edge",
    "id": "edge:sha256",
    "type": "calls",
    "source_id": "service:storefront",
    "target_id": "service:checkout-api",
    "attributes": {},
    "metrics": {
      "service_graph.request.total": 3,
      "service_graph.request.failed.total": 0
    }
  }
}

Delete

{
  "schema_version": "2.0",
  "event_id": "sha256",
  "event_type": "graph_element_state_changed",
  "operation": "delete",
  "element_id": "service:checkout-api",
  "observed_at_unix_nano": 1784977500000000000,
  "emitted_at_unix_ms": 1784977500100,
  "payload_hash": null,
  "element": null
}

Consumers must treat upsert as complete replacement, delete by element_id, preserve per-key Kafka order, and avoid independent extraction or expiry logic. Deterministic event IDs support deduplication under at-least-once delivery.