NoteWise - AI-Powered Study Assistant

A comprehensive AI-powered study platform that transforms educational content into interactive learning experiences. Features intelligent document ingestion, RAG-based Q&A chat, AI-generated teacher-student audio dialogues, scene-by-scene video summaries, and real-time progress tracking with streaming updates.

OVERVIEW

NoteWise is an advanced full-stack AI-powered study assistant designed to revolutionize how students interact with educational content. The platform enables users to upload PDF documents and YouTube video transcripts, which are automatically processed, embedded using Google's Gemini embedding models, and stored in a PostgreSQL database with pgvector extension for efficient semantic search. The system consists of multiple interconnected modules: a modern Next.js 16 web application with responsive UI components built using Radix UI and Tailwind CSS, a FastAPI backend with a multi-agent architecture powered by LangChain and LangGraph, and a PostgreSQL database with pgvector for vector similarity search. Key features include smart document ingestion with automatic OCR support for scanned PDFs via OCR.space API, YouTube transcript extraction and processing, RAG-based Q&A chat interface with streaming responses and markdown rendering, AI-generated teacher-student audio dialogues using LangGraph state machines with automatic text-to-speech conversion via Google TTS, scene-by-scene video summaries with visual presentation-style breakdowns and exam tips, real-time progress tracking with Server-Sent Events (SSE) for long-running operations, conversation persistence allowing users to continue previous chats, and comprehensive markdown support throughout the application for rich text rendering. The application uses a sophisticated multi-agent system where specialized agents handle different tasks: a retriever agent for semantic search across embedded content, a QA agent for answering questions grounded in uploaded materials, a dialogue orchestrator using LangGraph to manage teacher-student conversation flows, and a summary agent for generating structured video-style summaries. All agents are built with async-safe operations and lazy initialization for optimal performance. Built with production-grade reliability in mind, the application implements connection pooling for database operations, streaming responses for real-time user feedback, comprehensive error handling, and is optimized for deployment on platforms like Render. The frontend provides an intuitive, modern interface with smooth animations, responsive design, and excellent user experience across all features.