AI Service Hub

A unified TypeScript library for interacting with multiple AI services through a single, consistent interface.

Features

• 🚀 Single API - Access multiple AI services through one consistent interface
• 🔌 Multiple Providers - Support for OpenAI, Claude, Groq, Gemini, DeepSeek, Perplexity, Ollama, and LM Studio
• 🖼️ Vision Support - Process images with AI models that support vision capabilities
• 📊 Embeddings - Generate embeddings for semantic search and similarity applications
• 🔄 Auto-Detection - Automatically select the appropriate service based on the model
• 💬 Chat Interfaces - Easy-to-use chat completion APIs with system prompts
• 📝 Format Flexibility - Support for text and JSON response formats

Installation

npm install ai-service-hub

Updating Model Lists

The repository includes a helper script for refreshing the available models directly from each provider. Set the necessary API keys as environment variables and run:

npm run update-models

This will regenerate the Model*.ts files under src/ with the latest model IDs returned by the providers.

Supported Services

| Service | Chat | Vision | Embeddings | Description | | ---------- | ---- | ------ | ---------- | ------------------------------------ | | OpenAI | ✅ | ✅ | ✅ | GPT-4o, GPT-4.5-Preview, O1, etc. | | Claude | ✅ | ✅ | ❌ | Claude 3.7, 3.5 Sonnet & Haiku, Opus | | Groq | ✅ | ✅ | ✅ | Grok-2, Grok-2-Vision, Grok-Beta | | Gemini | ✅ | ❌ | ✅ | Gemini 2.5, 2.0, 1.5, and Imagen 3.0 | | DeepSeek | ✅ | ❌ | ❌ | DeepSeek Chat, Reasoner | | Perplexity | ✅ | ❌ | ❌ | Sonar models, R1-1776 | | Ollama | ✅ | ✅ | ✅ | Self-hosted open source models | | LM Studio | ✅ | ✅ | ✅ | Self-hosted Hugging Face models |

Current Model Support

OpenAI Models

• GPT-4.5 Preview
• GPT-4o Mini
• GPT-4o
• GPT-3.5 Turbo
• GPT-4
• GPT-4 Turbo
• GPT-4 Turbo Preview
• GPT-4 Vision Preview
• O1
• O1 Mini
• O1 Preview
• O3 Mini
• ChatGPT-4o Latest
• ChatGPT-4o Mini

Claude Models

• Claude 3.7 Sonnet Latest
• Claude 3.5 Haiku Latest
• Claude 3.5 Sonnet Latest
• Claude 3.5 Sonnet (20240620)
• Claude 3 Opus Latest
• Claude 3 Sonnet (20240229)
• Claude 3 Haiku (20240307)

Gemini Models

• Gemini 2.5 Pro Experimental (03/25)
• Gemini 2.0 Flash
• Gemini 2.0 Flash Lite
• Gemini 1.5 Flash
• Gemini 1.5 Flash 8B
• Gemini 1.5 Pro
• Gemini Embedding Experimental
• Imagen 3.0 Generate 002

DeepSeek Models

• DeepSeek Chat
• DeepSeek Reasoner

Perplexity Models

• Sonar Deep Research
• Sonar Reasoning Pro
• Sonar Reasoning
• Sonar Pro
• Sonar
• R1-1776

Basic Usage

import { GlobalInstance, ModelOpenAi } from 'ai-service-hub';

// Initialize with your API keys
const ai = new GlobalInstance({
  openAiKey: 'your-openai-key',
  claudeKey: 'your-claude-key',
  grokKey: 'your-groq-key',
  // Add other service keys as needed
});

// Basic chat completion
async function getCompletion() {
  const response = await ai.chat({
    prompt: "What's the capital of France?",
    model: ModelOpenAi.gpt4o,
    systemPrompt: "You are a helpful assistant."
  });
  
  console.log(response);
}

Examples

OpenAI Chat

import { OpenAiInstance, ModelOpenAi } from 'ai-service-hub';

const openai = new OpenAiInstance('your-api-key');

const response = await openai.chat(
  "Explain quantum computing in simple terms", 
  "You are a helpful assistant", 
  ModelOpenAi.gpt4o
);

Claude Chat

import { ClaudeInstance, ModelClaude } from 'ai-service-hub';

const claude = new ClaudeInstance('your-api-key');

const response = await claude.chat(
  "Write a short story about a robot learning to paint", 
  "You are a creative writer", 
  ModelClaude.claude37SonnetLatest
);

Vision API (OpenAI)

import { OpenAiInstance, ModelOpenAIVision } from 'ai-service-hub';

const openai = new OpenAiInstance('your-api-key');

// Convert your image to base64 first
const response = await openai.vision(
  "What's in this image?", 
  base64Image, 
  "Describe the image in detail",
  ModelOpenAIVision.gpt4o
);

Embeddings (OpenAI)

import { OpenAiInstance, ModelOpenAiEmbedding } from 'ai-service-hub';

const openai = new OpenAiInstance('your-api-key');

const embedding = await openai.embedding(
  "This is a sample text to embed", 
  ModelOpenAiEmbedding.textEmbedding3Large
);

Using Global Instance

import { GlobalInstance, ModelOpenAi, ModelClaude } from 'ai-service-hub';

const ai = new GlobalInstance({
  openAiKey: 'your-openai-key',
  claudeKey: 'your-claude-key'
});

// OpenAI will be automatically selected based on the model
const openaiResponse = await ai.chat({
  prompt: "Explain how rainbows form",
  model: ModelOpenAi.gpt4o
});

// Claude will be automatically selected based on the model
const claudeResponse = await ai.chat({
  prompt: "Write a poem about autumn",
  model: ModelClaude.claude37SonnetLatest
});

Configuration

Each service requires different configuration parameters:

const globalInstance = new GlobalInstance({
  openAiKey: 'your-openai-key',
  claudeKey: 'your-anthropic-key',
  grokKey: 'your-groq-key',
  geminiKey: 'your-gemini-key',
  deepSeekKey: 'your-deepseek-key',
  perplexityKey: 'your-perplexity-key',
  ollamaUrl: 'http://localhost:11434',
  lmstudioUrl: 'http://localhost:1234'
});

Copy the provided .env.example file to .env and fill in your API keys to run the library locally.

Recent Updates

• Updated to support newest models including Claude 3.7 Sonnet and Gemini 2.5
• Added comprehensive model definitions for all providers
• Added Claude AI integration with chat and vision capabilities
• Converted all enums to modern TypeScript object literals
• Added ESLint and Prettier for code quality
• Enhanced vision capabilities across multiple providers
• Improved type safety throughout the codebase

License

MIT

Author

Bartłomiej Zimny