Section 31 of 102
30. Agent Communication, Coordination and Deployment Topology
Stable section ID: S05-CON-010-SECTION-31 · 27 content blocks
System05 agents may communicate with humans, devices, services, robots, and other agents through authenticated, versioned, and permission-controlled interfaces.
Agent communications shall identify:
- Sender.
- Recipient.
- Message type.
- Time.
- Applicable project or building.
- Requested action.
- Authority basis.
- Priority.
- Validity period.
- Data classification.
- Required response.
- Correlation with related events or tasks.
Multi-agent coordination shall not create hidden authority. Each participating agent shall remain responsible for actions performed under its identity. One agent’s recommendation shall not automatically authorize another agent’s execution.
Coordination mechanisms may include task assignment, event publication, capability discovery, negotiation, plan synchronization, conflict detection, execution locking, progress reporting, and escalation.
Deployment topologies may include:
- Local device agents.
- Node or subsystem agents.
- Building-edge agents.
- Central Building Agent.
- Project or construction-site agents.
- Organizational agents.
- Approved cloud agents.
- Hybrid local-cloud agents.
- Distributed robotic agent networks.
Safety-critical building functions should retain sufficient local capability to maintain or enter a safe state during network or cloud-service failure. Deployment location shall not weaken identity, permission, privacy, cybersecurity, audit, or lifecycle requirements.