SentinelAI SDK Documentation

Installation

Install the SentinelAI SDK using your preferred package manager:

npm install @sentinel-ai/sdk

yarn add @sentinel-ai/sdk

pip install sentinel-ai

Quick Start

Initialize the SDK and analyze your first AI interaction:

import { SentinelAI } from '@sentinel-ai/sdk';

// Initialize with your API key
const sentinel = new SentinelAI({
  apiKey: 'sk_live_your_api_key_here',
  organizationId: 'your_org_id'
});

// Analyze an AI interaction
const result = await sentinel.analyze({
  prompt: 'User prompt text',
  response: 'AI response text',
  userId: 'user_123',      // Optional: for tracking
  sessionId: 'sess_456'    // Optional: for grouping
});

console.log('Risk Score:', result.final_risk_score);
console.log('Decision:', result.decision);
console.log('Flags:', result.flags);

Configuration

Configure the SDK with your organization's settings:

const sentinel = new SentinelAI({
  apiKey: process.env.SENTINEL_API_KEY,
  organizationId: process.env.SENTINEL_ORG_ID,
  
  // Optional: Custom configuration
  timeout: 5000,              // Request timeout in ms
  retries: 3,                 // Number of retries
  endpoint: 'https://api.sentinel-ai.com', // Custom endpoint
  
  // Optional: Callbacks
  onError: (error) => {
    console.error('SentinelAI Error:', error);
  },
  
  onDetection: (result) => {
    if (result.decision === 'block') {
      // Handle blocked content
    }
  }
});

Real-time Monitoring

Set up real-time monitoring for streaming AI responses:

// Enable real-time monitoring
sentinel.enableRealTimeMonitoring({
  // Called when risk is detected mid-stream
  onRiskDetected: (result, context) => {
    console.warn('Risk detected:', result.flags);
    
    // Optionally block the stream
    if (result.decision === 'block') {
      context.stopStream();
    }
  },
  
  // Called when compliance issues are detected
  onComplianceIssue: (result) => {
    console.error('Model compliance:', result.explanation);
  }
});

// Use with streaming AI APIs
const stream = await openai.chat.completions.create({
  model: 'gpt-4',
  messages: [{ role: 'user', content: prompt }],
  stream: true,
});

// Wrap the stream with SentinelAI monitoring
const monitoredStream = sentinel.monitorStream(stream, {
  prompt: prompt,
  userId: 'user_123'
});

for await (const chunk of monitoredStream) {
  process.stdout.write(chunk.choices[0]?.delta?.content || '');
}

API Reference

`analyze(options)`

Analyze a prompt/response pair for potential risks.

Parameter	Type	Required	Description
prompt	string	Yes	User input prompt
response	string	Yes	AI response text
userId	string	No	User identifier
sessionId	string	No	Session identifier

`getMetrics()`

Retrieve usage metrics for your organization.

const metrics = await sentinel.getMetrics({
  days: 30  // Time range
});

console.log(metrics.total_requests);
console.log(metrics.avg_latency_ms);
console.log(metrics.risk_distribution);

Integration Examples

Chatbot Integration

Protect your customer support chatbot from prompt injection

View example

Content Generation

Monitor AI-generated content for compliance issues

View example

Code Assistant

Secure your AI coding assistant from malicious requests

View example

API Gateway

Add SentinelAI as middleware to your API

View example