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:

  1. Guard System

    • Content validation engine

    • Configurable rule sets

    • Extensible plugin architecture

  2. 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

  1. API Key Management - Use environment variables - Rotate keys regularly - Never commit keys to source control

  2. Rate Limiting - Implement exponential backoff - Monitor usage quotas - Set up alerts for unusual activity