Offline AI on Android

This project explores running local language models entirely offline on Android devices, using open-source tools like llama.cpp, UserLAnd, and various Linux distributions (Arch, Alpine, Ubuntu, Kali) installed in a rooted/non-rooted environment.

The goal: turn old or underused Android devices into capable AI inference machines with complete user control and no cloud dependency.

---

What’s Included

• Install and run local LLMs (e.g., TinyLLaMA) on Android using UserLAnd
• Compile and use llama.cpp on-device (OnePlus 3T, 6GB RAM )
• Scripts for installing dependencies and launching models
• Architecture for reproducible experiments and logging

---

Requirements

• Android phone (6GB RAM or more recommended)
• UserLAnd app installed from Play Store
• Working session using Arch Linux (or other supported distro)
• 3–5GB free storage for model + tools

---

📁 Repo Structure

offline-ai-android/
├── scripts/                # install and run helpers
├── docs/                   # setup logs and markdown walkthroughs
├── userland-setup/         # Linux distro notes
├── experiments/            # test runs, logs, model benchmarks
├── models/                 # tested gguf models
└── assets/                 # screenshots, diagrams