Chatbot Agent Architecture

AI System

Comprehensive guide to SHELTR's multi-agent chatbot system with public orchestrator and admin control panel

Version 1.0.0August 22, 2025

System Overview

SHELTR-AI has a sophisticated multi-agent chatbot system with two main components:

  • Public Landing Page Chatbot - Orchestrator-based system for public users
  • Super Admin Chatbot Control Panel - Configurable agent system for internal use
Public Landing Page Chatbot
Orchestrator-based system for public users

Agent Types (7 Specialized Agents)

  • Emergency - Crisis response and safety resources
  • Participant Support - Service booking and resource navigation
  • Donor Relations - SmartFund™ explanation and impact tracking
  • Public Information - Platform education and SmartFund™ model
  • Public Support - Getting started and donation guidance
  • Shelter Operations - Participant management and reporting
  • Technical Support - Platform issues and account problems

How It Works

  • • Intent Classification - Analyzes user message for intent
  • • Role-Based Routing - Routes to appropriate agent
  • • RAG Enhancement - Uses knowledge base for enhanced responses
  • • Response Generation - Generates contextual responses
Super Admin Control Panel
Configurable agent system for internal use

Agent Types (5 Configurable Agents)

  • General Assistant - Various tasks and Q&A
  • SHELTR Support - Platform-specific help
  • Technical Expert - Technical guidance and architecture
  • Business Analyst - Business insights and analytics
  • Creative Writer - Writing assistance and marketing

Features

  • • Session Management - Persistent chat sessions
  • • Agent Selection - Choose from configurable agents
  • • Model Selection - Switch between AI models
  • • Real-time Configuration - Edit agent instructions

System Integration

Firebase Storage Status

  • Knowledge Base: ✅ 10 documents loaded
  • Storage Bucket: gs://sheltr-ai.firebasestorage.app
  • Collections: knowledge_documents, knowledge_chunks
  • Embeddings: 62+ embedding chunks for chatbot RAG

Current Status

  • Public Chatbot: 🟡 All 7 agents in development
  • Super Admin Control Panel: 🟡 All 5 configurable agents in development
  • Model Integration: 🟡 OpenAI GPT-4o and GPT-4o Mini in development
  • Session Management: 🟡 Chat sessions persisting correctly

🟡 Current Status: All Systems in Development

Public Orchestrator System: 7 specialized agents for public users with RAG enhancement
Admin Control Panel System: 5 configurable agents for Super Admin with model selection
Knowledge Base Integration: Both systems share the same Firebase Storage knowledge base