AI Course Support System

An AI-powered educational assistant that processes course materials and provides intelligent tutoring through a retrieval-augmented generation (RAG) system. The system automatically processes PDF documents, creates searchable indexes, and provides contextual answers with document citations.

Architecture Overview

This system consists of three main components:

1. Document Processing Pipeline (Azure Functions)

   • PDF chunking and text extraction
   • Vector embedding generation
   • Azure AI Search indexing

2. AI Chat Backend (FastAPI)

   • RAG implementation with Azure OpenAI
   • Semantic search across documents
   • Educational prompt engineering

3. Frontend Interface (Next.js)

   • Chat interface for student interactions
   • Document reference display
   • Course material navigation

🚀 Quick Start

Prerequisites

• Node.js 18.x or later
• Azure subscription with the following services:
  • Azure Functions
  • Azure Blob Storage
  • Azure AI Search
  • Azure OpenAI
• Python 3.9+ (for FastAPI backend)

1. Azure Services Setup

Required Azure Resources:

# Create resource group
az group create --name ai-course-support --location eastus

# Create storage account
az storage account create --name <storage-name> --resource-group ai-course-support --location eastus --sku Standard_LRS

# Create Azure AI Search service
az search service create --name <search-name> --resource-group ai-course-support --location eastus --sku Basic

# Create Azure OpenAI service
az cognitiveservices account create --name <openai-name> --resource-group ai-course-support --location eastus --kind OpenAI --sku S0

Required Blob Storage Containers:

• students-tools (for original PDF uploads)
• students-tools-chunked (for processed document chunks)

2. Backend Setup (FastAPI)

# Navigate to API directory
cd src/api

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Create .env file with your Azure credentials
cp .env.example .env
# Edit .env with your Azure service credentials

# Start the FastAPI server
python main.py

3. Azure Functions Setup

# Install Azure Functions Core Tools
npm install -g azure-functions-core-tools@4 --unsafe-perm true

# Install dependencies
npm install

# Copy and configure local settings
cp local.settings.example.json local.settings.json
# Edit local.settings.json with your Azure credentials

# Start functions locally
npm start

4. Frontend Setup (Next.js)

# Navigate to frontend directory
cd src/frontend

# Install dependencies
npm install

# Start development server
npm run dev

📋 Required Environment Variables

Azure Functions (local.settings.json)

{
  "IsEncrypted": false,
  "Values": {
    "AzureWebJobsStorage": "<storage-connection-string>",
    "FUNCTIONS_WORKER_RUNTIME": "node",
    "AZURE_SEARCH_ENDPOINT": "https://<search-name>.search.windows.net",
    "AZURE_SEARCH_API_KEY": "<search-admin-key>",
    "AZURE_SEARCH_INDEX_NAME": "document-chunks",
    "AZURE_OPENAI_ENDPOINT": "https://<openai-name>.openai.azure.com",
    "AZURE_OPENAI_API_KEY": "<openai-api-key>",
    "AZURE_OPENAI_EMBEDDING_DEPLOYMENT": "text-embedding-ada-002",
    "PAGES_PER_CHUNK": "10"
  }
}

FastAPI Backend (src/api/.env)

AZURE_SEARCH_ENDPOINT=https://<search-name>.search.windows.net
AZURE_SEARCH_API_KEY=<search-admin-key>
AZURE_SEARCH_INDEX_NAME=document-chunks
AZURE_OPENAI_ENDPOINT=https://<openai-name>.openai.azure.com
AZURE_OPENAI_API_KEY=<openai-api-key>
AZURE_OPENAI_COMPLETION_DEPLOYMENT=gpt-4o-mini
ALLOWED_ORIGINS=http://localhost:3000

Frontend (src/frontend/.env.local)

NEXT_PUBLIC_API_URL=http://localhost:8000

🔧 Key Components

1. PDF Chunking Function (src/functions/pdfchunker/PDFChunker.ts)

• Trigger: Blob storage events when PDFs are uploaded
• Process: Splits PDFs into overlapping chunks, extracts text
• Output: Chunked PDFs and metadata in separate container

2. Document Indexing Function (src/functions/indexing/PDFIndexer.ts)

• Trigger: Blob storage events for chunked documents
• Process: Generates embeddings, creates search index
• Dependencies: Embeddings Generator, Search Index Manager

