RAG Agent Chatbot

This project is a Retrieval-Augmented Generation (RAG) chatbot built using Node.js, Express, Socket.IO, MongoDB, Pinecone, and OpenAI's GPT-4. The chatbot leverages retrieval-augmented generation to answer user queries by retrieving relevant documents from a knowledge base and generating responses grounded in that knowledge.

---

Table of Contents

• Project Overview
• Features
• Architecture
• Requirements
• Setup
• Environment Variables
• Running the Application
• Troubleshooting
• License

---

Project Overview

The RAG Agent Chatbot is designed for customer support, helping users by providing responses based on information stored in a knowledge base. It can also escalate to human support when necessary.

Features

• Real-time Chat: Users interact with the chatbot in real time.
• Retrieval-Augmented Generation (RAG): Uses embeddings and a vector database to retrieve relevant documents for queries.
• Socket.IO Support: Enables seamless communication between the frontend and backend.
• Support Ticketing: Creates support tickets for unresolved queries.
• Feedback Collection: Allows users to provide feedback on chatbot responses.

---

Architecture

• Frontend: Built with React for a user-friendly chat interface.
• Backend: Developed with Node.js and Express for API endpoints, with Socket.IO for real-time communication.
• Database: MongoDB stores knowledge base entries, user interactions, and tickets.
• Vector Database: Pinecone is used for storing and querying vectorized embeddings of knowledge base documents.
• OpenAI: GPT-4 generates responses based on the retrieved documents.

---

Requirements

• Node.js: v16 or higher (v22 is not a valid version)
• MongoDB Atlas: A MongoDB cluster to store knowledge base and user interactions.
• Pinecone: A vector database account for embedding storage and querying.
• OpenAI API Key: Access to OpenAI’s GPT-4 model.

---

Setup

1. Clone the Repository

   git clone https://github.com/your-username/rag-agent-chatbot.git
   cd rag-agent-chatbot

2. Install Dependencies

   Navigate to the backend directory and install the required dependencies:

   cd backend
   npm install

3. Set Up Environment Variables

   Create a .env file in the backend directory with the following variables:

   PORT=5000
   MONGO_URI=mongodb+srv://<username>:<password>@<cluster>.mongodb.net/?retryWrites=true&w=majority
   OPENAI_API_KEY=your_openai_api_key
   PINECONE_API_KEY=your_pinecone_api_key
   PINECONE_ENVIRONMENT=your_pinecone_environment
   PINECONE_INDEX=your_pinecone_index_name

   Replace the placeholders with your actual credentials:

   • MONGO_URI: MongoDB connection string with username, password, and database name.
   • OPENAI_API_KEY: Your OpenAI API key for accessing GPT-4.
   • PINECONE_API_KEY: Your Pinecone API key.
   • PINECONE_ENVIRONMENT: Pinecone environment (e.g., us-west1-gcp).
   • PINECONE_INDEX: The name of your Pinecone index.

---

Running the Application

1. Start the Backend Server

   Navigate to the backend directory and start the server:

   node server.js

2. Run the Frontend (if applicable)

   If you have a frontend directory with a React app, navigate to it and start the development server:

   cd frontend
   npm install
   npm start

3. Access the Application

   Open your browser and go to http://localhost:3000 to interact with the chatbot.

---

Environment Variables

Variable	Description
PORT	Port for the backend server (default: 5000)
MONGO_URI	MongoDB connection string
OPENAI_API_KEY	API key for OpenAI GPT-4
PINECONE_API_KEY	API key for Pinecone
PINECONE_ENVIRONMENT	Pinecone environment (e.g., us-west1-gcp)
PINECONE_INDEX	Name of your Pinecone index

---

Troubleshooting

• CORS Issues

  If you encounter a Cross-Origin Request Blocked error, ensure that CORS is configured correctly in server.js:

  app.use(cors({ origin: 'http://localhost:3000', methods: ['GET', 'POST'], credentials: true }));

• MongoDB Authentication Failed

  • Verify your MongoDB URI, username, and password.
  • URL-encode special characters in the password if necessary.
  • Ensure the MongoDB user has access to the specified database.

• Pinecone searchSimilarDocs Error

  If you see a searchSimilarDocs is not a function error, make sure to use the .query method on the Pinecone client:

  const queryResponse = await pineconeIndex.query(queryRequest);

• OpenAI Rate Limit Error

  If you encounter a RateLimitError or 429 You exceeded your current quota error:

  • Check your OpenAI account quota and billing.
  • Implement error handling to manage quota limits gracefully.

• Check Node.js Version

  Ensure you’re using Node.js v16 or higher. Verify your Node version by running:

  node -v

• Logging for Debugging

  Use console.log statements to trace issues, especially in server.js, agent.js, and retrieveDocuments.js.

---

License

This project is licensed under the MIT License. See the LICENSE file for more information.

---