Elsai GuardrailsProgrammable Guardrails

for LLM-based Applications

View Documentation

Comprehensive Protection

Detect and block toxic content, sensitive data, jailbreak attempts, and malicious requests

Flexible Configuration

YAML-based or programmatic configuration with customizable thresholds and rules

Easy Integration

Simple Python API with async support, detailed results, and input/output validation

Multi-LLM Support

Works with OpenAI, Azure OpenAI, Anthropic, Gemini, AWS Bedrock, and more

Detailed Results

Get comprehensive information about guardrail checks and blocking reasons

Separate Checks

Full control with separate input and output validation for custom LLM integrations

Why Elsai Guardrails?

Elsai Guardrails provides a comprehensive solution for securing your LLM-based applications. With built-in protection against common threats and flexible configuration options, you can ensure your AI applications are safe and compliant.

Key Features

Toxicity Detection: Automatically detect and block offensive or harmful content
Sensitive Data Protection: Identify and protect personal information like emails, phone numbers, and credit cards
Content Classification: Detect jailbreak attempts, prompt injection, and malicious code using semantic routing
Multi-LLM Integration: Seamless integration with major LLM providers
Flexible Deployment: Use as a wrapper or perform separate input/output checks

Quick Example

python

from elsai_guardrails.guardrails import LLMRails

# Initialize with configuration
rails = LLMRails.from_config("config.yml")

# Safe LLM calls with automatic guardrails
response = rails.generate(
    messages=[{"role": "user", "content": "Hello!"}]
)

Get Started

Ready to secure your LLM application? Check out our Installation Guide to get started in minutes!