$ CoTeacher AI – Course RAG Assistant

$ cat project.md

A full-stack platform where instructors upload course materials and students chat with a course-specific AI.

// overview

CoTeacher AI is an intelligent course assistant platform that leverages Retrieval-Augmented Generation (RAG) to provide students with personalized, course-specific AI tutoring. The platform enables instructors to upload various course materials (PDFs, DOCX, PPTX) which are then processed, chunked, and embedded using OpenAI's API. Students can interact with a ChatGPT-style interface that answers questions based on the uploaded course content, ensuring accurate and contextually relevant responses.

// tech stack

frontend: [Next.js, TypeScript, Tailwind CSS, React]

backend: [Node.js, Next.js API Routes, OpenAI API]

database: [PostgreSQL, Supabase, Vector Embeddings]

infrastructure: [Vercel, Supabase Storage]

// features

Multi-role authentication system (Owner/TA/Student) with role-based access control
Document upload and processing pipeline supporting PDF, DOCX, and PPTX formats
Server-side RAG implementation with vector embeddings and semantic search
Per-course vector database isolation for accurate, course-specific responses
ChatGPT-style UI with conversation history and integrity guardrails
Real-time chat interface with streaming responses
File management system with Supabase Storage integration

// challenges

Implementing efficient chunking strategies for different document types while preserving context
Designing a scalable vector search system that maintains per-course data isolation
Balancing response accuracy with response time in the RAG pipeline
Creating an intuitive UI that handles complex document structures and long conversations

// tags

Next.jsTypeScriptTailwind CSSSupabasePostgresRAGOpenAI APIVercelNode.js

Tip: Highlight tradeoffs, constraints, and measurable outcomes.