KNOX (Knowledge-driven Natural Execution System)

KNOX (Knowledge-driven Natural Execution System) is a multimodal AI agent that enables users to interact with computers naturally using voice, hand gestures, and intelligent AI reasoning. Unlike traditional AI assistants that mainly answer questions, KNOX understands user intent, remembers context, retrieves relevant information, and performs real-world tasks autonomously. Built using React, Electron, FastAPI, LangGraph, and Fireworks AI, KNOX combines speech recognition (Faster-Whisper), gesture recognition (MediaPipe + OpenCV), long-term memory (ChromaDB), and Retrieval-Augmented Generation (LlamaIndex) to deliver context-aware and personalized assistance. Users can control applications, automate desktop tasks, interact with documents, and receive natural voice responses through a modern desktop interface. The system is designed to provide a seamless, hands-free experience by combining multiple input methods with AI-powered decision-making. By integrating advanced AI agent workflows, memory, and tool execution, KNOX moves beyond traditional chatbots to become an intelligent productivity assistant. Powered by Fireworks AI on AMD-supported infrastructure, KNOX demonstrates the future of natural, multimodal human-computer interaction for everyday productivity and accessibility.