https://github.com/benderscript/promptguardian

All-in-one App that Checks LLM prompts for Injection, Data Leaks and Malicious URLs.
https://github.com/benderscript/promptguardian

attacks injection llm openai prompt

Last synced: 17 days ago
JSON representation

All-in-one App that Checks LLM prompts for Injection, Data Leaks and Malicious URLs.

• Host: GitHub
• URL: https://github.com/benderscript/promptguardian
• Owner: BenderScript
• License: apache-2.0
• Created: 2024-01-21T10:09:17.000Z (10 months ago)
• Default Branch: main
• Last Pushed: 2024-03-27T06:54:41.000Z (8 months ago)
• Last Synced: 2024-08-08T05:07:05.605Z (3 months ago)
• Topics: attacks, injection, llm, openai, prompt
• Language: Python
• Homepage:
• Size: 34.6 MB
• Stars: 3
• Watchers: 2
• Forks: 2
• Open Issues: 3
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Prompt Guardian

This is application that will check your prompt against a database of 

malicious URLs and whether it is considered prompt injection attack.

It uses OpenAI to check prompt injection attacks. Why? Because it's fun.

Seriously, it has very good guard rails to prevent you from doing dumb things. Therefore, 

if you use another LLM, your prompt will be checked against OpenAI.

Finally it checks your prompt for DLP (Data Loss Prevention). It uses an abstracted

API to check your prompt against a database.

## OpenAI API Key

You should have an environment variable called `OPENAI_API_KEY` with your OpenAI API key. Alternatively,

you can create a `.env` file on the project root directory with the following content:

```bash

OPENAI_API_KEY=

```

## Installing Dependencies

```bash

pip3 install -r requirements.txt

```

## Running

On the project root directory, run the following command:

```bash

uvicorn prompt_guardian.server:prompt_guardian_app --reload --port 9001 

```

If Everything goes well, you should see the following page at http://127.0.0.1:9001

![Landing page](images/landing.png)

## Testing

See the demo below where the App checks a prompt with a malicious URL and injection.

![Demo](images/prompt_guardian_demo.gif)