CSIS Helpdesk Chatbot

A RAG-powered AI chatbot for university helpdesk queries built with a microservices architecture.

Architecture

Backend Services

Service	Port	Description
Knowledge Base	8000	Vector search with Qdrant, document ingestion
Chatbot	8001	RAG chat with GPT-3.5-turbo, escalation logic
Notification	8003	Escalation handling, ticket creation

Infrastructure

• Qdrant (6333) - Vector database
• RabbitMQ (5672, 15672) - Message queue for escalations
• MySQL - Conversation storage
• NGINX (80) - Load balancer

Frontend

React

Quick Start

Prerequisites

• Docker and Docker Compose
• Node.js 18+
• OpenAI API key

Environment Setup

Create backend/.env:

OPENAI_API_KEY=your-key-here

Running with Docker

cd backend
docker-compose up -d

Frontend runs at http://localhost:3000

Running Frontend on its own

cd frontend
npm install
npm run dev

API Endpoints

Chatbot Service (8001)

• POST /api/chat - Send message, get AI response
• GET /api/conversations - List conversations
• GET /api/conversations/{session_id} - Get conversation messages
• DELETE /api/conversations/{session_id} - Delete conversation

Knowledge Base Service (8000)

• POST /api/knowledge/search - Vector search
• POST /api/knowledge/ingest - Add document
• POST /api/knowledge/upload - Upload file (PDF/DOCX/TXT/JSON)
• GET /api/knowledge/documents - List documents
• DELETE /api/knowledge/documents/{id} - Delete document

Notification Service (8003)

• GET /api/escalations - List escalations
• GET /api/escalations/stats - Escalation statistics
• PATCH /api/escalations/{id}/status - Update status

Features

• RAG-powered responses using OpenAI embeddings and GPT-3.5
• Automatic document chunking for better retrieval
• Query escalation based on confidence scores and keywords
• Conversation history per user
• Admin dashboard for managing knowledge base and escalations
• Voice input support

Project Structure

backend/
  chatbot/          # Main chat service
  knowledge_base/   # Vector DB and document processing
  auth/             # Authentication service
  notification_service/  # Escalation handling
  database/         # MySQL schema
frontend/
  src/
    components/     # React components
    services/       # API client
    utils/          # Helper functions
sample_data/        # Sample knowledge base documents

Configuration

Escalation Thresholds

In backend/chatbot/main.py:

• MIN_CONFIDENCE_SCORE = 0.3 - Below this triggers escalation
• ESCALATION_KEYWORDS - Words that trigger escalation

Chunking

In backend/knowledge_base/main.py:

• CHUNK_SIZE = 1500 - Characters per chunk
• CHUNK_OVERLAP = 200 - Overlap between chunks