Geneline-X/AfricaQuantLM

AfricaQuantLM: A browser-based demo of a 1B-parameter Gemma3 model quantized to INT4, enabling fast, offline AI for African users. Features education, agriculture, health, and legal modes, it cuts memory by 75% and speeds inference 2-4x. See arXiv paper (arXiv:2506.XXXXX). Join us to advance equitable AI in Africa! # GeniGuideSL - Offline AI Assistant for Sierra Leone GeniGuideSL is a modular offline AI assistant designed specifically for users in Sierra Leone. It provides helpful information across multiple domains without requiring an internet connection after initial setup. ## Features - **100% Offline Operation**: Once loaded, works completely without internet - **Local Storage**: Conversations are saved locally on your device - **Multiple Specialized Modes**: - 🧑🏽‍🏫 **Study Mode**: Educational assistance for students - 🌾 **Farming Mode**: Agricultural advice and techniques - 🏥 **Health Mode**: First aid and health education (not medical diagnosis) - ⚖️ **Law Mode**: Basic legal information and rights - 💼 **Career Guide**: Job preparation and entrepreneurship tips ## Technical Details - Built using MediaPipe's LLM Inference API - Powered by Gemma 3 1B-IT (quantized to INT4 for efficiency) - Fully client-side processing - no data leaves your device ## Setup Instructions 1. Make sure you have the model file `gemma3-1b-it-int4.task` in the same directory as the application 2. Open `index.html` in a web browser 3. Wait for the model to load (this may take a moment) 4. Start interacting with the assistant! ## Usage Tips - Select the appropriate mode for your question using the buttons at the top - Clear conversation history with the "Clear Chat" button - Press Enter to send your message (Shift+Enter for a new line) - Each mode provides specialized assistance in its domain ## Development GeniGuideSL is designed to be modular and extensible. The system uses specialized prompts for each mode to provide contextually relevant responses. ## License This project includes components licensed under the Apache License, Version 2.0.