Abbilaash/SnapClass

AI-powered offline classroom assistant for under-resourced schools. Real-time speech-to-text, textbook parsing, quiz generation, and analytics—all on-device, no internet required. Empowers teachers with personalized insights and students with interactive learning via local WiFi access point. # SnapClass SnapClass is an advanced on-device edge AI solution designed for low-connectivity, high-density classroom environments. Powered by Snapdragon’s Hexagon NPU, it runs open-source large language models (LLMs), image captioning, and audio transcription entirely offline. SnapClass automates personalized learning by transcribing lectures, analyzing textbook content, and generating adaptive quizzes—bridging educational gaps in underserved regions where internet access and qualified educators are limited. ## App architecture ### Models - **Whisper-small (242M params)** via "openai/whisper-small" - **nougat-small (247M params)** via "facebook/nougat-small" - **blip-image-captioning-base** via "Salesforce/blip-image-captioning-base" running parallelly with whisper-small - **Phi-3.5-mini-instruct (3.82B params)** via "AnythingLLM" running locally via ONNX accelerated by Snapdragon's X Elite's NPU ## Features - 🖼Teacher dashboard with file upload and analytics view - Lecture and textbook PDF/audio upload - AI-based question and answer evaluation - Uses Both CPU and NPU for faster on-device processing - Identifies weak syllabus topics per student or group - Fully functional offline – no internet needed - Lightweight and fast inference using sentence embeddings ## Setup & Usage ### Step 1: Setup Local Hotspot (No Internet) We use MyPublicWifi to create a local area dead network 1. Download and Install MyPublicWifi. 2. Open the app, set: - Network Access = No Internet Sharing - Turn on hotspot 3. Note the IP address shown in the app. This IP will be used to access the server from other devices. ### Step 2:Install Python Dependencies 1. Clone this repository ``` git clone github.com cd SnapClass ``` 2. Install requirements ``` pip install -r requirements.txt ``` Install AnythingLLM and activate AnythingLLM NPU to process LLM models in Qualcomm Hexagon NPU. Download ```Phi 3.5 Mini Instruct 4K 2.00GB``` model. Get you AnythingLLM Dev …