Unified JARVIS: A Comprehensive Survey and Framework for AI-Enhanced Desktop Virtual Assistants

Virtual Assistants (VAs) have progressed from simple command-based systems to intelligent, multimodal, and context-aware frameworks capable of complex reasoning, desk top automation, and accessibility support. Despite commercial success, academic research prototypes reveal significant oppor tunities for assistants that combine natural language processing, large language models (LLMs), computer vision (CV), and GUI automation into a unified architecture. This survey synthe sizes findings from two Jarvis-inspired assistants—“Jarvis: AI Enhanced Desktop Virtual Assistant” and “JARVIS- Smart AI Support System”—while embedding them into a broader landscape of conversational AI research. We review advances in speech recognition, intent classification, OCR, face recognition, gesture detection, LLM-guided automation, sandboxing, and privacy-preserving AI. The paper proposes a unified architecture that integrates perception, orchestration, execution, knowledge, and safety layers. With more than 80 references, detailed algo rithm designs, evaluation protocols, and ethical case studies, this paper contributes a comprehensive 30-page IEEE-style resource for researchers and practitioners.