🤖 AI Customer Support Bot

A smart, LLM-powered customer support system designed to answer FAQs, maintain conversation context, escalate unresolved queries, and provide a simple chat interface. This project was created for the Placement Drive Assignment.

🚀 Features

• AI-Powered Responses: Utilizes Groq LLM (openai/gpt-oss-120b) for intelligent, contextual replies when FAQs are insufficient.
• FAQ Matching (Semantic Search): Retrieves the best matching FAQ from MongoDB, falling back to AI when similarity is low using embedding-based similarity and word-overlap fallback.
• Contextual Memory: Stores user messages, bot messages, and optional summaries, allowing the bot to remember past messages during the session.
• Escalation System: Automatically escalates queries that the bot cannot answer after repeated attempts or when FAQ similarity is too low, storing them in the DB for review.
• Clean REST API: Well-structured backend with controllers, services, models, validators, error handling, and logging.
• Simple React Chat UI: A clean chat interface built with React and Axios.

🏗️ Tech Stack

Backend

• Node.js
• Express.js
• MongoDB + Mongoose
• Groq SDK (LLM)
• Zod Validation
• Dotenv
• CORS

Frontend

• React
• Axios
• UUID

📁 Project Structure

backend/
├── src/
│   ├── controllers/
│   ├── models/
│   ├── services/
│   ├── routes/
│   ├── validators/
│   ├── utils/
│   └── app.js
├── seed/
│   └── insertFaqs.js
└── package.json

client/
├── src/
│   ├── components/
│   ├── api/
│   └── App.jsx
└── package.json

⚙️ Installation & Setup

1. Clone the repository

   git clone https://github.com/jadhav045/AI-Customer-Support-Bot..git
   cd backend

2. Install backend dependencies

   npm install

3. Add environment variables Create .env inside the backend:

   PORT=5000
   MONGODB_URI=mongodb://localhost:27017/ai-customer-support-bot
   GROQ_API_KEY=your_groq_api_key_here
   GROQ_MODEL=openai/gpt-oss-120b
   GROQ_EMBEDDING_MODEL=llama-embed-english-v1
   FAQ_MATCH_THRESHOLD=0.6
   FAQ_ESCALATION_THRESHOLD=0.2
   

4. Seed FAQs into MongoDB

   node seed/insertFaqs.js

5. Run backend

   npm run dev

6. Run frontend

   cd ../client
   npm install
   npm run dev

📡 API Endpoints

POST /api/query

Send a user message.

Request:

{
  "sessionId": "abc123",
  "query": "How do I reset my password?"
}

Response:

{
  "success": true,
  "botResponse": "Go to Account Settings → Security → Reset Password...",
  "matchScore": 0.92,
  "usedFAQ": true,
  "escalated": false
}

🖼️ Screenshots

Demo Video

The demo video file referenced previously (./assets/demo.mp4) is not present in this branch and will return a 404 on GitHub if missing.

You can also host the demo externally. Example (Google Drive public/shareable link):

• Watch demo (Google Drive): https://drive.google.com/file/d/11I6A-HxVRRu-gg2y8MSAuBIXwI1HkP1Y/view?usp=sharing

Note: Ensure the Google Drive file is set to "Anyone with the link can view" so collaborators can open it. Other options:

• Add demo.mp4 to this repository (small files)
• Upload to YouTube/Vimeo (recommended for large files)
• Convert to GIF for inline preview

📌 Author

Developed as part of the Placement Drive Assignment, including an AI-powered customer support pipeline with modern backend and frontend technologies.