🤖 Amazon Analytics Chatbot

RAG + SQL Powered Decision Support System for Amazon Seller Analytics

---

🚀 Business Impact

This project transforms raw Amazon Seller data into an AI-powered analytics assistant that enables fast, data-driven decision making.

• 💬 Ask questions in natural language (no SQL required)
• 📊 Retrieve insights directly from structured business data
• ⚡ Combine database queries with AI-generated explanations
• 🧠 Support operational and strategic decisions

👉 Goal: Make complex e-commerce analytics accessible through a chatbot interface

---

🎯 Use Case

This system is designed for:

• 🛒 E-commerce Managers → analyze sales, performance, and trends
• 📊 Data Analysts → speed up data exploration
• 💰 Business Teams → access insights without technical knowledge

Example Queries

• “Which products generated the highest revenue last month?”
• “Show sales trends by country”
• “Which categories are declining?”
• “Top performing ASINs in the last 30 days”

👉 The system converts these into SQL + contextual answers automatically.

---

🧠 Project Overview

This project implements a Retrieval-Augmented Generation (RAG) pipeline combined with SQL-based querying.

It integrates:

• Structured database queries (SQLite)
• Vector search (FAISS)
• Natural language understanding (LLM)
• Business data sources (Amazon Seller data)

👉 Result: Hybrid AI system combining symbolic + semantic retrieval

---

🏗️ System Architecture

🔹 Indexing Pipeline (Offline)

• Load CSV data sources
• Convert to structured tables (SQLite)
• Create document chunks
• Generate embeddings (Sentence Transformers)
• Store vectors in FAISS index

🔹 Retrieval Pipeline (Online)

• User query (natural language)
• Intent detection (SQL vs semantic search)
• Retrieve relevant context
• Execute SQL query (if needed)
• Combine results with LLM
• Generate final answer

---

⚙️ Features

• 💬 Natural language query interface
• 🧠 RAG-based context retrieval
• 🗄️ SQL database integration (SQLite)
• 🔎 Vector search with FAISS
• 📊 Structured + unstructured data support
• ⚡ Hybrid reasoning (SQL + LLM)
• 🧩 Modular pipeline (easy to extend)

---

🧬 Data Sources

The system uses multiple Amazon datasets:

• 📁 Orders (time series revenue data)
• 📊 Business Reports
• 📢 Advertising reports
• 🔍 Search term data

⚠️ Note: Data is partially masked / synthetic for portfolio purposes.

---

🧪 Core Technologies

• Python
• Streamlit
• SQLite
• FAISS
• Sentence Transformers
• Pandas / NumPy
• Local LLM (via Ollama or API)

---

📊 Example Workflow

1. User asks: → “Show revenue trend for last 3 months”

2. System:

   • detects analytical intent
   • generates SQL query
   • retrieves data
   • enriches with LLM explanation

3. Output:

   • structured data result
   • human-readable insight

---

🛠️ Installation

git clone https://github.com/Mst-KrgZ/amazon-analytics-chatbot.git
cd amazon-analytics-chatbot

pip install -r requirements.txt
streamlit run app.py

---

📦 Requirements

• Python 3.11
• Streamlit
• Pandas / NumPy
• FAISS
• Sentence Transformers
• SQLite
• LLM backend (Ollama / API)

---

👨‍💻 Author

Mesut Karagöz Data Scientist

🔗 GitHub: https://github.com/Mst-KrgZ 🔗 LinkedIn: https://www.linkedin.com/in/mesut-karagöz-181733260/

---

⚡ Final Note

This project goes beyond a chatbot — it is a data-driven decision support system powered by AI and structured analytics.