🚗 VariantWise — Your Personalized Car Consultant

VariantWise is an AI-powered car recommendation platform that helps users discover the perfect car variant tailored to their needs. Leveraging advanced machine learning and real-world insights, it simplifies the complex car selection process into a few intuitive steps.

VariantWise enhances recommendations by combining technical specifications with real-life experience metrics such as comfort, material quality, throttle response, and brake performance. These features are scored on a scale of 1 to 5, manually curated in the initial project phase, with plans to automate or validate them in future iterations. Behind the scenes, VariantWise builds a custom preference profile for each user by analyzing their inputs through semantic embeddings and rule-based scoring mechanisms. This personalized vector is then matched against every variant in the database to shortlist the top 5 best-matching variants. User interaction is handled through an intelligent chatbot powered by Retrieval-Augmented Generation (RAG). The chatbot gathers user preferences and then retrieves relevant insights and variant data from a curated knowledge base and provides context-aware recommendations.

VariantWise simplifies what is traditionally a complex and overwhelming process into a smart, conversational, and deeply personalized experience — helping every user find not just the right car, but the right variant.

---

📁 Project Structure

This monorepo is divided into five main components:

Folder	Description
backend/	Node.js (Express) backend for authentication, sessions, and proxying.
data/	Datastore for all the scraped raw specs, the final dataset and the reviews
frontend/	Next.js frontend providing a sleek and interactive user interface.
model/	Python (Flask) service for AI/ML-powered car recommendations and Q&A.
scraper/	Scraping the specs and reviews for all cars in car_models.json

---

⚙️ Prerequisites

Make sure the following are installed/configured:

• Node.js (v16 or later)
• npm or yarn
• Python (v3.8+)
• pip
• MySQL server
• AWS account with access to Amazon Bedrock (Mistral model) and appropriate IAM permissions

---

🚀 Setup Instructions

1. Clone the Repository

git clone https://github.com/Prakhar-4/VariantWise
cd VariantWise

---

2. Scraper Setup (scraper/)

cd scraper

To collect technical specs of all variants of the car models listed in car_models.json

python main.py

To collect expert reviews of all car models listed in car_models.json

python reviews.py

To merge all the technical specs scraped into final_dataset.csv

python merge_data.py

Make sure the car_models.json file is present in the /scraper folder — it's used as the master list of car models to scrape.

Manual Scoring of Experience Features

After running merge_specs.py, you’ll need to manually score each car variant for some user experience features. These features go beyond the technical specifications, capturing the real-world feel of the car.

The following (or some similar) user experience features need to be scored on a scale of 1 to 5:

• Front seat comfort
• Rear seat comfort
• Under-thigh support
• Rear seat recline angle
• Gear lever feel
• Bump absorption capability
• Cabin noise insulation
• Suspension quality
• Body roll control
• Small item storage practicality
• Steering feedback
• Throttle response
• Clutch feel
• Brake responsiveness
• Material quality
• Instrument cluster readability
• Infotainment system responsiveness
• Sound system quality

These scores power the experience metrics used by the recommendation engine to deliver more human-centric suggestions.

---

2. Backend Setup (backend/)

cd ../backend
npm install
# or
yarn install

Create .env File

# MySQL Database
DB_HOST=your_mysql_host
DB_USERNAME=your_mysql_username
DB_PASSWORD=your_mysql_password
DB_NAME=your_mysql_database_name
DB_PORT=your_database_port

# Session Secret
Secret_Key=your_strong_session_secret

# Frontend URL
FRONTEND_URL=http://localhost:3000

# Google OAuth (Optional)
GOOGLE_CLIENT_ID=your_google_client_id
GOOGLE_CLIENT_SECRET=your_google_client_secret
GOOGLE_CALLBACK_URL=http://localhost:3001/api/auth/google/callback

# Server Settings
PORT=3001
NODE_ENV=development

Database Setup

• Ensure your MySQL server is running.
• Create the database as specified in .env.
• Create necessary tables like users (based on backend/models/user.js).
• Note: Migration scripts are not included.

Start Backend Server

npm start
# or
node index.js

Backend runs at: http://localhost:3001

---

3. Model Setup (model/)

cd ../model
pip install -r requirements.txt

Create .env File

AWS_ACCESS_KEY_ID=your_aws_access_key_id
AWS_SECRET_ACCESS_KEY=your_aws_secret_access_key
AWS_REGION=your_aws_region

If using an IAM user, make sure they have Amazon Bedrock permissions.

Required Files

• Ensure final_dataset.csv exists in data/.
• Ensure reviews/ directory exists in data/.

Start Flask Server

python app.py

Model server runs at: http://localhost:5000

---

4. Frontend Setup (frontend/)

cd ../frontend
npm install
# or
yarn install

Create .env.local

NEXT_PUBLIC_BACKEND_URL=http://localhost:3001
NEXT_PUBLIC_MODEL_URL=http://localhost:5000

Start Frontend Server

npm run dev
# or
yarn dev

Frontend runs at: http://localhost:3000

---

🌐 Access the Application

Once all servers are up and running, open your browser and visit:

http://localhost:3000

---

📬 Feedback & Contributions

We welcome feedback and contributions! Feel free to open issues or submit PRs to make VariantWise better.

---

🛡 License

This project is licensed under the MIT License.