Hotel Registry with AI-powered Chatbot
Project Overview
This project implements a microservice-based application that simulates the core functionality of a hotel booking website. Key features include:
- Hotel Information & Booking Services
- RESTful APIs for backend processing and data retrieval
- React.js frontend for simple, component-based user interface
- Secure Authentication
- Stateless JWT-based authentication for users to access secured content
- AI-powered Chatbot Integration
- An NLP-driven conversational assistant embedded into the platform
- Leverages basic OpenAI API calls on user queries combined with conversational and/or data context
- Vector Similarity Search for Data Retrieval
- Vector embeddings and vector similarity search for retrieving contextually relevant records from the database
Project Architecture
The project is split into three different microservice applications:
Frontend - React UI
The React UI microservice techstack uses React.js and its related packages to build a dynamic, component-based UI, enable routing within the single-paged applications (SPA), and integrated Bootstrap for styling. Webpack is used to compile and bundle JS, CSS, and other static assets into optimized files that are ready to be served to the browser. Babel serves as the JS transpiler, converting modern JS (ES6+ features) and JSX syntax into backwards-compatible ES5 code for a wider range of browser support.
Backend - Database Access
The backend microservice is built with Spring Boot and powered by Spring Data JPA for database interactions with a PostgreSQL instance enhanced by the pgvector extension for vector-based similarity search. It uses Apache HTTP Client 5 to communicate with OpenAI’s APIs for both text responses and embedding generation. The generated embeddings are stored in the database and leveraged for semantic search (using vector cosine similarity), enabling the system to return relevant hotel or hotel-room data for natural language queries and responses. Spring AI helps streamline AI model integration, and Jackson is used for efficient JSON processing. This service is responsible for creating, retrieving, and managing all hotel, room, and booking records in the system.
JWT Security & Authentication
The JWT Security microservice is built with Spring Boot and powered by Spring Security for authentication and authorization. It uses Spring Data JPA with a PostgreSQL database for secure and efficient data persistence, and JJWT (Java JWT) for JSON Web Token handling. This service acts as a gateway, blocking unauthorized access to database resources and ensuring that every request from the React UI passes through token verification. JWTs are issued and validated using RSA256 signing with public and private keys. When a user logs out, the issued token is added to a “revoked list,” preventing it from being reused.
Key Takeaways & Future Steps
To expand on this project into a more fully-developed production style application, future steps could include:
- Conversation storage per user, allow user to view past chats (also clear stale chats)
- More robust multi-lingual support
- Currently, user must prompt for a certain language first
- Greater hotel and hotel-room data variety to test the effectiveness of vector similarity search
- More robust data validation when submitting a booking request
- Integration with a banking/payment service
Links
View the project’s installation, setup, and run steps.
View the Project Presentation.
View the GitHub Repository.