PragyaShield AI Insurance Intelligence Platform

Enterprise-grade Java + Angular insurance platform upgraded from the previous portal with a visibly different UI and a real AI chatbot module.

Preserved login options

The original three login options are still present and unchanged in concept:

1. Customer
2. Agent
3. Admin

The new AI modules are added on top of those roles.

What is new in this version

1. AstraClaim AI ChatOps chatbot

A dedicated insurance AI chatbot is now added in three places:

• Sidebar route for Customer, Agent and Admin
• Full chatbot page: /customer/astraclaim-chatops, /agent/astraclaim-chatops, /admin/astraclaim-chatops
• Floating chatbot widget visible across the portal after login

Technical keywords shown in the UI and Java code:

• RAG Pipeline
• Vector Embeddings
• Semantic Search
• Tool Calling
• Prompt Guardrails
• PII Redaction
• Audit Ledger
• Redis Risk Cache
• Kafka Event Replay
• Circuit Breaker Fallback
• p95 Latency
• SLO Burn Rate
• Human-in-the-Loop Review
• Model Drift
• Manual Override Rate
• LLM Gateway
• Prompt Versioning
• Agentic Tool Router

2. Java backend AI chatbot APIs

New Java chatbot endpoint aliases:

POST /api/ai/astraclaim/chat
GET  /api/ai/astraclaim/profile

Backward-compatible aliases are also kept:

POST /api/ai/copilot/chat
GET  /api/ai/copilot/profile

3. enterprise-grade backend design

New Java modules added under:

backend/src/main/java/com/pragyashield/insurance/ai/chatops

Important classes:

• AstraClaimChatOpsService
• AstraClaimToolRouter
• AstraClaimPromptTemplateFactory
• AstraClaimPromptEnvelope
• AstraClaimToolResult
• ChatMemoryRepository
• AstraClaimConversationEvent

These modules show RAG, tool routing, prompt construction, guardrails, telemetry, conversation memory, audit reference creation and deterministic fallback design.

4. Different UI

The UI is changed from a standard insurance portal style to a dark AI-command-center style with:

• Floating chatbot launcher
• Dark glass dashboard shell
• AI runtime chips
• Technical prompt lab
• Tool-calling trace cards
• RAG source panels
• Guardrail badges
• Governance and observability wording

Run backend

cd backend
mvn spring-boot:run

Backend runs on:

http://localhost:8080

Run frontend

cd frontend
npm install
npm start

Frontend runs on:

http://localhost:4200

Demo login roles

Use the login screen role tabs:

• Customer
• Agent
• Admin

The project intentionally keeps the three login options while adding AstraClaim AI ChatOps on top.

Architecture talking points

This project can be explained as a production-minded insurance intelligence platform where AI is not placed directly in the critical path. The design uses deterministic rules, RAG-grounded answers, prompt guardrails, tool-calling traces, circuit breaker fallback, audit logging and human approval for high-risk decisions. This is safer than a simple chatbot because every AI recommendation is explainable, observable and reversible.