AI Model Client SDK

A unified TypeScript/JavaScript SDK for interacting with multiple AI model providers, including OpenAI, Anthropic, Cohere, Gemini, Mistral, DeepSeek, Llama, and XAI. This SDK provides a consistent interface for generating text and working with various AI models, making it easy to integrate state-of-the-art language models into your applications.

Features

• Unified API for multiple AI providers
• Easy switching between models and providers
• TypeScript support with rich typings
• Extensible and customizable
• Supports custom system prompts, temperature, max tokens, and more

Installation

npm install @craftily/ai-sdk
# or
yarn add @craftily/ai-sdk

Usage: General AIClient

Note: The value of response returned by client.chat(...) depends on the provider. It may be a string, or it may be an object (e.g., with a .text property or other fields).

import AIClient from '@craftily/ai-sdk';

const client = new AIClient({
  modelId: 'gpt-4.1', // or any supported modelId
  provider: 'openai', // "openai", "google", "anthropic", "cohere", "meta", "mistral", "xai", "deepseek", "custom",
  apiKey: 'YOUR_API_KEY',
  maxTokens: 256,      // optional
  temperature: 0.7,    // optional
});

const response = await client.chat("What are the top tourist attractions in Tokyo, Japan?");

// Handle both string and object responses
if (typeof response === 'string') {
  console.log(response);
} else if (response && typeof response === 'object' && 'text' in response) {
  console.log(response.text);
} else {
  console.log(response); // fallback for other shapes
}

Usage Individual Clients

import { OpenAIClient, AnthropicClient, CohereClient /*, ... */ } from '@craftily/ai-sdk';
import { AIRequest } from '@craftily/ai-sdk';

// Example: Using OpenAIClient
const client = new OpenAIClient({ apiKey: 'YOUR_OPENAI_API_KEY' });

const request: AIRequest = {
  prompt: "What are the top tourist attractions in Tokyo, Japan?",
  modelId: "gpt-4.1", // or any supported model
  temperature: 0.7,
  maxTokens: 256,
};

const response = await client.generate(request);

// Handle both string and object responses
if (typeof response === 'string') {
  console.log(response);
} else if (response && typeof response === 'object' && 'text' in response) {
  console.log(response.text);
} else {
  console.log(response); // fallback for other shapes
}

This approach allows you to switch between providers and models dynamically by changing the modelId and provider in the configuration.

Supported Providers & Models

• OpenAI: GPT-4, GPT-3.5, etc.
• Anthropic: Claude models
• Cohere: Command models
• Google Gemini: Gemini models
• Mistral: Mistral models
• DeepSeek: DeepSeek models
• Llama: Llama models
• XAI: XAI models

API Reference

AIRequest

| Field | Type | Description | | ------------- | -------- | ---------------------------------------------- | | prompt | string | The prompt/question for the model | | modelId | string | The model identifier (see supported models) | | temperature | number | (Optional) Sampling temperature | | maxTokens | number | (Optional) Maximum tokens in the response |

Client Classes

Each provider has its own client class (e.g., OpenAIClient, AnthropicClient, etc.), all exposing a .generate(request: AIRequest) method that returns an AIResponse.

AIResponse

| Field | Type | Description | | ------------- | -------- | ----------------------------- | | text | string | The generated text | | ... | ... | Other provider-specific fields |

Contributing

1. Fork the repository
2. Create a new branch
3. Make your changes
4. Submit a pull request

License

MIT