An open API service indexing awesome lists of open source software.

https://github.com/xinzhang-chen/tosense

TOSense is a Chrome browser extension that demonstrates a proof-of-concept workflow for analyzing website Terms of Service (TOS) using a large language model (LLM). It enables basic LLM-powered interaction with TOS content, designed for academic exploration, prototyping, and feedback-driven development.
https://github.com/xinzhang-chen/tosense

chrome-extension proof-of-concept terms-of-service

Last synced: 2 months ago
JSON representation

TOSense is a Chrome browser extension that demonstrates a proof-of-concept workflow for analyzing website Terms of Service (TOS) using a large language model (LLM). It enables basic LLM-powered interaction with TOS content, designed for academic exploration, prototyping, and feedback-driven development.

Awesome Lists containing this project

README

        

TOSense Logo

# TOSense – We Read, You Click

**TOSense** is a Chrome browser extension that demonstrates a proof-of-concept workflow for analyzing website Terms of Service (TOS) using a large language model (LLM). It enables basic LLM-powered interaction with TOS content, designed for academic exploration, prototyping, and feedback-driven development.

TOSense is part of the following research paper:

> **"ToSense: We Read, You Click"**\
> *Xinzhang Chen, Hassan Ali, Arash Shaghaghi, Salil S. Kanhere, Sanjay Jha*\
> *Under review at IEEE/IFIP DSN 2025*

> ⚠️ This is a **proof-of-concept** version. The current release focuses on demonstrating the workflow and integration with an external server hosting our LLM backend.

---

## ✨ Key Features

- 🧠 Simple LLM-based interface for querying Terms of Service
- πŸ”„ Communicates with a backend server to demonstrate the TOS-to-LLM pipeline
- πŸ› οΈ Ideal for research demonstrations and future feature prototyping

---

## πŸš€ Installation

### πŸ“Ί Installation Video

https://github.com/user-attachments/assets/f47141f4-f1ca-42b9-a3ff-69ac477b2c8f

### Manual Installation Steps

1. Clone or download this repository.
2. Open [chrome://extensions](chrome://extensions) in your browser.
3. Enable **Developer mode** (top right corner).
4. Click **Load unpacked** and select the `chrome-mv3-prod` folder from this repo.
5. You should see **TOSense** appear in your extension bar.

---

## ℹ️ How to Use

### πŸ“Ί Usage Demo

https://github.com/user-attachments/assets/67a9b9cc-6e47-41c1-a8fa-96fbdab6213c

1. Open a new tab inside your browser (a landing page also works).
2. Click the **TOSense** extension icon (if it is hidden [by default], you need to click the extension icon first).
3. Enter a question or select a pre-configured question (e.g., β€œCan I permanently delete my account and data?”).
4. Send the question.
5. View the LLM-generated response in the popup interface.

> ⚠️ The current server is hosted on a low-spec testing server. Expect possible delays or limited performance in LLM responses. Server will be adjusted in the next iteration.

> ⚠️ Please note that the provided version will not use the ToS content extract from the current page; all the answers returned from the server will base on our pre-loaded
> LinkedIn ToS document. (Due to computing power limitations)

---

## 🧠 Architecture Overview

```txt
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β” Sends Request (User Prompt) β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ TOSense β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β–Ίβ”‚ Backend β”‚
β”‚ (Chrome) │◄───────────────────────────────────── Server β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜ Returns Response β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
β”‚ β–²
└─ Page Context [❌ Not supported in this version]β”€β”˜

```

---

## πŸ›‘ Current Limitations

- ❌ Does not detect or extract TOS from arbitrary web pages (but we already have the key component; more info checkout [tos-crawl](https://github.com/Xinzhang-Chen/tos-crawl))
- ❌ Does not provide fast access to pre-crawled TOS repositories
- 🐒 It relies on a low-powered backend for LLM responses
- 🌐 LLM model is a reduced version for performance reasons

The next version of the tool will focus on addressing these limitations.

---

## ⚠️ Disclaimer

This tool is provided solely for **non-commercial, academic research, and demonstration** purposes. The maintainers do not guarantee the performance, uptime, or accuracy of the LLM responses, especially under heavy usage.

> Users are responsible for complying with applicable laws and the website's terms of service.

---

## πŸ“„ License

This project is licensed under the **GNU Affero General Public License v3.0 (AGPL-3.0)**.
See the [LICENSE](./LICENSE) file for more information.

---

## πŸ‘©β€πŸ’» Maintainer

- Xinzhang Chen - [email protected]
- Hassan Ali - [email protected]
- Dr Arash Shaghaghi - [email protected]

🚧🚧🚧🚧🚧🚧🚧🚧🚧🚧

This project is under **active development** πŸ› οΈ.

New features and improvements are being added continuously. **Stay tuned!**