Agents Overview

Customer: "I need towels for room B12"
    → Agent output:
      { intent: "room_service", actions: [{ type: "hospitality.room_service.request", payload: { room: "B12", items: "towels" } }] }

Message received (WhatsApp)
  → Debounce (batch fast messages, 10s window)
  → PreFilterRules (regex, spam — no LLM, free)
  → Human Takeover check (skip if paused)
  → InboundRouter (1 LLM call → classify + route)
  → InboundFilter (per-agent rules)
  → AgentRunner (1 LLM call → structured output)
  → Context saved (context_summary for next message)
  → WebhookDispatcher (agent_run.completed → your backend)
  → [If confirmation_mode: "confirm"] Widget/API confirms → agent_run.actioned

agent = client.agents.create(
  name: "Hotel Concierge",
  system_prompt: "You handle hotel guest requests: bookings, room service, and maintenance.",
  auto_run_inbound: true,          # process inbound messages automatically
  debounce_seconds: 10,            # batch fast messages
  context_ttl_minutes: 60,         # remember conversation for 1 hour
  confirmation_mode: "confirm",     # require human approval before execution
  confirmation_timeout_minutes: 30, # auto-expire unconfirmed actions
  allowed_actions: [               # constrain output types
    "hospitality.booking.modify",
    "hospitality.room_service.request",
    "hospitality.maintenance.report"
  ],
  payload_schema: [                # constrain output fields
    { name: "room", field_type: "text", required: true },
    { name: "guest_name", field_type: "text", required: false },
    { name: "details", field_type: "text", required: true }
  ]
)

{
  "intent": "room_service",
  "confidence": 0.95,
  "actions": [
    {
      "type": "hospitality.room_service.request",
      "payload": { "room": "B12", "items": "extra towels" }
    }
  ],
  "language": "en",
  "suggested_reply": "I'll have extra towels sent to room B12 right away!",
  "analysis_summary": "Guest requested extra towels for room B12",
  "context_summary": "Guest in B12 requested towels, fulfilled"
}

{
  "event": "agent_run.completed",
  "data": {
    "run_id": "run_abc123",
    "agent_id": "agt_def456",
    "status": "completed",
    "from": "33612345678",
    "output": { "...structured output above..." },
    "tokens": { "input": 850, "output": 120 },
    "latency_ms": 420,
    "duration_ms": 450
  }
}