I GAVE JARVIS VISION | Building a VLM for my Omarchy Assistant

I gave my voice-controlled Linux computer eyes. OmaVision adds on-demand camera vision to OmaVoice, my AI voice assistant built on Omarchy Linux—so I can talk to my computer, show it physical objects, and ask it to take action. In this demo, OmaVision examines a Raspberry Pi, identifies a watch and searches for it online, recognizes a Lightning McQueen collectible, interprets a drawing, and plays tic-tac-toe on paper. The same assistant can switch workspaces and control the desktop through voice commands. I also walk through the architecture behind the prototype: GPT Live handles the voice interaction, an action model operates the computer, and a separate vision companion examines camera captures. I explain how I chose models for responsiveness and visual recognition, and show the microphone, audio interface, and camera setup. Omarchy, created by David Heinemeier Hansson (DHH), provides the foundation. OmaVoice and OmaVision are my experiments in making that desktop more personal and interactive—a step toward a Jarvis-like assistant that can see, talk, and act. • GitHub repository: https://github.com/wombatoperator/oma... • Previous video — building OmaVoice on Omarchy:    • I BUILT JARVIS IN OMARCHY   • Omarchy: https://omarchy.org/ What would you want a computer that can see and respond to help you do? #Omarchy #ComputerVision #Linux