OpenAI Integration

Overview

The OpenAI integration lets you evaluate GPT model inputs and outputs against your governance policies using the Aguardic SDK. Your application calls OpenAI directly — Aguardic never touches your LLM traffic. Policy evaluation happens in your code before and after each LLM call.

Setup

npm install @aguardic/sdk

Code Example

import Aguardic from "@aguardic/sdk";
import OpenAI from "openai";
 
const aguardic = new Aguardic(process.env.AGUARDIC_API_KEY);
const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });
 
const messages = [
  { role: "user" as const, content: "Summarize our compliance requirements" },
];
 
// 1. Evaluate input before sending to OpenAI
const inputCheck = await aguardic.evaluate({
  input: {
    provider: "openai",
    model: "gpt-4o",
    messages,
  },
  targetKey: "chat-completion",
});
 
if (inputCheck.enforcementAction === "BLOCK") {
  console.log("Blocked:", inputCheck.violations);
  // Do not call OpenAI
} else {
  // 2. Call OpenAI directly
  const completion = await openai.chat.completions.create({
    model: "gpt-4o",
    messages,
  });
 
  const responseText = completion.choices[0]?.message?.content ?? "";
 
  // 3. Evaluate output after receiving from OpenAI
  const outputCheck = await aguardic.evaluate({
    input: {
      provider: "openai",
      model: "gpt-4o",
      response: responseText,
    },
    targetKey: "chat-completion-output",
  });
 
  if (outputCheck.enforcementAction === "BLOCK") {
    console.log("Output blocked:", outputCheck.violations);
  } else {
    console.log(responseText);
  }
}

What Gets Evaluated

Input Evaluation

Before sending a request to OpenAI, pass the request content to aguardic.evaluate():

Output Evaluation

After receiving a response from OpenAI, evaluate the output content:

Output evaluation is optional but recommended. You control whether to block, log, or ignore output violations.

Enforcement Actions

The enforcementAction field in the evaluation response tells you what to do:

Why Sidecar Over Proxy

Unlike proxy-based approaches that route all LLM traffic through a third party, Aguardic's sidecar model means:

• No added latency — Evaluation runs in parallel, not in the request path
• No single point of failure — Your LLM calls go direct; if Aguardic is down, you decide the fallback
• Your keys stay with you — Aguardic never sees your OpenAI API key
• Full control — You choose exactly when and what to evaluate

Next Steps

• Anthropic Integration — Same pattern for Claude models
• Gemini Integration — Same pattern for Google AI
• JavaScript SDK — Full SDK reference
• Your First Policy — Create policies to evaluate against