A fully offline AI system built to run voice commands, AI tasks, and system automation directly on the local machine without cloud dependency. Built a real-time voice system using Faster-Whisper for speech-to-text and Ollama local models for reasoning. Supports voice commands, automation workflows, debugging tasks, and local AI execution. with low latency and maintains conversation context. Designed a modular architecture where each feature is a separate Python skill module. New features can be added by dropping a file into the system without touching core code. Built system automation using ADB, allowing voice control of Android devices — opening apps, sending messages, navigation, and checking device status. Implemented a dual-mode system: Fast mode for instant voice responses Heavy mode for complex AI tasks like debugging and automation Added worker isolation using multiprocessing so heavy AI tasks never block voice input. Built a live monitoring dashboard showing system status, active tasks, CPU/RAM usage, and execution logs in real time. Includes automatic recovery system that restarts failed or frozen AI workers without stopping the main system. Added process monitoring and watchdog recovery that automatically detects frozen workers and safely restarts them without affecting the main voice system. Implemented separate fast-response and heavy-workflow execution paths to keep voice interaction responsive during long AI tasks. 🔗 Demo Video: https://drive.google.com/file/d/15aT0SbL-bgEsEy-pLC1-Gyj09cFdAkgW/view?usp=sharing