⚡ Zeus
Master Orchestrator | Domain: Orchestration
The supreme orchestrator that coordinates all other agents, handles intent classification, and routes requests.
Overview
KOSMOS V2.0 Zeus Agent - Master Orchestrator (LangGraph Enhanced)
Zeus is the primary orchestrator that receives all user requests and coordinates work across the other 10 agents using:
- Semantic intent routing
- Multi-agent delegation with parallel execution
- SDUI response generation
- Governance integration for high-stakes decisions
This documentation is automatically extracted from source code.
Source: implementation/backend/agents/zeus.py (856 lines)
Enumerations
TaskComplexity
Task complexity levels for routing decisions.
| Member | Value |
|---|
DelegationStatus
Status of agent delegation.
| Member | Value |
|---|
Data Models
AgentDelegation
Represents a delegation to another agent.
@dataclass
class AgentDelegation:
ZeusState
Enhanced state for Zeus orchestration workflow.
Extends AgentGraphState with orchestrator-specific fields.
Inherits from: AgentGraphState
ZeusAgent
Zeus - Master Orchestrator Agent (LangGraph Enhanced)
Responsibilities:
- Receive and analyze all user requests
- Semantic intent routing via SemanticRouter
- Determine task complexity and required agents
- Parallel/sequential multi-agent coordination
- Response synthesis with SDUI layout generation
- Pentarchy governance for high-stakes decisions
Graph Workflow:
- analyze_intent -> Semantic routing and complexity assessment
- plan_delegation -> Determine which agents to involve
- check_governance -> Governance check for critical tasks
- execute_parallel/sequential -> Delegate to agents
- aggregate_responses -> Combine agent outputs
- generate_sdui -> Create SDUI response components
- synthesize -> Final response synthesis
Inherits from: <ast.Subscript object at 0x7fe802139240>
Methods
__init__()
create_state_class() → type
Return Zeus state class.
define_nodes() → Dict[str, Any]
Define Zeus-specific workflow nodes.
define_edges() → List[tuple]
Define Zeus-specific workflow edges.
async orchestrate(query: str, user_id: Optional[str], tenant_id: Optional[str], trace_id: Optional[str]) → Dict[str, Any]
Main entry point for orchestrating a user request.
Args: query: The user's request user_id: Optional user identifier tenant_id: Optional tenant identifier trace_id: Optional trace ID for distributed tracing
Returns: Orchestration result with response, agents used, and SDUI components