Purrsurance: AI Multi-Agent Pet Insurance Platform

Purrsurance is an intelligent pet insurance management platform that leverages AI agents powered by AWS Bedrock to provide instant policy information, veterinary consultations, and appointment booking

Comment

Purrsurance - AI-Powered Pet Insurance Platform

Overview

Purrsurance is an intelligent pet insurance management platform that leverages AI agents powered by AWS Bedrock to provide instant policy information, veterinary consultations, and appointment booking through a conversational interface. The platform combines insurance management with AI-driven pet healthcare assistance to deliver a seamless user experience for pet owners.

Problem Statement

Pet owners often face multiple challenges:

  • Understanding complex insurance policies and coverage details
  • Determining if their pet's symptoms require immediate veterinary attention
  • Finding and booking appointments with appropriate veterinarians
  • Accessing reliable veterinary medical information quickly

Purrsurance solves these problems by providing an intelligent, conversational assistant that can handle insurance inquiries, provide preliminary veterinary assessments, and facilitate appointment booking - all in one place.

Key Features

🤖 Multi-Agent AI System

  • Intention Classifier Agent: Routes user queries to the appropriate specialized agent
  • Policy Manager Agent: Retrieves and explains insurance policy details, coverage, claims
  • Vet Doctor Agent: Provides preliminary health assessments using RAG-enhanced medical knowledge base
  • Booking Manager Agent: Searches for veterinary clinics and schedules appointments

💬 Conversational Interface

  • Real-time chat interface with streaming responses
  • Server-Sent Events (SSE) for live updates
  • Session isolation ensuring privacy across multiple browser tabs
  • Context-aware conversations with memory

🏥 Veterinary Knowledge Base (RAG)

  • Professional veterinary medical literature database
  • Clinical medicine guides, internal medicine references
  • Infectious disease manuals and dermatology resources
  • Evidence-based symptom analysis and urgency assessment

📋 Management Dashboard

  • Policy information display
  • Appointment scheduling and tracking
  • Pet and owner information management
  • Real-time status updates

Technology Stack

Frontend

  • Nuxt.js 3 - Vue 3 framework with TypeScript
  • Tailwind CSS - Modern, responsive UI styling
  • Pinia - State management
  • Server-Sent Events - Real-time event streaming

Backend & Infrastructure

  • AWS Bedrock - AI foundation models and agent orchestration
  • AWS Lambda - Serverless compute for microservices
  • Amazon DynamoDB - NoSQL database for policies and appointments
  • Amazon SNS - Event publishing and messaging
  • Amazon EventBridge - Event routing
  • AWS SAM - Infrastructure as Code deployment
  • API Gateway - RESTful API management

AI & Machine Learning

  • AWS Bedrock Agents - Multi-agent orchestration
  • Knowledge Bases for Amazon Bedrock - RAG implementation
  • Claude (Anthropic) - Large Language Model for conversations
  • Retrieval Augmented Generation (RAG) - Enhanced medical knowledge retrieval

Development Tools

  • pnpm - Fast, disk-efficient package manager
  • Node.js 20.x - Runtime environment
  • Git - Version control

Architecture

High-Level Architecture

┌─────────────────┐
│  Web Frontend   │
│   (Nuxt.js)     │
└────────┬────────┘
         │
         ├──────────────┐
         ▼              ▼
┌─────────────┐  ┌─────────────┐
│ Service     │  │ SSE Stream  │
│ Router      │  │ Lambda      │
└──────┬──────┘  └──────┬──────┘
       │                │
       ▼                ▼
┌─────────────────────────────┐
│   AWS Bedrock Agent         │
│   ┌──────────────────────┐  │
│   │ Intention Classifier │  │
│   └─────────┬────────────┘  │
│             │                │
│   ┌─────────┴────────┐      │
│   │                  │      │
│   ▼                  ▼      │
│ ┌──────────┐  ┌──────────┐ │
│ │ Policy   │  │ Vet Doc  │ │
│ │ Agent    │  │ Agent    │ │
│ └────┬─────┘  └────┬─────┘ │
│      │             │        │
│      │   ┌──────────┐       │
│      │   │ Booking  │       │
│      │   │ Agent    │       │
│      │   └────┬─────┘       │
│      │        │             │
└──────┼────────┼─────────────┘
       │        │
       ▼        ▼
┌─────────────────────────────┐
│     Action Group Tools      │
│ ┌───────────────────────┐   │
│ │ • Policy Details      │   │
│ │ • Context Save/Load   │   │
│ │ • Find Vet Clinic     │   │
│ │ • Book Appointment    │   │
│ │ • Doctor Visit Rec.   │   │
│ └───────────────────────┘   │
└──────────┬──────────────────┘
           │
           ▼
    ┌──────────────┐
    │  DynamoDB    │
    │  • Policies  │
    │  • Appts     │
    │  • Events    │
    └──────────────┘

