AutoLXB  by wuwei-crg

On-device Android automation framework

Created 4 months ago
271 stars

Top 94.9% on SourcePulse

GitHubView on GitHub
Project Summary

AutoLXB is an experimental, on-device Android automation framework designed for repetitive, linear, and triggerable tasks. It targets users needing to automate daily phone actions like scheduled check-ins, notification-based replies, or fixed-page information retrieval and submission, offering a persistent daemon for background execution. The primary benefit is enabling hands-free automation of routine mobile interactions.

How It Works

The framework employs a "Route-Then-Act" design. Tasks first attempt to replay a saved, deterministic navigation route. If the route is insufficient or unavailable, a vision model handles dynamic UI interactions. This hybrid approach aims to improve reliability and reduce reliance on computationally expensive model calls for predictable workflows. Task routes are generated from recorded user interactions, combining screenshots, model actions, and accessibility data.

Quick Start & Requirements

  1. Install: Download and install the APK from Releases.
  2. Setup: Enable Developer Options, USB debugging, and Wireless debugging (for non-root). Configure battery optimization for AutoLXB to "Unrestricted". Rooted devices require su permission; non-rooted devices require initial Wireless ADB pairing.
  3. LLM/VLM Config: Configure an OpenAI Chat Completions-compatible LLM/VLM endpoint with its API Base URL and Key. The model must support image understanding.
  4. Task Creation: Define tasks concretely (e.g., "Open app, enter check-in page, complete check-in"). Tasks can be scheduled or triggered by notifications.
  5. Route Saving: Optionally save task routes after execution for future deterministic replay.

Prerequisites: Android 11 (API 30) or above, real Android device, Developer Options enabled, root access or Wireless debugging enabled, configured LLM/VLM endpoint. Documentation: User manual available at AutoLXB Docs, demo video at Bilibili (BV114RbBfEou).

Highlighted Details

  • Supports scheduled tasks (e.g., daily app check-ins) and notification-triggered tasks (e.g., specific message replies).
  • Sample tasks include Baidu Tieba one-click sign-in, Bilibili text posting, and Luckin Coffee order automation.
  • Task routes are portable and can be imported/exported.
  • Includes structured Trace logs for troubleshooting.

Maintenance & Community

The project README does not detail specific contributors, sponsorships, or community channels like Discord/Slack. Links to official documentation and demo videos are provided.

Licensing & Compatibility

The project is licensed under the MIT license. This permissive license generally allows for commercial use and integration into closed-source projects without significant restrictions.

Limitations & Caveats

AutoLXB is explicitly marked as an "Experimental" framework. Certain Chinese Android ROMs (MIUI/HyperOS, ColorOS, Flyme) may require additional specific permission adjustments beyond standard setup. The effectiveness of the vision model component is dependent on the quality and capabilities of the configured LLM/VLM endpoint.

Health Check
Last Commit

2 weeks ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
4
Star History
79 stars in the last 30 days

Explore Similar Projects

Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems").

macOS-use by browser-use

0.1%
2k
AI agent for macOS app automation
Created 1 year ago
Updated 1 year ago
Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems") and Gregor Zunic Gregor Zunic(Cofounder of Browser Use).

mobilerun by droidrun

0.6%
8k
Framework for controlling Android devices via LLM agents
Created 1 year ago
Updated 23 hours ago
Feedback? Help us improve.