ai-gateway

version	1
name	ai-gateway
description	Add AI/LLM capabilities to deployed applications. Provision an API key for your deployment to access OpenAI-compatible endpoints (GPT, Claude, Gemini) with usage billed to your organization.

AI Gateway for Deployed Applications

Provision an AI Gateway API key to access LLM APIs (OpenAI, Anthropic, Google) through an OpenAI-compatible interface.

Requires Rebyte API auth — $AUTH_TOKEN and $API_URL are set up per the agent's system prompt; use them as Bearer token and base URL.

When to Use This Skill

Use this skill when:

• Building an AI chat application
• Adding AI-powered features to a web app
• Need LLM access from a serverless function
• Setting up OpenClaw or other AI agents
• Want to use Vercel AI SDK, LangChain, or OpenAI SDK

Quick Start

Step 1: Provision the API Key

# Provision AI Gateway key (uses default deployment for workspace)
curl -X POST "$API_URL/api/data/aigateway/provision" \
  -H "Authorization: Bearer $AUTH_TOKEN" \
  -H "Content-Type: application/json"

Response:

{
  "deployId": "myapp-f33ad2defb",
  "apiKey": "aig_myapp-f33ad2defb_a1b2c3d4e5f6...",
  "baseUrl": "https://api.rebyte.ai/api/ai",
  "isNew": true
}

Step 2: Deploy Your App

node /home/user/.skills/rebyteai-rebyte-app-builder/bin/rebyte.js deploy

That's it! The REBYTE_AI_GATEWAY_KEY and REBYTE_AI_GATEWAY_URL environment variables are automatically injected into your Lambda function during deployment.

---

Using the AI Gateway in Your Code

With Vercel AI SDK

// app/api/chat/route.ts
import { openai } from '@ai-sdk/openai';
import { streamText } from 'ai';

const client = openai.provider({
  baseURL: process.env.REBYTE_AI_GATEWAY_URL,
  apiKey: process.env.REBYTE_AI_GATEWAY_KEY,
});

export async function POST(req: Request) {
  const { messages } = await req.json();

  const result = streamText({
    model: client('gemini-3-flash'), // or gpt-4o, claude-sonnet-4.5
    messages,
  });

  return result.toDataStreamResponse();
}

With OpenAI SDK

// functions/chat.func/index.js
const OpenAI = require('openai');

const openai = new OpenAI({
  apiKey: process.env.REBYTE_AI_GATEWAY_KEY,
  baseURL: process.env.REBYTE_AI_GATEWAY_URL,
});

exports.handler = async (event) => {
  const { messages } = JSON.parse(event.body || '{}');

  const response = await openai.chat.completions.create({
    model: 'gemini-3-flash',
    messages,
  });

  return {
    statusCode: 200,
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify(response.choices[0].message),
  };
};

With Raw Fetch

// functions/api.func/index.js
exports.handler = async (event) => {
  const { messages } = JSON.parse(event.body || '{}');

  const response = await fetch(`${process.env.REBYTE_AI_GATEWAY_URL}/chat/completions`, {
    method: 'POST',
    headers: {
      'Authorization': `Bearer ${process.env.REBYTE_AI_GATEWAY_KEY}`,
      'Content-Type': 'application/json',
    },
    body: JSON.stringify({
      model: 'gemini-3-flash',
      messages,
      stream: false,
    }),
  });

  const data = await response.json();

  return {
    statusCode: 200,
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify(data.choices[0].message),
  };
};

---

Available Models

Model	Provider	Best For
gemini-3-flash	Google	Fast responses, cost-effective
gemini-3-pro	Google	Complex reasoning
gpt-4o	OpenAI	General purpose, vision
gpt-5.2-codex	OpenAI	Code generation
claude-sonnet-4.5	Anthropic	Long context, analysis
claude-opus-4.5	Anthropic	Most capable

---

API Endpoints

The AI Gateway provides OpenAI-compatible endpoints:

Endpoint	Description
POST /v1/chat/completions	Chat completions (streaming and non-streaming)
POST /v1/responses	Responses API (for Codex models)
GET /v1/models	List available models

Base URL: https://api.rebyte.ai/api/ai

---

API Operations

Provision Key

# Create or get existing key for your workspace
curl -X POST "$API_URL/api/data/aigateway/provision" \
  -H "Authorization: Bearer $AUTH_TOKEN" \
  -H "Content-Type: application/json"

Get Key Info

# Check if key exists (doesn't create new one)
curl -X POST "$API_URL/api/data/aigateway/info" \
  -H "Authorization: Bearer $AUTH_TOKEN" \
  -H "Content-Type: application/json"

Revoke Key

# Revoke key (call provision to generate new one)
curl -X POST "$API_URL/api/data/aigateway/revoke" \
  -H "Authorization: Bearer $AUTH_TOKEN" \
  -H "Content-Type: application/json"

---

Environment Variables

When you provision an AI Gateway key and then deploy a Lambda function, these REBYTE_* environment variables are automatically injected:

Variable	Description
REBYTE_AI_GATEWAY_KEY	Your AI Gateway API key (aig_...)
REBYTE_AI_GATEWAY_URL	The base URL (https://api.rebyte.ai/api/ai/v1)

These are platform-managed variables. You do not need to set them manually.

