🌍 Smart Travel System — Backend (Developed by Ei Ei Khaing)

This repository contains my backend contributions to the Fanshawe College Capstone project Smart Travel System, an AI-powered travel planning assistant that integrates real-time Google data with generative AI for personalized travel experiences.

I developed two key backend modules — Smart Itinerary and Sentiment Insights — using FastAPI, Google Places API, and Gemini Flash 2.5.
These features enable the system to generate personalized travel itineraries and analyze user reviews dynamically.

---

🚀 Project Overview

🧭 Smart Itinerary

Generates personalized, day-by-day travel itineraries using Google Places API and Gemini LLM.

🔹 Core Workflow

1. Receives user input — destination city, trip duration, and preferences
2. Builds filtered Google Places queries for attractions and restaurants
3. Retrieves photo references and metadata (name, rating, price, types)
4. Sends the structured data + user filters to Gemini Flash 2.5
5. Gemini generates a natural-language itinerary (Morning / Afternoon / Evening) with descriptions tailored to preferences

🔹 Personalization Filters

Filter	Description	Example
Travel Type	Solo / Couple / Family / Friends	“Perfect for couples who enjoy peaceful escapes.”
Activity Theme	Adventure / Cultural / Relaxing / Shopping	“Start your morning with an exciting hike...”
Kid-Friendly	Boolean filter	Excludes bars, nightlife venues
Budget Level	0–4 (Free → Luxury)	Controls both price level and descriptive tone
Days	Number of days to plan	Adjusts day count dynamically

🔹 Output Example

{
  "city": "Toronto",
  "days": 2,
  "itinerary": [
    {
      "day": 1,
      "morning": "Explore the CN Tower and nearby cafes.",
      "afternoon": "Visit Ripley's Aquarium.",
      "evening": "Enjoy dinner at a lakeside restaurant."
    }
  ]
}

---

💬 Sentiment Insights

Analyzes Google Maps reviews for selected places to produce concise AI-generated summaries.

🔹 Process Flow

1. Retrieve reviews from Google Places Details API via Place ID
2. Aggregate multiple review texts
3. Use Gemini Flash 2.5 or a pretrained sentiment model (e.g. BERT/DistilBERT)
4. Generate a structured sentiment summary

🔹 Output Example

{
  "place_id": "ChIJmzrzi9Y0K4gR3Y1XxJ8zN7M",
  "overall_sentiment": "Positive",
  "summary": "Visitors love the atmosphere and friendly staff, though parking is limited."
}

🔹 Highlights

• Real-time, zero-dataset review analysis using Google Places data
• Supports multiple languages and summarization styles via Gemini
• Scalable for integration with itinerary recommendations

---

🗣️ Language Buddy (Planned 🚧)

Upcoming module for multilingual translation and conversational travel assistance using Gemini Flash 2.5 multilingual capabilities.

---

🧠 Architecture

SmartTravelSystem_Backend_EiEiKhaing/
│
├── app.py
├── routers/
│   ├── itinerary.py          # Smart Itinerary endpoint
│   ├── sentiment.py          # Sentiment Insights endpoint
│
├── services/
│   ├── gemini_service.py     # Handles Gemini LLM API interactions
│   ├── places_service.py     # Handles Google Places API requests
│   └── sentiment_service.py  # Performs review sentiment analysis
│
├── models/
│   ├── schemas.py            # Pydantic request/response models
│
├── docs/
│   ├── Smart_Itinerary_filters_guide.md
│
└── requirements.txt

Framework: FastAPI
LLM Integration: Gemini Flash 2.5
External APIs: Google Places API, Google Photos API, Google Maps Reviews API

---

⚙️ API Endpoints

Endpoint	Method	Description
/itinerary/generate	POST	Generate itinerary using city + filters
/sentiment/{place_id}	GET	Analyze sentiment for a Google Place
/chatbot/query	POST	(Coming soon) Multilingual chatbot

---

🧰 Tech Stack

• FastAPI for API framework
• Python 3.10+
• Gemini Flash 2.5 for AI-based text generation
• Google Places / Photos / Maps APIs
• SQLAlchemy & SQLite for local persistence
• Requests, Pydantic, Uvicorn

---

🔐 Environment Variables

Copy .env.example and fill in your own API keys:

GEMINI_API_KEY=your_gemini_api_key_here
GOOGLE_API_KEY=your_google_places_key_here

---

⚙️ How to Run

# 1️⃣ Install dependencies
pip install -r requirements.txt

# 2️⃣ Run FastAPI server
uvicorn app:app --reload

# 3️⃣ Test endpoints
http://127.0.0.1:8000/docs

---

📄 Status Summary

Module	Status	Description
Smart Itinerary	✅ Completed	Integrated Google Places & Gemini
Sentiment Insights	✅ Completed	Real-time review summarization
Language Buddy	🚧 Planned	Multilingual conversational support

---

📈 Future Enhancements

• Integrate user authentication for personalized itinerary history
• Add caching layer to reduce Google API cost
• Expand Gemini prompts for richer itinerary storytelling
• Deploy on Render or AWS Lambda for public demo

---

👩‍💻 Author

Ei Ei Khaing
Graduate Certificate in Artificial Intelligence & Machine Learning
Fanshawe College | London, Ontario, Canada

📧 LinkedIn
🔗 GitHub Portfolio

---

🏷️ Keywords

FastAPI Gemini API Google Places AI Backend Travel Itinerary Sentiment Analysis Generative AI Python