LectureMind: AI-Powered PDF Study Assistant

---

About This Project

This repository hosts the source code for LectureMind. LectureMind is a mobile app that allows students to upload lecture slides or notes and interact with them through a built-in chat interface, all within a single, side-by-side workspace. Instead of manually searching through lengthy materials, students can open a PDF and ask questions directly within the app.

📸 Screenshots

Home page

Key Features

• PDF Upload: Import lecture notes and slide decks directly from the mobile app.
• Automatic Text Extraction: Parses PDF contents on upload using PyMuPDF.
• Vector Search: Stores document chunks with embeddings in PostgreSQL using pgvector.
• AI Q&A: Answers study questions by retrieving the most relevant chunks from the uploaded lecture.
• Split Workspace View: Displays the PDF alongside the chat assistant on larger screens.
• Mobile-First Experience: Built with Expo and React Native for an accessible cross-platform workflow.

Tech Stack

Frontend

• Framework: React Native with Expo Router
• Language: TypeScript
• PDF Viewing: react-native-pdf
• File Upload: expo-document-picker

Backend

• Framework: FastAPI
• Language: Python 3.11
• Database: PostgreSQL
• Vector Search: pgvector
• ORM: SQLAlchemy
• PDF Processing: PyMuPDF
• AI Models:
  • Embeddings: text-embedding-3-small
  • Chat: gpt-4o-mini

Getting Started

Prerequisites

• Node.js and npm
• Python 3.11+
• Docker Desktop
• An OpenAI API key

1. Backend Setup

cd backend
cp .env.example .env

Add the following values to .env:

OPENAI_API_KEY=your_openai_api_key
DATABASE_URL=postgresql://postgres:postgres@db:5432/lecturemind

Then start the backend services:

docker-compose up --build

The API will be available at http://localhost:8000.

2. Frontend Setup

From the project root:

npm install
npm start

You can then run the app with Expo on:

• iOS simulator
• Android emulator
• Expo Go
• Web

Important Notes

• The current frontend code sends requests to http://localhost:8000, so device testing works best on a simulator running on the same machine as the backend.
• The backend currently expects PDF uploads and OpenAI access to be available before chat can work.
• There is no committed .env.example file in the repository yet, so you may need to create .env manually inside backend/.