TA AI - Q&A Assistant for University Courses

A cost-effective, privacy-first Q&A assistant that helps university courses provide 24/7 student support while maintaining the professor's tone and citing exact sources.

Features

• Course Material Ingestion: Process slides (PDF/PPTX) and lecture transcripts
• Intelligent Q&A: Answer student questions using professor's tone with exact citations
• Professor Oversight: Review logs, flag incorrect answers, maintain quality control
• Privacy & Security: All data stays within Azure region using Private Endpoints
• Cost Optimized: Serverless architecture targeting <£80/month for 200 students

Architecture

┌──────────┐   upload   ┌──────────────┐     chunk/     ┌────────────┐
│Professor │──────────►│Ingestion Fn  │──embed + store►│Postgres +  │
│Portal    │           │(Azure Fn)    │                │pgvector DB │
└──────────┘           └──────────────┘                └────────────┘
                                         query
     SSO (OIDC)                        ┌───────────┐   top-k vectors
Student  ───► Next.js chat ► API Fn ──►│GPT-4 Turbo│◄──(Azure OpenAI)  
browser     (Azure Fn, private VNet)   └───────────┘      (UK South PE)
                                   ↑    citations
                      log Q&A in Postgres; expose to prof dashboard

Technology Stack

• Frontend: Next.js 14 with TypeScript, Tailwind CSS
• Backend: Python 3.11 with FastAPI on Azure Functions
• Database: PostgreSQL 15 with pgvector extension
• LLM: GPT-4 Turbo via Azure OpenAI Service
• Embeddings: text-embedding-3-small
• Auth: Microsoft Entra ID (OIDC)
• Infrastructure: Azure (Functions, Static Web Apps, VNet, Private Endpoints)
• IaC: Terraform

Project Structure

├── frontend/          # Next.js application
├── backend/           # Python Azure Functions
├── infrastructure/    # Terraform modules
├── docs/             # Documentation and planning
└── tests/            # Test suites

Development Timeline

• Week 1: Foundation & Infrastructure
• Week 2: Data Ingestion Pipeline
• Week 3: Query & Retrieval System
• Week 4: Frontend Development
• Week 5: Professor Dashboard
• Week 6: Production Readiness

Getting Started

Prerequisites

• Node.js 18+
• Python 3.11+
• Azure CLI
• Terraform
• PostgreSQL with pgvector

Local Development

# Clone repository
git clone <repo-url>
cd ta-ai

# Install frontend dependencies
cd frontend
npm install

# Install backend dependencies
cd ../backend
pip install -r requirements.txt

# Set up infrastructure
cd ../infrastructure
terraform init
terraform plan

Cost Estimation

Component	Monthly Cost
GPT-4 Turbo tokens	$32
Embeddings	$1
Azure Postgres	$25
Azure Functions	$2
Static Web App	$0
Misc	$5
Total	$65

Security

• All resources within Azure VNet
• Private Endpoints for OpenAI API
• Encrypted storage and backups
• 30-day log retention
• SSO authentication via Microsoft Entra

License

MIT License - see LICENSE file for details.

Support

For questions or issues, please see the documentation in the docs/ directory or contact the development team.