This repository is dedicated to addressing the prevalent challenge of prompt injection encountered by Language Model Models (LLMs). By providing insights, strategies, and solutions, we aim to equip developers and practitioners with effective approaches to mitigate and overcome prompt injection issues in the LLM domain. Explore the resources and knowledge shared here to enhance your understanding and effectively tackle prompt injection problems within LLMs.
To see the project in action here:
The project report that gives background and compares different methods to circumvent this problem and explaination of proposed solution is stored here: Project Report
./CatchPromptInjection/
├── src
│ ├── config
│ │ └── config.json
│ ├── tests
│ │ └── test_detector.py
│ └── detector.py
├── Dockerfile
├── LICENSE
├── Makefile
├── README.md
├── app.py
└── requirement.txtPlease remember to the config.json file to control parameters.
- Make sure you have python 3+ installed on your computer.
- Make sure you have pip3 package installed on your computer.
- Make sure you clone this package to your computer.
In your terminal/command-line go to the project folder and execute the command below:
pip install -r requirement.txtPlease run these make commands to make sure the formatting, linting and tests are working.
make format
make lint
make test
To run the app locally just run the following command in bash in the root directory of the folder:
streamlit run app.pyYour app is available at: https://localhost:8501
To run the project with docker simply run the following command to build:
docker build -t <IMAGE_NAME_TO_GIVE> ./Once the image is built please run the container using this command:
docker run -p 8501:8501 <IMAGE_NAME_TO_GIVE>Once started the app is available at: https://localhost:8501