3. Chat API (src/api/main.py)

• Endpoints: /api/ChatCompletion, /health
• Features: RAG with Azure AI Search, educational prompt engineering
• Response: AI answers with document citations

4. Frontend Interface (src/frontend/)

• Components: Chat interface, document panel, navigation
• Features: Real-time chat, document references, course selection

📁 Project Structure

├── src/
│   ├── functions/                    # Azure Functions
│   │   ├── pdfchunker/              # PDF processing
│   │   │   └── PDFChunker.ts        # Main chunking function
│   │   ├── indexing/                # Document indexing
│   │   │   └── PDFIndexer.ts        # Indexing function
│   │   └── services/                # Shared services
│   │       ├── indexing/            # Search and embedding services
│   │       └── utils/               # Utility functions
│   ├── api/                         # FastAPI backend
│   │   ├── main.py                  # Main API application
│   │   ├── requirements.txt         # Python dependencies
│   │   └── .env                     # Environment variables
│   ├── frontend/                    # Next.js frontend
│   │   ├── app/                     # App router pages
│   │   ├── components/              # React components
│   │   ├── lib/                     # Utilities and services
│   │   └── package.json             # Frontend dependencies
│   └── models/                      # Shared TypeScript models
├── host.json                        # Function app configuration
├── local.settings.json              # Local development settings
├── package.json                     # Function dependencies
└── README.md                        # This file

🔄 Data Flow

1. Document Upload: PDFs uploaded to students-tools container
2. Chunking: Azure Function processes PDFs into chunks
3. Indexing: Another Function creates embeddings and search index
4. Query: User asks question through web interface
5. Search: System performs hybrid search on indexed content
6. Generate: Azure OpenAI generates response using retrieved documents
7. Display: Response shown with document citations

🚀 Deployment

Azure Functions Deployment

# Build and deploy functions
npm run build
func azure functionapp publish <your-function-app-name>

FastAPI Deployment (Azure Container Instances)

# Build Docker image
docker build -t ai-course-api src/api/

# Deploy to Azure Container Instances
az container create --resource-group ai-course-support --name ai-course-api --image ai-course-api --cpu 1 --memory 2 --port 8000

Frontend Deployment (Azure Static Web Apps)

# Build frontend
cd src/frontend
npm run build

# Deploy to Azure Static Web Apps
az staticwebapp create --name ai-course-frontend --resource-group ai-course-support --source . --branch main --app-location "src/frontend" --output-location ".next"

🔍 Usage

1. Upload Documents: Place PDF course materials in the students-tools blob container
2. Wait for Processing: Functions automatically chunk and index documents
3. Start Chatting: Use the web interface to ask questions about course content
4. Review References: Check document citations provided with each response

🛠️ Customization

Adding New Document Types

• Modify PDFChunker.ts to handle additional file formats
• Update text extraction logic in the chunking function

Adjusting Chunk Size

• Change PAGES_PER_CHUNK environment variable
• Rebuild and redeploy functions

Modifying AI Behavior

• Edit the system prompt in src/api/main.py
• Adjust OpenAI parameters (temperature, max_tokens)

📊 Monitoring

• Function Logs: View in Azure Portal or Application Insights
• API Health: Check /health endpoint
• Search Performance: Monitor Azure AI Search metrics
• OpenAI Usage: Track token consumption in Azure OpenAI

🔧 Troubleshooting

Common Issues

1. Functions not triggering: Check blob storage connection strings
2. Search returning no results: Verify index exists and has data
3. OpenAI errors: Check API key and deployment names
4. Frontend not connecting: Verify API URL in environment variables

Debug Commands

# Check function logs
func start --verbose

# Test API health
curl http://localhost:8000/health

# Check search index
curl -H "api-key: <search-key>" "https://<search-name>.search.windows.net/indexes/document-chunks?api-version=2023-11-01"

📝 Support

For issues and questions:

1. Check the troubleshooting section above
2. Review Azure service logs
3. Verify all environment variables are set correctly
4. Ensure all required Azure services are running

🔄 Version History

• v1.0.0: Initial release with PDF processing, RAG chat, and web interface
• Current: Enhanced error handling, improved document citations, markdown support