Extending older smartphone lifespans with local-first AI and zero cloud uploads

Setting For Me, an open-source, local-first AI 'Device Doctor' for older Android smartphones, aims to extend device lifespans by letting users describe issues in plain language and applying pre-validated actions without cloud uploads. The project, distributed under the MIT License, outlines a phased roadmap from a rule-based core to predictive maintenance using on-device SmolLM2-360M-Instruct (Q4) and LoRA fine-tuning.

Powered by Vision 50 Years PhoneSoftware should adapt to hardware, not force hardware replacement. Setting For Me is a lightweight, local-first, open-source AI "Device Doctor" designed for existing and older Android smartphones. Instead of forcing users to navigate complex system menus, users simply state what they want in plain language. A tiny, local on-device model maps their intent to safe, verified actions. Before The Problem : Older phones suffer from software degradation, unwanted background bloat, and thermal throttling. After Setting For Me : Your phone explains what is wrong in plain, honest language and fixes it with a single tap—no cloud uploads, no fake boosts, and no technical jargon required. No Adware or Trackers: We will never bundle third-party cleaner tools, bloatware, or ad networks. No Fake RAM Boosts: We will never run fake memory clearing scripts that kill background apps only for Android to immediately restart them. No Data Uploads: Local-first architecture. All processing, logs, and execution happen strictly on-device. No Unverified Execution: The AI model cannot generate or execute arbitrary shell commands; it can only select pre-validated Action IDs from settings map.json. Phase 0 Rule-Based Core : Search bar, 20-30 validated system actions, Device Health Report, Safety Gate, Evidence Ledger. Phase 1 Model Frontend : Integrate on-device SmolLM2-360M-Instruct Q4 for handling un-scripted query variations. Phase 2 LoRA Fine-Tuning : Fine-tune the 360M model specifically on device-health intent mapping. Phase 3 Cross-Service Diagnostics : Correlate thermal, battery, storage, and wake-lock telemetry to explain root causes. Phase 4 Predictive Maintenance : Multi-week telemetry modeling to predict hardware/battery degradation before failure. Distributed under the MIT License.