LLM Adapter Kit

A universal TypeScript library for interacting with multiple LLM providers through a unified interface. Supports 600+ models across 8 major providers with built-in cost calculation, streaming, and advanced features.

🚀 Features

• Universal Interface: Single API for all LLM providers
• 8 Major Providers: OpenAI, Anthropic, Google, Mistral, Groq, OpenRouter, Requesty, Perplexity
• 600+ Models: Access to the latest models including GPT-4o, Claude 4, Gemini 2.5, Sonar
• Cost Tracking: Built-in token counting and cost calculation
• Streaming Support: Real-time response streaming
• Advanced Features: Function calling, JSON mode, vision support
• TypeScript First: Full type safety and IntelliSense support
• Comprehensive Testing: Extensive test suite with real API validation

📦 Installation

npm install llm-adapter-kit

🔑 Setup

1. Copy the environment file:

cp .env.example .env

2. Add your API keys to .env:

# Required: At least one provider
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=AIza...
MISTRAL_API_KEY=...

# Optional: Multi-provider aggregators
OPENROUTER_API_KEY=sk-or-...  # 400+ models
REQUESTY_API_KEY=...          # 150+ models
PERPLEXITY_API_KEY=pplx-...   # Web search & reasoning

🎯 Quick Start

import { OpenAIAdapter, AnthropicAdapter, GoogleAdapter, GroqAdapter, PerplexityAdapter } from 'llm-adapter-kit';

// Initialize adapters
const openai = new OpenAIAdapter();
const claude = new AnthropicAdapter();
const gemini = new GoogleAdapter();
const groq = new GroqAdapter();
const perplexity = new PerplexityAdapter();

// Basic text generation
const response = await openai.generate('Write a haiku about coding');
console.log(response.text);

// With options
const jsonResponse = await claude.generate('Return user data as JSON', {
  maxTokens: 100,
  temperature: 0.7,
  jsonMode: true
});

// Streaming
await gemini.generateStream('Count to 10', {
  onToken: (token) => process.stdout.write(token),
  onComplete: (response) => console.log('\\nDone!')
});

🌟 Supported Providers

Provider	Models	Key Features
OpenAI	GPT-4o, GPT-4o-mini, o1	Function calling, vision, reasoning
Anthropic	Claude 4, Claude 3.5	Computer use, thinking, large context
Google	Gemini 2.5, Gemini Pro	Multimodal, thinking mode, long context
Mistral	Large, Medium, Codestral	Code generation, OCR, agents API
Groq	Llama, Gemma, Mixtral	Ultra-fast inference, 750+ tokens/sec
OpenRouter	400+ models	Multi-provider access, cost optimization
Requesty	150+ models	Premium model access, high availability
Perplexity	Sonar, Reasoning, Research	Real-time web search, citations, reasoning

📝 Usage Examples

Provider-Specific Features

// OpenAI function calling
const openai = new OpenAIAdapter();
const result = await openai.generate('What\'s the weather?', {
  tools: [{
    type: 'function',
    function: {
      name: 'get_weather',
      description: 'Get weather data',
      parameters: { /* ... */ }
    }
  }]
});

// Anthropic Claude 4 with system prompt
const claude = new AnthropicAdapter('claude-sonnet-4-20250514');
const response = await claude.generate('Analyze this code', {
  systemPrompt: 'You are a senior code reviewer',
  maxTokens: 1000
});

// Google Gemini with vision
const gemini = new GoogleAdapter();
const analysis = await gemini.generate('Describe this image', {
  model: 'gemini-2.5-flash',
  // images: [imageData] // Add image data
});

// Mistral code generation
const mistral = new MistralAdapter();
const code = await mistral.generate('Write a Python function', {
  model: 'codestral-25.01',
  maxTokens: 500
});

// Groq ultra-fast inference
const groq = new GroqAdapter();
const fastResponse = await groq.generate('Explain quantum computing', {
  model: 'llama-3.1-70b-versatile',
  maxTokens: 200
});

// Perplexity web search with citations
const perplexity = new PerplexityAdapter();
const searchResult = await perplexity.generate('What are the latest developments in AI?', {
  model: 'sonar-pro',
  maxTokens: 300
});
console.log(searchResult.citations); // Real-time web sources

Multi-Provider Access

// OpenRouter - access 400+ models
const router = new OpenRouterAdapter();

// Use different providers through OpenRouter
const gptResponse = await router.generate('Hello', { 
  model: 'openai/gpt-4o' 
});
const claudeResponse = await router.generate('Hello', { 
  model: 'anthropic/claude-3.5-sonnet' 
});
const llamaResponse = await router.generate('Hello', { 
  model: 'meta-llama/llama-3.1-70b-instruct' 
});

// Requesty - premium model access
const requesty = new RequestyAdapter();
const premium = await requesty.generate('Complex analysis', {
  model: 'gpt-4o',
  maxTokens: 2000
});

Cost Tracking

import { CostCalculator } from 'llm-adapter-kit';

const response = await openai.generate('Hello world');

// Automatic cost calculation
if (response.usage) {
  const cost = await CostCalculator.calculateCostWithTokenCounting(
    'openai',
    response.model,
    'Hello world',
    response.text,
    response.usage
  );
  
  console.log(`Cost: $${cost.totalCost.toFixed(6)}`);
  console.log(`Tokens: ${response.usage.totalTokens}`);
}

🧪 Testing

The library includes comprehensive tests for all adapters:

# Run all tests
npm test

# Test specific providers
npm test -- --testPathPattern="OpenAI"
npm test -- --testPathPattern="Anthropic"

# Test with coverage
npm test:coverage

🔧 Configuration

Model Selection

// Use specific models
const gpt4o = new OpenAIAdapter('gpt-4o');
const claude4 = new AnthropicAdapter('claude-sonnet-4-20250514');
const gemini2_5 = new GoogleAdapter('gemini-2.5-flash');

// List available models
const models = await openai.listModels();
console.log(models.map(m => m.id));

Adapter Capabilities

// Check what each adapter supports
const caps = openai.getCapabilities();
console.log({
  streaming: caps.supportsStreaming,
  json: caps.supportsJSON,
  functions: caps.supportsFunctions,
  images: caps.supportsImages,
  context: caps.maxContextWindow
});

📄 License

MIT License - see LICENSE file for details.

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Add tests for new functionality
4. Ensure all tests pass
5. Submit a pull request

📞 Support

• Issues: GitHub Issues
• Documentation: See example files in /examples
• API Reference: Generated TypeScript definitions

---

Made with ❤️ by Synaptic Labs