XÂY DỰNG WEBSITE THƯƠNG MẠI ĐIỆN TỬ KINH DOANH THỨC ĂN VÀ PHỤ KIỆN THÚ Y TÍCH HỢP CHATBOT TƯ VẤN THÔNG MINH

Abstract

This thesis presents the design and implementation of PetHub, a web-based ecommerce system for pet food and accessories. The project addresses two related requirements: providing a coherent online purchasing process and supporting customers who require additional information before selecting products for dogs and cats. The system adopts a decoupled client-server architecture. The frontend is implemented with React and Vite, while the backend is developed with Spring Boot and exposes RESTful APIs. MySQL stores transactional and application data, and Amazon S3 stores product images. Spring Security, short-lived JSON Web Tokens, and rotating opaque refresh tokens enforce authentication, role-based authorization, and session revocation. Implemented functions include catalog search and filtering, cart and wishlist management, coupons, COD and VNPay payment, order tracking, reviews, inventory history, audit logging, and administrative management. PetHub AI is implemented using retrieval-augmented generation. Administratormanaged documents are normalized, divided into overlapping chunks, embedded, and stored with their metadata. At query time, relevant chunks and recent conversation messages are supplied to the language model. Product-related intents may also produce structured ChatAction objects; subsequent cart or checkout operations remain subject to the same authenticated backend validation as conventional interface requests. Evaluation in the reported test environment covered authentication, catalog operations, cart boundaries, coupon eligibility, checkout idempotency, payment notification processing, order transitions, inventory restoration, reviews, and chatbot behavior. The observed results were consistent with the expected business rules for the tested scenarios