Firewall for AI API Sample Code

Overview

Firewall for AI enables rapid identification and anonymization to protect sensitive data from disclosure through GenAI prompts, models, analytics pipelines, and automated workflows.

Key Features

• Comprehensive Coverage: Supports over 100 entity types in text, files, and images, including names, addresses, Social Security numbers, credit card details, credit card images, banking account numbers, and protected health information. For a full list, refer to the detector glossary.
• Enhanced Accuracy: Delivers up to twice the precision and recall compared to leading solutions like Google DLP, Microsoft Presidio, and AWS Guardrails. Benchmarks are available upon request at [email protected].
• Low Latency: ≤100ms P99
• High Throughput: >= 1000 rps

Repository Contents

This repository contains sample code demonstrating how to use Firewall for AI APIs to sanitize GenAI prompts within popular LLM services and frameworks using Python.

1. OpenAI ChatGPT
2. Anthropic Claude
3. LangChain / Anthropic Claude

Prerequisites

1. If you still need to get a Nightfall account, get one here.

2. Create a Nightfall key and add it to your .env file. Here are instructions.

3. Ensure you have the following packages installed:

   pip install -r requirements.txt

Contributing

Feel free to submit issues or pull requests if you find any bugs or have suggestions for improvements.

License

This project is licensed under the Apache 2 license.

Formatting and style

To sort imports run:

isort .

To format code run:

black .

To check code style run:

pylint examples/