Kibo.AgenticAdmin

Agent Management

Overview

Agents are the core entities in Kibo Agent Admin. Each agent represents a conversational AI assistant with its own configuration, playbooks, tools, and conversation history.

Agent Properties

Basic Information

• Name: Unique display name for the agent
• ID: System-generated unique identifier (UUID)
• Status: Current operational state
  • Online (🟢): Agent is active and can handle conversations
  • Offline (🔴): Agent is inactive or unavailable

Configuration

• Language: Primary language code (e.g., EN, ES, FR, DE)
• Timezone: Location-based timezone (e.g., America/Los_Angeles)
• Site Context: Associated brand or context

Metrics

• Total Chats: Cumulative number of conversations handled
• Playbooks: Number of configured playbooks
• Uptime: Percentage of time the agent has been available
• Created Date: When the agent was first created
• Updated Date: Last modification timestamp

Agent Selection Interface

The main agents page provides an intuitive card-based interface for agent selection.

Search and Filter Options

1. Text Search
   • Real-time filtering as you type
   • Searches agent names and descriptions
2. Language Filter
   • Dropdown menu with available languages
   • Options include:
     • All Languages (default)
     • English
     • Spanish
     • French
     • German
     • Additional languages as configured
3. Pagination
   • Displays current page and total pages
   • Previous/Next navigation buttons
   • Shows count of displayed vs total agents

Agent Cards

Each agent is displayed as an interactive card containing:

┌─────────────────────────────────┐
│ [Status] Agent Name             │
│                                 │
│ 🌐 EN   📍 Los_Angeles        │
│                                 │
│ 💬 541    📚 8     ⏱️ 96.2%   │
│ Chats    Books    Uptime       │
│                                 │
│ Created: [Date]                 │
│ Updated: [Date]                 │
└─────────────────────────────────┘

Quick Actions

After selecting an agent, you’re presented with four main action areas:

1. Manage Playbooks
   • Create, edit, and organize conversational playbooks
   • Define different scenarios and user intents
2. Configure Tools
   • Set up external tools and integrations
   • Enable API connections and functions
3. Conversation History
   • View and analyze past conversations
   • Export conversation data for analysis
4. Agent Testing
   • Test agent responses in real-time
   • Validate playbook and tool configurations

Agent States and Lifecycle

Online State

When an agent is online:

• Can receive and process new conversations
• All tools and integrations are active
• Real-time metrics are collected
• Testing interface is available

Offline State

When an agent is offline:

• Cannot process new conversations
• Historical data remains accessible
• Configuration can still be modified
• Testing may be limited

State Transitions

Agents can transition between states based on:

• Administrative actions
• System maintenance
• Error conditions
• Scheduled downtime

Multi-Agent Management

Organization Strategies

1. By Purpose
   • Customer Service Agent
   • Sales Assistant Agent
   • Technical Support Agent
2. By Language/Region
   • English US Agent
   • Spanish LATAM Agent
   • French EU Agent
3. By Brand/Product
   • Brand A Shopping Assistant
   • Brand B Support Agent

Best Practices

1. Naming Conventions
   • Use descriptive, consistent names
   • Include purpose or region in name
   • Avoid special characters
2. Metric Monitoring
   • Regular uptime checks
   • Chat volume analysis
   • Performance benchmarking
3. Maintenance Windows
   • Schedule offline periods
   • Communicate downtime
   • Test changes thoroughly

Agent Performance Metrics

Key Performance Indicators

1. Chat Volume
   • Daily/weekly/monthly trends
   • Peak usage times
   • Average session duration
2. Uptime Percentage
   • Target: >99% for production agents
   • Calculated over rolling period
   • Excludes scheduled maintenance
3. Playbook Utilization
   • Most/least used playbooks
   • Success rates by playbook
   • User satisfaction scores

Optimization Tips

1. High Chat Volume
   • Review and optimize playbooks
   • Add more specific intents
   • Consider load balancing
2. Low Uptime
   • Check error logs
   • Verify tool configurations
   • Review system resources
3. Poor Performance
   • Simplify complex playbooks
   • Optimize tool response times
   • Update training data

Access Control

Permission Levels

1. Admin
   • Full access to all agents
   • Can create/delete agents
   • Modify all configurations
2. Editor
   • Modify playbooks and tools
   • View conversation history
   • Cannot delete agents
3. Viewer
   • Read-only access
   • View metrics and history
   • Cannot make changes

Tenant Management

• Agents are organized by tenant
• Search across tenants (if permitted)
• Tenant context shown in header

Integration with External Systems

API Access

• RESTful API for agent management
• Webhook notifications for events
• Bulk operations support

Data Export

• Export agent configurations
• Backup playbooks and tools
• Conversation history archives

Monitoring Integration

• Connect to monitoring platforms
• Custom alerts and notifications
• Performance dashboards