Jarvis Assistant (Windows Desktop AI Assistant)
Voice-first desktop AI assistant for Windows with real action routing and deployment-ready releases.
Project Snapshot
Jarvis Assistant is a production-focused Windows desktop AI assistant built in Python with a modern custom UI. It combines speech recognition, LLM reasoning, actionable command routing, and packaging into an end-to-end user flow: speak or type commands, execute real actions, receive spoken and visual responses, and distribute as EXE plus installer.
Key Contributions
- Designed a polished multi-screen desktop UI (Home, Chat, Settings, Commands) with tkinter and ttkbootstrap.
- Built end-to-end voice flow: microphone input to STT to action/LLM orchestration to TTS response.
- Implemented UX controls like mic toggle, echo suppression, noise cancellation level, and rolling transcript highlights.
- Added safer automation via git-action detection, approval popup before execution, and visible repo path context.
- Improved reliability for startup/audio flows and clarified incomplete-request handling with better command fallbacks.
- Prepared release pipeline with installer scripts and trust-focused release notes for smoother SmartScreen onboarding.
Tech Stack
Python, OpenAI API, SpeechRecognition, PyAudio, pyttsx3, tkinter, ttkbootstrap, Pillow, requests, psutil, pycaw, comtypes, python-dotenv, pyinstaller, pytest.
Architecture
jarvis/ui for desktop UX, jarvis/speech for voice I/O, jarvis/brain for LLM reasoning and memory, jarvis/actions for system/web/dev routing, plus scripts and installer for packaging.
AI/RAG Positioning
This implementation is RAG-inspired and tool-augmented rather than full vector-database RAG. It uses a command contract for executable intents, conversation memory for continuity, and deterministic action routing. The current architecture is ready for future vector retrieval extension.
Portfolio Impact
Built a production-focused Windows AI assistant that merges conversational AI with practical desktop automation, with strong UX polish, safety-first command execution, voice reliability improvements, and deployment-ready packaging.