Abstract

This project introduces a large language model (LLM)-powered chatbot designed to enhance online shopping experiences by providing personalized product recommendations, assisting customers in decision-making, and managing shopping carts. The chatbot is developed using Python, Streamlit, OpenAI API, and PostgreSQL to deliver seamless customer interactions. By leveraging retrieval-augmented generation (RAG), it ensures relevant responses based on user queries. This publication details the methodology, system architecture, experimental setup, and results, highlighting how AI-powered assistants can improve online retail experiences.

Introduction

E-commerce platforms often struggle with improving user engagement, decision-making, and conversion rates. AI-powered chatbots offer a solution by personalizing recommendations, understanding natural language, and managing transactions efficiently. This project focuses on integrating an LLM chatbot within an online shopping environment, offering interactive customer support, product suggestions, and purchase management. The chatbot interacts with a PostgreSQL database to store and retrieve shopping data, ensuring continuity across sessions.

Methodology

The chatbot follows a retrieval-augmented generation (RAG) architecture, combining knowledge retrieval with language generation to respond dynamically to customer queries. The methodology consists of:

1. Data Processing & Storage
   • PostgreSQL database stores product catalogs and shopping cart data.
   • SQL queries enable efficient product retrieval and recommendation.
2. LLM Integration & Response Generation
   • OpenAI’s LLM processes user queries and generates natural language responses.
   • The chatbot uses context-aware prompt engineering to personalize responses.
3. User Interaction via Streamlit
   • The chatbot interface is built using Streamlit, allowing for real-time conversations and cart modifications.
   • Users can add/remove products, receive suggestions, and finalize purchases.
   • The chatbot suggests products based on user preferences and database queries.

Results

Key Features:

• Works as a clerk at a shop: ask customers about things they really want to purchase, recommend corresponding products, etc.
  
• Show a list of products according to customer need. Price can be filtered out.
  
• Interact with a shopping cart
  
• Make a transaction (available if connect to the 3rd party)
  

Conclusion

This project successfully demonstrates the integration of LLMs with e-commerce platforms to provide an intelligent shopping assistant. The chatbot enhances user experience by offering accurate recommendations, interactive responses, and a seamless transaction process. Future improvements could focus on multi-turn conversational memory, integration with voice assistants, and enhanced personalization models.