This repository contains the source code for AutoTask
AutoTask
is a system that can automatically execute arbitrary voice commands by exploring and learning from mobile GUI based on Large Language Models (LLMs).
AutoTask
is built and tested in Android devices. To use AutoTask
, you need to follow the steps below:
- Download and install Android Debug Bridge (adb)
- Download the apk file in the release page and install it in your Android device. (We recommend using Android emulator to run
AutoTask
, which is the experimental environment we used in evaluation. Android emulator can be downloaded from here in Android Studio.) - Open the app and grant the required permissions to start the ActionRecord service.
- Connect your Android device to your computer via USB cable (if you use a real device, if you use Android emulator, you do not need to do this step), enter
adb reverse tcp:5002 tcp:5002
in the command line to forward the port.
- Install dependencies:
pip install -r requirements.txt
cd AutoTask
pip install -r requirements.txt
- Start the back-end server:
python main.py --task "[YOUR_TASK]"
(More parameters can be accessed bypython main.py --help
) - Press
Enter
to start the exploration process.
If you are using OpenAI models (such as the default gpt-3.5-turbo
), please obtain an OpenAI API key on their website then set the environment variable OPENAI_API_KEY
to your API key by running the following command in your terminal:
export OPENAI_API_KEY=<your key>
If you use AutoTask in your research, please cite our paper:
@misc{pan2023autotask,
title={AutoTask: Executing Arbitrary Voice Commands by Exploring and Learning from Mobile GUI},
author={Lihang Pan and Bowen Wang and Chun Yu and Yuxuan Chen and Xiangyu Zhang and Yuanchun Shi},
year={2023},
eprint={2312.16062},
archivePrefix={arXiv},
primaryClass={cs.HC}
}
The MIT license.