AIGNE Hub

The AIGNE Hub provides a unified proxy layer for accessing a variety of Large Language Models (LLMs), image, and video generation services from multiple providers. By using the @aigne/aigne-hub package, you can seamlessly switch between different AI models without altering your client-side application logic, directing all requests through a single, consistent API endpoint.

This guide covers the installation, configuration, and usage of the AIGNEHubChatModel, AIGNEHubImageModel, and AIGNEHubVideoModel classes to connect your application to the AIGNE Hub.

Overview#

AIGNE Hub acts as a gateway, aggregating major AI providers such as OpenAI, Anthropic, Google, and others. This architecture simplifies integration by abstracting the specific requirements of each provider's API. You interact with any supported model by simply passing its unique identifier, which includes the provider prefix (e.g., openai/gpt-4o-mini or anthropic/claude-3-sonnet).

Key Features#

  • Unified Access: A single endpoint for all LLM, image, and video generation requests.
  • Multi-Provider Support: Access models from OpenAI, Anthropic, AWS Bedrock, Google, DeepSeek, Ollama, xAI, and OpenRouter.
  • Secure Authentication: Manage access through a single API key (apiKey).
  • Chat, Image, and Video Models: Supports chat completions, image generation, and video creation.
  • Streaming: Real-time, token-level streaming for chat responses.
  • Seamless Integration: Designed to work with the broader AIGNE Framework.

Supported Providers#

AIGNE Hub supports a wide range of AI providers through its unified API.

Provider

Identifier

OpenAI

openai

Anthropic

anthropic

AWS Bedrock

bedrock

DeepSeek

deepseek

Google

google

Ollama

ollama

OpenRouter

openRouter

xAI

xai

Installation#

To get started, install the @aigne/aigne-hub and @aigne/core packages in your project.

npm install

npm install @aigne/aigne-hub @aigne/core

yarn add

yarn add @aigne/aigne-hub @aigne/core

pnpm add

pnpm add @aigne/aigne-hub @aigne/core

Configuration#

The chat, image, and video models require configuration to connect to your AIGNE Hub instance. The primary options include the Hub's URL, an access key, and the desired model identifier.

Model Configuration#

The configuration options are consistent across AIGNEHubChatModel, AIGNEHubImageModel, and AIGNEHubVideoModel.

baseUrl
string
required

The base URL of your AIGNE Hub instance (e.g., https://your-aigne-hub-instance/ai-kit).

apiKey
string
required

Your API access key for authenticating with the AIGNE Hub.

model
string
required

The model identifier, prefixed with the provider (e.g., openai/gpt-4o-mini).

modelOptions
object

Optional model-specific parameters to pass through to the provider's API.

Usage#

Chat Completions#

To perform a chat completion, instantiate AIGNEHubChatModel with your configuration and call the invoke method.

Basic Chat Completion

import { AIGNEHubChatModel } from "@aigne/aigne-hub";

const model = new AIGNEHubChatModel({
  baseUrl: "https://your-aigne-hub-instance/ai-kit",
  apiKey: "your-access-key-secret",
  model: "openai/gpt-4o-mini",
});

const result = await model.invoke({
  messages: [{ role: "user", content: "Hello, world!" }],
});

console.log(result);

Example Response

{
  "text": "Hello! How can I help you today?",
  "model": "openai/gpt-4o-mini",
  "usage": {
    "inputTokens": 8,
    "outputTokens": 9
  }
}

Example Models:

  • openai/gpt-4o-mini
  • anthropic/claude-3-sonnet
  • google/gemini-pro
  • xai/grok-1
  • openRouter/mistralai/mistral-7b-instruct
  • ollama/llama3

Streaming Chat Responses#

For real-time responses, set the streaming option to true in the invoke call. This returns an async iterator that yields response chunks as they become available.

Streaming Example

import { AIGNEHubChatModel } from "@aigne/aigne-hub";
import { isAgentResponseDelta } from "@aigne/core";

const model = new AIGNEHubChatModel({
  baseUrl: "https://your-aigne-hub-instance/ai-kit",
  apiKey: "your-access-key-secret",
  model: "openai/gpt-4o-mini",
});

const stream = await model.invoke(
  {
    messages: [{ role: "user", content: "Hello, who are you?" }],
  },
  { streaming: true },
);

let fullText = "";
for await (const chunk of stream) {
  if (isAgentResponseDelta(chunk)) {
    const text = chunk.delta.text?.text;
    if (text) {
      fullText += text;
      process.stdout.write(text);
    }
  }

See all 4 lines

Image Generation#

AIGNE Hub supports image generation from multiple providers. Instantiate AIGNEHubImageModel and provide a prompt and model-specific parameters.

