TurboAz AI Assistant

An AI-powered RAG chatbot for exploring used car listings from Turbo.az using natural language.

---

Overview

TurboAz AI Assistant is a Retrieval-Augmented Generation (RAG) application that enables users to search and analyze used car listings through natural language.

Instead of manually filtering hundreds of advertisements, users can simply ask questions such as:

• Which SUVs can I buy for under 20,000 AZN?
• Show me automatic Toyota Corollas after 2018.
• Is this BMW overpriced?
• Recommend the best sedan under my budget.

The assistant combines structured filtering, vector search and Llama 3.3 to generate accurate answers based on real vehicle listings.

---

Dataset

The project uses approximately 1,800 real vehicle listings collected from Turbo.az.

Each listing contains structured information including:

• Brand
• Model
• Price
• Year
• Mileage
• Engine
• Transmission
• Fuel Type
• Color
• City
• Vehicle Condition
• Listing URL

---

Architecture

Turbo.az

↓

Web Scraping (BeautifulSoup)

↓

Raw CSV

↓

Data Cleaning (Pandas)

↓

ChromaDB Vector Database

↓

Streamlit Interface

↓

User Question

↓

Filter Extraction (LLM)

↓

Pandas Filtering

↓

Intent Detection

↓

Vector Search (if required)

↓

Llama 3.3 (Groq)

↓

Answer

---

Features

• Web scraping from Turbo.az
• Data preprocessing and cleaning
• Vector database with ChromaDB
• Natural language vehicle search
• Intelligent recommendation system
• Price analysis
• Automatic filter extraction
• Intent detection
• Azerbaijani language responses
• Interactive Streamlit interface

---

Tech Stack

Programming

• Python

Libraries

• Pandas
• BeautifulSoup
• Requests
• ChromaDB
• Streamlit
• Groq API

AI

• Llama 3.3 70B
• Retrieval-Augmented Generation (RAG)

---

Installation

Clone the repository

git clone https://github.com/alkyramazann/TurboAz-AI-Assistant.git

Install dependencies

pip install -r requirements.txt

Create a .env file

GROQ_API_KEY=YOUR_API_KEY

Run the application

streamlit run app.py

---

Project Structure

TurboAz-AI-Assistant/

src/
chatbot.py
scraper.py
vectordb.py

data/

database/

assets/

screenshots/

README.md

requirements.txt

---

Screenshots

Landing Page

---

AI Chat Interface

---

Smart Recommendations

Future Improvements

• Hybrid Search (BM25 + Vector Search)
• Vehicle image understanding
• Voice assistant
• Multi-language support
• User authentication
• Favorites system
• Fine-tuned recommendation model

---

Author

Ramazan Allahverdizada

LinkedIn

www.linkedin.com/in/ramazan-allahverdizada-8b541431a

GitHub

https://github.com/alkyramazann

This project was developed for educational and portfolio purposes. All listing data belongs to Turbo.az.