Microservices Architecture

The platform consists of independent, scalable Lambda functions:

Core Services:

  • service-router - Main entry point, orchestrates Bedrock agents
  • service-backend - REST API for data retrieval
  • service-sse-stream - Real-time event streaming to clients
  • service-event-publisher - Publishes events to SNS

AI Agents:

  • agent-intention-classifier - Routes queries to specialized agents
  • agent-policy-manager - Handles insurance policy queries
  • agent-vet-doctor - Provides veterinary consultations with RAG
  • agent-booking-manager - Manages clinic search and appointments

Action Group Tools:

  • tool-policy-details - Retrieves policy information from DynamoDB
  • tool-context-save / tool-context-details - Session context management
  • tool-find-vet-clinic - Searches for veterinary clinics
  • tool-book-vet-clinic - Creates appointment bookings
  • tool-recommend-doctor-visit - Analyzes urgency and recommends care level

Data Models

Policies

Insurance policies for pets including:

  • Pet information (name, species, breed, age, medical history)
  • Owner information (contact details, address)
  • Policy details (coverage, limits, deductibles)
  • Medical history (conditions, vaccinations, medications)

Appointments

Veterinary appointments with:

  • Appointment details (date, time, duration, type)
  • Pet and owner information
  • Clinic details (name, address, specialty)
  • Status tracking (scheduled, confirmed, completed, cancelled)

Events

Real-time events for:

  • Session-isolated event streams
  • Policy updates and claim status
  • Appointment confirmations
  • Agent responses

Key Innovations

1. Session-Based Event Isolation

Each browser tab receives only its own events through unique session IDs, ensuring privacy and multi-tab support.

2. RAG-Enhanced Veterinary Agent

The Vet Doctor agent uses Amazon Bedrock Knowledge Base with professional veterinary literature to provide evidence-based preliminary assessments.

3. Multi-Agent Orchestration

Intelligent routing system automatically directs user queries to the most appropriate specialized agent without explicit user selection.

4. Serverless Scalability

Fully serverless architecture scales automatically from zero to thousands of concurrent users without infrastructure management.

5. Real-Time Conversational Experience

Streaming responses via SSE provide immediate feedback, creating natural conversation flow.

Use Cases

1. Policy Inquiry

User: "What does my insurance cover for dental procedures?"
Agent: Policy Manager retrieves coverage details and explains dental benefits, deductibles, and limits.

2. Health Concern

User: "My cat is vomiting and lethargic since morning"
Agent: Vet Doctor queries medical knowledge base, assesses symptoms, determines urgency, and recommends emergency care.

3. Appointment Booking

User: "I need to book a vet appointment for my dog next week"
Agent: Booking Manager searches for nearby clinics, shows available slots, and creates appointment.

4. Mixed Queries

User: "My dog needs vaccination - is it covered and where can I get it?"
Agent: System routes to Policy Manager for coverage info, then to Booking Manager for clinic search.

Deployment

The platform is deployed entirely on AWS using Infrastructure as Code:

  • Frontend: Can be deployed to S3 + CloudFront or containerized deployment
  • Backend Services: AWS SAM templates for Lambda functions
  • Database: DynamoDB tables with appropriate indexes
  • AI Infrastructure: AWS Bedrock agents with configured action groups

All services are independently deployable and scalable.

Security & Privacy

  • Session-based isolation for multi-user support
  • No hardcoded credentials (IAM roles and policies)
  • CORS-enabled APIs for secure frontend communication
  • Data encryption at rest (DynamoDB)
  • Secure event routing with session validation

Getting Started

Prerequisites

  • AWS Account with Bedrock access
  • Node.js 20.x or higher
  • AWS SAM CLI
  • pnpm package manager

Quick Start

  1. Deploy Backend Services

    make setup

  2. Configure Frontend

    cd apps/frontend
# Update .env with deployed API URLs
pnpm install
pnpm dev

  3. Access Application

    http://localhost:3001

Conclusion

Purrsurance demonstrates how modern AI capabilities can transform traditional insurance services into intelligent, conversational experiences. By combining AWS Bedrock's agent orchestration, RAG-enhanced knowledge retrieval, and serverless architecture, the platform delivers a scalable, cost-effective solution for pet insurance management with integrated healthcare guidance.

The multi-agent approach ensures specialized, accurate responses while maintaining a seamless user experience through intelligent routing and real-time streaming. This architecture can be adapted for various insurance domains, healthcare applications, and customer service scenarios.

Built with ❤️ and AWS Bedrock

Share this project:

Updates