NaviBlu Travel Assistant

Project Overview

NaviBlu is an AI-powered travel assistant designed to make trip planning effortless by allowing the user to ask natural language questions. Built as a group project for UNC Charlotte's Software System Design course, NaviBlu combines real-time flight and hotel data with Meta's Llama 3.3 70B language model to provide personalized travel recommendations.

The chatbot handles everything from finding the cheapest flights across 400+ airlines to recommending hotels from a database of 150,000+ properties worldwide. Users can simply chat naturally about their travel plans without filling out complex forms.

Key Features

• Flight Search - Real-time flight pricing from 400+ airlines using the fast-flights API
• Hotel Search - 150,000+ hotels across 190 countries using the Amadeus API
• Destination Info - Tourist attractions and activity recommendations
• AI-Powered - Built on Meta's Llama 3.3 70B model with inference via Groq API

Technical Implementation

Architecture

NaviBlu uses an agent-based architecture where specialized agents handle different aspects of travel planning:

• Flight Agent - Queries the fast-flights API for real-time pricing
• Hotel Agent - Searches Amadeus API for accommodation options
• Location Agent - Provides destination information and recommendations
• General Agent - Handles conversational queries and context management

Technology Stack

The project combines modern web technologies with powerful AI capabilities:

• Frontend - Streamlit for the chat interface, HTML/CSS/JS for the home page on GitHub Pages.
• Backend - Python with agent-based logic running on Hugging Face Spaces.
• LLM - Meta Llama 3.3 70B via Groq API for fast inference

Team Collaboration

Developed as part of ITCS 6112 (Software System Design and Implementation) at UNC Charlotte with a team of 4 students. I focused on the AI agent architecture and streamlit interface.

Technologies Used

Project Links

View on GitHub →

Try on Hugging Face →