@authensor/connector-openai

Safety connector for the OpenAI API. Wraps the official openai SDK so every chat.completions.create() call goes through Authensor policy evaluation and Aegis content scanning.

Quickstart

npm install @authensor/connector-openai openai

import { createSafeOpenAIClient } from '@authensor/connector-openai';

const client = await createSafeOpenAIClient({
  apiKey: process.env.OPENAI_API_KEY,
  policy: {
    id: 'my-policy',
    name: 'My Policy',
    version: '1.0.0',
    rules: [
      { id: 'allow-all', effect: 'allow' },
    ],
  },
});

const response = await client.chat.completions.create({
  model: 'gpt-4o',
  messages: [{ role: 'user', content: 'Hello' }],
});

What it does

1. Before each chat.completions.create() call, the request content is evaluated against your policy using the Authensor engine.
2. If the policy says deny, an AuthensorDeniedError is thrown.
3. If the policy says escalate (require_approval), an AuthensorEscalationError is thrown.
4. If the policy says allow, the request is forwarded to the OpenAI API.
5. After receiving a response, Aegis scans the output for threats (PII, injection, credentials, exfiltration).
6. If Aegis flags the response, an AuthensorContentThreatError is thrown.

Configuration

| Option | Type | Default | Description | |--------|------|---------|-------------| | apiKey | string | env var | OpenAI API key | | policy | Policy \| string | required | Authensor policy object or YAML string | | principalId | string | 'openai-connector' | ID for audit trail | | aegisEnabled | boolean | true | Enable Aegis response scanning | | onEvaluation | function | - | Callback after policy evaluation | | onAegisThreat | function | - | Callback when Aegis flags content |

License

MIT