Primary Structures:
-
Structured JSON for all machine-agent communications (enforced via JSON Schema, validated pre-transmission)
-
Natural language payloads embedded within JSON for Roundtable outputs (includes agent ID, rationale type, and timestamp per turn, enabling downstream auditability)
-
Provenance, confidence, and timestamp metadata included in every transmission to support interpretability and traceability
-
Globally unique
message_id(generated via UUID v4 with collision-check at dispatcher level) ensures end-to-end correlation across systems
{
"message\_id": "msg-9471a2f2d9",
"origin": "Keymaker",
"destination": "Vault",
"timestamp": "2025-04-30T02:15:00Z",
"job\_id": "LIO-2025-04-29-0001",
"step\_id": "Step\_3",
"payload": {
"type": "AdjudicationRequest",
"content": "\[Serialized Q\&A, Roundtable discussion logs, proposed cluster outputs\]",
"confidence\_scores": \[0.91, 0.87, 0.93\],
"provenance\_links": \["Doc-23\#Para5", "Doc-25\#Summary"\]
},
"status": "PROCESSING"
}
Example Schema (Keymaker → Locksmith):
{
"message\_id": "msg-913cc01",
"origin": "Keymaker",
"destination": "Locksmith",
"timestamp": "2025-04-30T03:04:00Z",
"payload": {
"type": "CognitionPackage",
"mom\_graph\_file": "MoM\_certified.json",
"memory\_snapshot": "Keymaker\_memorygraph.json",
"vault\_summary": "vault\_summary.yaml",
"contradictions": "contradiction\_log.json"
},
"locksmith\_injection": true
}