AI · Case study
Zentrax
FRIDAY-inspired Windows desktop assistant enabling voice commands and gesture-based control, with local LLM execution for privacy-focused automation.
- Timeline
- October 2025 to March 2026
- Commits
- 12 of 16
- Stack
- Python · Whisper · MediaPipe · WebSocket · Ollama · Docker
Situation
Cloud voice assistants send everything you say to a server. For a personal desktop assistant that controls your machine, that is both a privacy and a latency problem.
Task
Build a privacy-first desktop assistant that runs its intelligence locally and supports both voice and gesture control.
Action
I built a Windows assistant that runs an LLM locally through Ollama, uses Whisper for on-device STT and intent recognition, and adds hands-free gesture control via MediaPipe with fallback pipelines for reliability. A WebSocket dashboard provides real-time monitoring, and the components are containerised with Docker.
Result
A fully local voice-and-gesture assistant that keeps data on-device while automating desktop tasks.
Local LLM
privacy-first (Ollama)
Voice + Gesture
dual control
Real-time
monitoring dashboard