---

Security: Server-Side Only

CRITICAL: The AI Gateway key must ONLY be used in server-side code.

DO: Use in Lambda/Server Functions

// functions/api.func/index.js - SERVER-SIDE, SAFE
exports.handler = async (event) => {
  const response = await fetch(`${process.env.REBYTE_AI_GATEWAY_URL}/chat/completions`, {
    headers: { 'Authorization': `Bearer ${process.env.REBYTE_AI_GATEWAY_KEY}` },
    // ...
  });
};

DON'T: Expose in Frontend Code

// src/App.tsx - CLIENT-SIDE, DANGEROUS
const response = await fetch('https://api.rebyte.ai/api/ai/v1/chat/completions', {
  headers: { 'Authorization': 'Bearer aig_...' }  // KEY EXPOSED!
});

Why: Frontend code is visible to users. Anyone can copy the key from browser DevTools.

Note: The AI Gateway does NOT allow cross-origin requests (no CORS), so frontend calls will fail anyway. This is intentional security.

---

Billing

• Usage is automatically tracked and billed to your organization's credits
• View usage in the Rebyte dashboard under Billing
• Each API call is logged with model, tokens, and cost

---

Complete Example: AI Chat App

1. Create the App

# Create Next.js app with AI
npx create-next-app@latest my-ai-chat --typescript --tailwind
cd my-ai-chat

# Install dependencies
npm install ai @ai-sdk/openai
npm install -D @opennextjs/aws

2. Create API Route

// app/api/chat/route.ts
import { openai } from '@ai-sdk/openai';
import { streamText } from 'ai';

const client = openai.provider({
  baseURL: process.env.REBYTE_AI_GATEWAY_URL,
  apiKey: process.env.REBYTE_AI_GATEWAY_KEY,
});

export async function POST(req: Request) {
  const { messages } = await req.json();

  const result = streamText({
    model: client('gemini-3-flash'),
    messages,
  });

  return result.toDataStreamResponse();
}

3. Create Chat UI

// app/page.tsx
'use client';
import { useChat } from 'ai/react';

export default function Chat() {
  const { messages, input, handleInputChange, handleSubmit } = useChat();

  return (
    <div className="flex flex-col h-screen p-4">
      <div className="flex-1 overflow-auto">
        {messages.map(m => (
          <div key={m.id} className="mb-4">
            <strong>{m.role}:</strong> {m.content}
          </div>
        ))}
      </div>
      <form onSubmit={handleSubmit} className="flex gap-2">
        <input
          value={input}
          onChange={handleInputChange}
          className="flex-1 border p-2 rounded"
          placeholder="Say something..."
        />
        <button type="submit" className="bg-blue-500 text-white px-4 rounded">
          Send
        </button>
      </form>
    </div>
  );
}

4. Build and Deploy

# Build with OpenNext
npx @opennextjs/aws build

# Create .rebyte/ directory
mkdir -p .rebyte/static .rebyte/functions/default.func
cp -r .open-next/assets/* .rebyte/static/
cp -r .open-next/server-functions/default/* .rebyte/functions/default.func/

cat > .rebyte/config.json << 'EOF'
{
  "version": 1,
  "routes": [
    { "handle": "filesystem" },
    { "src": "^/(.*)$", "dest": "/functions/default" }
  ]
}
EOF

# Deploy
node /home/user/.skills/rebyteai-rebyte-app-builder/bin/rebyte.js deploy

5. Provision AI Gateway

# Provision the key
curl -X POST "$API_URL/api/data/aigateway/provision" \
  -H "Authorization: Bearer $AUTH_TOKEN"

# Redeploy to pick up the auto-injected REBYTE_* env vars
node /home/user/.skills/rebyteai-rebyte-app-builder/bin/rebyte.js deploy

Your AI chat app is now live at https://<deploy-id>.rebyte.pro!

---

Troubleshooting

Issue	Cause	Fix
"Invalid API key"	Key revoked or wrong	Run aigateway/provision again
CORS error in browser	Calling from frontend	Move API call to server function
401 Unauthorized	Missing/wrong env var	Provision key, then redeploy
"Model not found"	Invalid model name	Check available models list