TaskSense

Overview

Paper: https://dl.acm.org/doi/abs/10.1145/3715014.3722070

Task Sense is a project designed to use Large Language Models (LLMs) to coordinate sensor systems to complete complicated user queries. It provides a system that includes modules for execution, planning, responding, and testing, as well as data handling and configuration management.

Project Structure

./src
├── api
│   ├── __init__.py
│   └── llm_apis.py
├── check
│   ├── activity_label_list
│   │   └── inlab_activity.txt
│   └── prompt
│       └── prompt_inlab.txt
├── execution
│   ├── dynamic_plan_adaptation
│   │   ├── adjuster.py
│   │   ├── alternative_path_manager.py
│   │   └── metrics.py
│   ├── model_zoo
│   ├── __init__.py
│   ├── executor.py
│   └── toolbox.py
├── planning
│   ├── __init__.py
│   ├── checker.py
│   ├── example_library_manager.py
│   ├── grammar_rules_manager.py
│   ├── plan_validator.py
│   ├── planner.py
│   └── vocabulary_set_manager.py
├── responding
│   ├── __init__.py
│   ├── respondent.py
│   └── formatter.py
├── __init__.py
├── coordinator.py
├── process_pool.py
└── utils.py

Installation

To set up the project locally, follow these steps:

1. Clone the repository:

   git clone https://github.com/KAIWEILIUCC/tasksense.git
   cd tasksense_demo_code

2. Create a virtual environment (optional but recommended):

   conda create -n your_env_name python=3.10
   conda activate your_env_name

3. Install the dependencies:

   pip install -r requirements.txt

Configuration

Update the configuration file at config/config.yaml with the necessary settings.

Please apply for the corresponding API key from the LLM API service provider (e.g., Volcengine) and fill it in the corresponding section of config/config.yaml.

Please register on the official Cohere website and apply for an API key, then fill it in the config/config.yaml file under:

cohere:
    api_key: 

Usage Examples

To run the project, execute the cli.py script:

python cli.py

Here are 4 example commands:

1. Please check whether anyone made a phone call to someone on Nov 16, 2023?

2. Did anyone show aggressive behaviors on 2023.11.16?

3. Was Bob troubled by a bad dream on November 16th, 2023?

4. Does Bob have difficulty falling asleep (cannot fall asleep within 30 minutes) on Nov 16, 2023?

Response for Example 1:

Response for Example 2:

Response for Example 3:

Response for Example 4:

Acknowledgments

This project is built upon HuggingGPT. We would like to express our gratitude to the original authors for their excellent work.