---

E-commerce Chatbot

A robust, embeddable chatbot designed for eCommerce websites to enhance customer engagement and optimize customer acquisition. This chatbot uses a Retrieval-Augmented Generation (RAG) pipeline to answer user queries based on uploaded files such as product catalogs, FAQs, and other resources.

---

Features

1. File Upload for RAG Pipeline

• Upload files in formats such as PDF, TXT, or CSV.
• Automatically processes files and integrates them into the chatbot's knowledge base.

2. Customer Engagement

• Handles user queries based on the knowledge base.
• Allows follow-up questions with conversational context retained.
• Provides concise and accurate answers to user queries.

3. Embeddable Chat Interface

• User-friendly chat interface for integration into any eCommerce website.
• Easily embeddable using a React-based frontend.

---

Demo

Deployed Links

• Streamlit Admin Panel: Streamlit App
• React Chatbot Interface: Vercel Deployment
• Backend API Test: Render Deployment

Screenshots

---

Setup and Deployment

1. Clone the Repository

git clone https://github.com/karthik738/e-commerce-chatbot
cd e-commerce-chatbot

2. Backend Setup

Prerequisites:

• Python 3.8+
• Pinecone API Key

Install dependencies:

pip install -r requirements.txt

Configuration:

1. Pinecone Setup: Set your Pinecone API key and environment in the backend/main.py file or as environment variables:

   export PINECONE_API_KEY=<your_api_key>
   export PINECONE_ENVIRONMENT=<your_environment>

2. Run the Backend:

   cd backend
   uvicorn main:app --host 0.0.0.0 --port 8000

---

3. Streamlit Frontend (Admin Panel)

Prerequisites:

• Install Streamlit:

  pip install streamlit

Run the App:

cd streamlit_app_final
streamlit run app.py

---

4. Chatbot UI (React Interface)

Prerequisites:

• Node.js and npm/yarn installed.

Steps:

1. Navigate to the chatbot_ui directory:

   cd chatbot_ui

2. Install dependencies:

   npm install

3. Configure API Endpoint: Update the BASE_URL in api.js to point to your backend's deployment URL:

   const BASE_URL = "https://e-commerce-chatbot-1-ydcl.onrender.com";

4. Start the Development Server:

   npm start

5. Deploy the Chatbot UI: Use Vercel for deployment:

   npm install -g vercel
   vercel deploy

---

Architecture

Retrieval-Augmented Generation (RAG) Pipeline

1. File Upload: Files uploaded via Streamlit are processed and indexed into Pinecone.
2. Query Handling:
   • When a user asks a question, the chatbot retrieves relevant chunks from Pinecone.
   • Uses a language model (LLM) to generate answers based on the retrieved context.
3. Answer Delivery: The chatbot sends back concise and relevant answers.

---

Embedding Process

For the React-based Chatbot UI:

1. Deploy the chatbot to a service like Vercel.
2. Copy the provided embed script:

   <iframe
     src="https://e-commerce-chatbot-black.vercel.app/"
     width="400"
     height="600"
     style="border:none;"
   ></iframe>

3. Add the iframe to your eCommerce website's HTML.

---

File Structure

ecommerce-chatbot/
│
├── backend/                 # Backend API with FastAPI
│   ├── main.py              # Entry point for the backend
│   ├── services/            # File processing and embedding logic
│   └── requirements.txt     # Backend dependencies
│
├── streamlit_app_final/     # Streamlit admin interface
│   ├── app.py               # Entry point for Streamlit app
│   └── utils/               # Utility functions for file upload and query
│
├── chatbot_ui/              # React-based chatbot interface
│   ├── src/
│   ├── public/
│   └── package.json         # React dependencies
│
└── README.md                # Documentation

---

FAQs

1. Can the chatbot handle follow-up queries?
   Yes, it retains conversational history for seamless follow-ups.

2. How is the knowledge base updated?
   Upload updated files via the admin panel to refresh the knowledge base.

3. What file formats are supported?
   PDF, TXT, and CSV.

---

Contributing

We welcome contributions! Please create a pull request or open an issue for any improvements or bugs.

---

License

This project is licensed under the MIT License. See LICENSE for details.

---