Module Overview
AI Sentinel is designed as a modular system for content validation and security checking using Large Language Models (LLMs). Here’s how the components work together:
Guard System
Content validation engine
Configurable rule sets
Extensible plugin architecture
LLM Integration Layer
Multi-provider support
Standardized API interface
Automatic failover capabilities
Environment Setup
Required Environment Variables:
# Open Source OpenAI Configuration
OPENAI_API_BASE=your_base_url_here
# Azure OpenAI Configuration
AZURE_API_KEY=your_azure_key_here
AZURE_API_BASE=your_endpoint_here
AZURE_API_VERSION=your_version_here
# Google Gemini AI Configuration
GEMINI_API_KEY=your_gemini_key_here
Supported LLM Services
AI Sentinel supports multiple LLM API services through a unified interface. It is Model-agnostic, meaning you can switch between different LLM providers with minimal code changes. Currently, the following LLM services are supported:
Provider |
Models (Examples) |
Authentication |
Notes |
|---|---|---|---|
OpenAI Compatible Server |
Qwen, Meta Llama, etc. |
Base URL |
Used for open-source models using OpenAI’s compatibility server |
Azure OpenAI |
gpt-4o-mini |
API Key + Endpoint + Version |
Closed source, requires Azure subscription |
Google Gemini AI |
gemini-2.5-flash |
API Key |
Closed source, requires Google Cloud subscription |
Security Considerations
API Key Management - Use environment variables - Rotate keys regularly - Never commit keys to source control
Rate Limiting - Implement exponential backoff - Monitor usage quotas - Set up alerts for unusual activity