woaiys3/deepseek-harness-android-app
DeepSeek Harness 手机版:可直接安装的 Android APK,AI 免 Root 操作手机(Shizuku/root 可选),文件编辑只需所有文件访问权限,前台保活 + AI 通知
What it solves
This project packages the DeepSeek Harness (DSH) kernel into a standalone Android APK, transforming a chat-based AI into a functional mobile agent. It solves the difficulty of deploying AI agents on Android by removing the need for Termux or complex command-line setups, allowing an AI to interact with the phone's system, apps, and web pages without requiring root access (via Shizuku).
How it works
The app embeds a Node.js runtime, Python 3.14.6, and the DSH kernel. It uses a series of specialized plugins to grant the AI control over the device:
- Privileged Access: Uses Shizuku or Root to execute system shell commands, manage apps, and modify system settings.
- Accessibility Services: Enables the AI to read the screen's control tree and simulate taps, typing, and scrolling without root.
- Virtual Screen (vscreen): Creates a separate virtual display where the AI can launch apps and perform tasks in the background, streaming the view to the user via an H.264 video feed.
- AI Browser: A dedicated browser process that allows the AI to navigate the web using DOM semantic snapshots and stable references rather than just screenshots.
Who it’s for
- Android users who want to run a fully autonomous AI agent on their mobile device.
- Developers interested in mobile AI automation and agentic workflows.
- Users who want an AI to perform system-level tasks (like installing apps or changing settings) without rooting their device.
Highlights
- Non-Root System Control: Leverages Shizuku to provide system-level privileges for app installation and setting modifications.
- Virtual Display: AI can operate in a virtual screen, allowing the user to continue using their main screen uninterrupted.
- Semantic Web Navigation: An AI browser that reads structured DOM elements for precise interaction instead of guessing from images.
- Self-Healing Core: Includes a kernel self-check and recovery ledger to fix missing or corrupted files without wiping user data.
- Built-in Dev Environment: Comes with Python and npm/npx, enabling the AI to write and execute scripts for data processing.
- Customizable Console: A highly flexible UI where users can modify layouts, colors, and behaviors via a JSON configuration file.
Related
- Project
- Project
- Project
- Project
- Project