 | Turned a Xiaomi 12 Pro into a dedicated local AI node. Here is the technical setup: OS Optimization: Flashed LineageOS to strip the Android UI and background bloat, leaving ~9GB of RAM for LLM compute. Headless Config: Android framework is frozen; networking is handled via a manually compiled wpa_supplicant to maintain a purely headless state. Thermal Management: A custom daemon monitors CPU temps and triggers an external active cooling module via a Wi-Fi smart plug at 45°C. Battery Protection: A power-delivery script cuts charging at 80% to prevent degradation during 24/7 operation. Performance: Currently serving Gemma4 via Ollama as a LAN-accessible API. Happy to share the scripts or discuss the configuration details if anyone is interested in repurposing mobile hardware for local LLMs. [link] [comments] |
![[link]](https://i.redd.it/fo3jf5vk85vg1.jpeg){kind=link}