Pantheon Agents
The Pantheon is a collection of 15 specialized LangGraph AI agents, each named after a Roman god and responsible for a specific business domain. Rex AI COO orchestrates these agents, routing user intents to the appropriate specialist.
Agent Roster
| Agent | Domain | Key Capabilities |
|---|---|---|
| Rex | Orchestration (AI COO) | Intent classification (14 categories), delegation, conversation management |
| Saturn | Executive/Strategy | Strategic planning, board presentations, competitive analysis |
| Pluto | Finance | Revenue tracking, P&L generation, cash flow, tax planning |
| Jupiter | Infrastructure | Service health monitoring, deployment management, capacity planning |
| Mercury | Accounting | Invoice tracking, reconciliation, accounts payable/receivable |
| Proserpina | People/HR | Employee lookup, benefits enrollment, org chart, onboarding |
| Fortuna | Revenue/Business | Revenue forecasting, pricing optimization, sales metrics |
| Aequitas | Legal/Compliance | Contract review, compliance checking, regulatory alerts |
| Apollo | Analytics/Insights | Data querying, trend analysis, custom report generation |
| Ceres | Inventory | Stock levels, reorder alerts, supplier management, waste tracking |
| Juno | HR/Scheduling | Shift planning, availability tracking, labor cost optimization |
| Vulcan | Production/Ops | Kitchen metrics, equipment status, quality control, production rates |
| Diana | Customer Support | Ticket management, response generation, escalation routing |
| Neptune | Logistics | Delivery tracking, route optimization, fleet status monitoring |
| Mars | Security | Threat detection, access audit trails, vulnerability scanning |
Architecture
Each agent follows the same LangGraph pattern:
backend/python/app/agents/pantheon/{agent_name}/
├── graph.py # LangGraph StateGraph definition
├── nodes.py # Node implementations with Spanner queries
└── state.py # TypedDict state definition
LangGraph Flow
# Typical agent graph structure
graph = StateGraph(AgentState)
graph.add_node("analyze", analyze_node)
graph.add_node("query", query_node)
graph.add_node("respond", respond_node)
graph.add_edge(START, "analyze")
graph.add_edge("analyze", "query")
graph.add_edge("query", "respond")
graph.add_edge("respond", END)
Intent Classification
Rex classifies incoming messages across 14 intent categories:
| Category | Routed To | Examples |
|---|---|---|
finance | Pluto | "What's our revenue?", "Show P&L" |
operations | Saturn | "How are stores performing?" |
infrastructure | Jupiter | "Service health status" |
accounting | Mercury | "Outstanding invoices" |
people | Proserpina | "Employee directory" |
revenue | Fortuna | "Revenue forecast" |
legal | Aequitas | "Compliance status" |
analytics | Apollo | "Generate report" |
inventory | Ceres | "Stock levels" |
scheduling | Juno | "Schedule for next week" |
production | Vulcan | "Kitchen efficiency" |
support | Diana | "Open tickets" |
logistics | Neptune | "Delivery status" |
security | Mars | "Security audit" |
API Routes
All Pantheon agents are accessible through the Python Analytics Service:
POST /api/v1/agent/founder/chat # Rex conversation
POST /api/v1/agent/founder/voice # Voice command
GET /api/v1/agent/founder/agents/list # List available agents
POST /api/v1/agent/founder/agents/{name}/invoke # Direct agent invocation
Custom Agents
Organizations can create custom agents that integrate with the Pantheon:
POST /api/v1/agent/founder/agents/custom # Create custom agent
GET /api/v1/agent/founder/agents/custom # List custom agents
Custom agents are stored in the custom_agents Spanner table and can be granted specific agent_delegations for data access.