To start the user study, ensure that the fields in user_study_data/next/config.yml are correct.
Then run export ARC_API_KEY=your_api_key and python server.py and open a browser to localhost:8000.
The server and main_utils scripts do the following:
- Read from
user_study_data/next/config.ymlto find the next user ID and task to perform - Read from
config/tasks/tasks.ymlto find the order of tasks to perform - Executes these tasks, saving to
user_study_data/global_lorebook.pklanduser_study_data/user_(id) - Once the time is up, the simulation closes at the next opportunity. This writes to
next/config.yml
pip install numpy sentence_transformers openai termcolor PyYaml pybullet imageio[ffmpeg]
Or simply run
uv init
uv sync
Then for any python files, run
uv run python *.py
export ARC_API_KEY=YOUR_API_KEY_HERE
python main.py
To provide feedback for a subtask, ctrl+c during execution (or the 5 second grace period after execution)