Allow me to introduce VYSP.AI

What is VYSP.AI?

VYSP.AI is a platform that simplifies the configuration and implementation of guardrails for GenAI applications. Guardrails (or what we call Rules) are ML models that perform NLP (and soon, Computer Vision) tasks to keep your applications and users secure. You can easily configure Rules through the web UI and you'd never have to worry about having the processing power to run the machine learning models yourself, because VYSP.AI handles that for you.

So, how does it work?

As you can see in the image above, it works by sitting in between your application and whatever LLM (or LLM service API) you're using, scanning inputs and outputs to ensure risk can be minimized. The flow goes like this:

Before your application ever interacts with VYSP.AI, it will collect some input from your end user, maybe through a web or mobile application.
We start off by checking if the input is risky by sending it to VYSP.AI
VYSP.AI then runs the configured input guardrails (Rules) and returns whether or not the input was flagged and the result from each Rule filter applied
Based on the result, you may want to re-prompt the user if the input was flagged.
Send the input to your LLM of choice.
Upon receiving a response from yout LLM, send the input and the output to VYSP.AI to check the model's output against the applied Rules.
Based on the result, you may want to re-ask the LLM.
Display the output to your user!

Why should I use this service?

VYSP.AI was built for AI application developers and teams looking to easily secure and evaluate inputs and outputs from their LLMs!

Benefits of using VYSP.AI

VYSP.AI is your go-to solution for keeping AI applications safe. Our platform leverages advanced machine learning models to catch and block harmful or inappropriate inputs and outputs in your AI software, ensuring robust protection for your AI-driven services.

Key Features

Real-time Monitoring and Logging: Track AI application traffic in real time.
Detailed Analytics and Reporting: Generate comprehensive reports and analytics on detected threats, blocked inputs/outputs, and overall firewall performance.
Customizable Rules and Policies: Create and customize your own security rules and policies tailored to your specific needs.
User Behavior Analytics: Detect unusual activities or potential security threats by analyzing user behavior patterns.
Integration with Datadog (Coming Soon): Seamlessly integrate with Datadog for centralized logging and threat management.
Support for Multiple ML Models: Choose from a variety of machine learning models or integrate your own for optimal security measures.
API Rate Limiting and Throttling: Protect against API abuse with rate limiting and throttling features.

Real-world Applications

Following are some example of use cases for integrating VYSP.AI in LLM-enabled applications:

Enhanced Customer Support

AI-driven customer support systems can greatly benefit from VYSP.AI's ability to filter out inappropriate or harmful language, ensuring that interactions remain professional and respectful.

Secure Content Moderation

For platforms that rely on user-generated content, VYSP.AI can automatically flag and block harmful or inappropriate content, maintaining a safe and welcoming environment for all users.

Regulatory Compliance

Organizations operating in regulated industries can use VYSP.AI to ensure that their AI applications comply with industry-specific regulations and standards, reducing the risk of non-compliance.

AI Ethics and Bias Mitigation

By applying VYSP.AI's customizable rules and policies, developers can mitigate biases in AI models, promoting fair and ethical AI usage.

Future Developments

We are continuously working on expanding the capabilities of VYSP.AI. Upcoming features include:

Integration with More Third-party Services: Beyond Datadog, we plan to integrate with other popular third-party services to enhance the versatility and usability of VYSP.AI.
Enhanced Computer Vision Capabilities: Expanding beyond NLP, we will introduce computer vision guardrails to secure applications involving image and video processing.
Automated Threat Response: Advanced automation features to enable proactive threat mitigation, reducing the time and effort required to manage security incidents.