OpenAI DALL-E#

Generate with DALL-E 3

import { AIGNEHubImageModel } from "@aigne/aigne-hub";

const model = new AIGNEHubImageModel({
  baseUrl: "https://your-aigne-hub-instance/ai-kit",
  apiKey: "your-access-key-secret",
  model: "openai/dall-e-3",
});

const result = await model.invoke({
  prompt: "A futuristic cityscape with flying cars and neon lights",
  n: 1,
  size: "1024x1024",
  quality: "standard",
  style: "natural",
});

console.log(result.images[0].url);

Google Gemini Imagen#

Generate with Imagen

import { AIGNEHubImageModel } from "@aigne/aigne-hub";

const model = new AIGNEHubImageModel({
  baseUrl: "https://your-aigne-hub-instance/ai-kit",
  apiKey: "your-access-key-secret",
  model: "google/imagen-4.0-generate-001",
});

const result = await model.invoke({
  prompt: "A serene mountain landscape at sunset",
  n: 1,
  aspectRatio: "1:1",
});

console.log(result.images[0].base64); // Note: Gemini models return base64 data

Ideogram#

Generate with Ideogram

import { AIGNEHubImageModel } from "@aigne/aigne-hub";

const model = new AIGNEHubImageModel({
  baseUrl: "https://your-aigne-hub-instance/ai-kit",
  apiKey: "your-access-key-secret",
  model: "ideogram/ideogram-v3",
});

const result = await model.invoke({
  prompt: "A cyberpunk character with glowing blue eyes, cinematic style",
  aspectRatio: "1:1",
  styleType: "cinematic",
});

console.log(result.images[0].url);

Video Generation#

AIGNE Hub extends its unified API to AI-powered video generation from leading providers. To create a video, instantiate AIGNEHubVideoModel with the appropriate configuration.

OpenAI Sora#

Generate with Sora

import { AIGNEHubVideoModel } from "@aigne/aigne-hub";

const model = new AIGNEHubVideoModel({
  baseUrl: "https://your-aigne-hub-instance/ai-kit",
  apiKey: "your-access-key-secret",
  model: "openai/sora-2",
});

const result = await model.invoke({
  prompt: "A serene beach scene with gentle waves at sunset",
  size: "1280x720",
  seconds: "8",
  outputFileType: "url",
});

console.log(result);

Example Response

{
  "videos": [{ "url": "https://...", "type": "url" }],
  "usage": {
    "inputTokens": 0,
    "outputTokens": 0,
    "aigneHubCredits": 200
  },
  "model": "openai/sora-2",
  "seconds": 8
}

Google Gemini Veo#

Generate with Veo

import { AIGNEHubVideoModel } from "@aigne/aigne-hub";

const model = new AIGNEHubVideoModel({
  baseUrl: "https://your-aigne-hub-instance/ai-kit",
  apiKey: "your-access-key-secret",
  model: "google/veo-3.1-generate-preview",
});

const result = await model.invoke({
  prompt: "A majestic eagle soaring through mountain valleys",
  aspectRatio: "16:9",
  size: "1080p",
  seconds: "6",
  outputFileType: "url",
});

console.log(result);

Example Response

{
  "videos": [{ "url": "https://...", "type": "url" }],
  "usage": {
    "inputTokens": 0,
    "outputTokens": 0,
    "aigneHubCredits": 150
  },
  "model": "google/veo-3.1-generate-preview",
  "seconds": 6
}

Summary#

The @aigne/aigne-hub package simplifies multi-provider LLM integration by offering a unified client for the AIGNE Hub service. By abstracting provider-specific logic for chat, image, and video models, it enables developers to build more flexible and maintainable AI-powered applications.

For more detailed information on specific models and their capabilities, please refer to the documentation provided by each respective AI provider. To explore other model integrations, see the Models Overview.

LMStudio