SDK Documentation

Getting Started

Learn how to install and use the Akioudai Safety SDK to add robust safety measures to your AI systems.

Prerequisites

Before you begin, ensure you have the following requirements in place:

System Requirements

  • Node.js 14 or higher
  • npm 6+ or yarn 1.22+
  • An active Akioudai account
  • API credentials (obtain from your dashboard)

The SDK is compatible with most modern JavaScript frameworks and runtime environments, including:

Node.js Environments

Server-side Node.js applications and services

Browser Applications

React, Vue, Angular, and other frontend frameworks

Serverless Functions

AWS Lambda, Azure Functions, Vercel Functions

Installation

The Akioudai Safety SDK is available as a private npm package. To install it, you need to first authenticate with our private registry.

Important

To request access to our private registry, please contact our sales team or use the Request Access button on our SDK documentation page.

Installing with npm

npm installation
# Authenticate with our private registry
npm config set @akioudai:registry https://registry.akioudai.ai/
npm config set //registry.akioudai.ai/:_authToken="YOUR_AUTH_TOKEN"

# Install the SDK
npm install @akioudai/safety-sdk
bash

Installing with yarn

yarn installation
# Authenticate with our private registry
yarn config set @akioudai:registry https://registry.akioudai.ai/
yarn config set //registry.akioudai.ai/:_authToken "YOUR_AUTH_TOKEN"

# Install the SDK
yarn add @akioudai/safety-sdk
bash

Authentication

Once installed, you need to authenticate the SDK with your API credentials. We recommend using environment variables to securely store these credentials.

Setting up environment variables

Create a .env file in your project root with the following content:

.env file
# .env file
AKIOUDAI_API_KEY=your_api_key_here
AKIOUDAI_API_SECRET=your_api_secret_here
env

Security Best Practices

  • Never hardcode API credentials in your source code
  • Add .env to your .gitignore file
  • Provide different access levels for development, staging, and production
  • Regularly rotate your API keys

Loading environment variables

In Node.js, you can use the dotenv package to load these environment variables:

Loading environment variables
// Install dotenv first:
// npm install dotenv

// At the top of your entry file (e.g., index.js)
require('dotenv').config();

// Then access environment variables
const apiKey = process.env.AKIOUDAI_API_KEY;
const apiSecret = process.env.AKIOUDAI_API_SECRET;
javascript

Basic Usage

Here's a simple example of how to use the Safety SDK to implement safety guardrails for your AI model:

Basic Usage Example
1// Import the SDK
2import { SafetyGuard } from '@akioudai/safety-sdk';
3
4// Initialize the safety guard with your API credentials
5const guard = new SafetyGuard({
6  apiKey: process.env.AKIOUDAI_API_KEY,
7  apiSecret: process.env.AKIOUDAI_API_SECRET,
8  boundaries: ['ethics', 'bias', 'toxicity'],
9  threshold: 0.95
10});
11
12// Function to check AI model outputs
13async function checkModelOutput(output) {
14  try {
15    // Monitor the model output for safety violations
16    const result = await guard.monitor(output, {
17      realtime: true,
18      callbacks: {
19        onViolation: (violation) => {
20          console.log('Safety violation detected:', violation);
21          // Take appropriate action (e.g., filter, log, alert)
22        }
23      }
24    });
25    
26    // If the output is safe, return it
27    if (result.safe) {
28      return output;
29    } else {
30      // Otherwise, return a modified or filtered output
31      return result.filteredOutput || 'I cannot provide that response';
32    }
33  } catch (error) {
34    console.error('Error monitoring model output:', error);
35    // Fallback response in case of error
36    return 'Sorry, I encountered an error processing your request';
37  }
38}
javascript

Understanding the response

The monitor method returns a response object with the following properties:

Response Object
{
  safe: boolean,           // Whether the output is safe
  violations: [            // Array of detected violations (if any)
    {
      type: string,        // Type of violation (e.g., 'ethics', 'bias')
      severity: number,    // Severity score (0-1)
      details: string,     // Description of the violation
      span: {              // Location of the violation in the output
        start: number,
        end: number
      }
    }
  ],
  filteredOutput: string,  // Modified output with violations removed/replaced
  metrics: {               // Performance metrics
    latency: number,       // Processing time in ms
    score: number          // Overall safety score (0-1)
  }
}
javascript

Next Steps

Now that you have the SDK installed and configured, you can explore more advanced features:

Agent Alignment

Learn how to align AI agents with specific objectives and safety guidelines.

Learn more

Safety Guardrails

Explore advanced safety guardrails for different types of content and contexts.

Learn more

Runtime Validation

Implement validation checks throughout the execution of your AI system.

Learn more

Examples

View complete code examples for common use cases and integrations.

Browse examples
PreviousSDK OverviewNextAPI Reference