This Swift library provides clients for interacting with various AI providers, including OpenAI, Gemini, Anthropic, and others. It aims to simplify AI API integration into iOS and macOS applications, offering a unified interface for making requests. The library supports both direct API calls and proxied requests through a secure backend for API key protection.

How It Works

The library abstracts the complexities of different AI provider APIs into a consistent Swift interface. Users can choose to initialize services directly with their own API keys (BYOK) or use a proxied approach. The proxied approach leverages a backend service that enhances security by implementing certificate pinning, DeviceCheck verification, split key encryption, and per-user/per-IP rate limiting, protecting API keys and managing costs.

Quick Start & Requirements

• Installation: Add the package via Xcode's Package Dependencies by entering github.com/lzell/aiproxyswift.
• Configuration: Call AIProxy.configure() during app launch.
• Dependencies: Requires Xcode and Swift. Specific AI provider models may have their own requirements.
• Documentation: See AIProxy integration video and AIProxyIdentifier.swift for details.

Highlighted Details

• Supports a wide range of AI providers: OpenAI, Gemini, Anthropic, Stability AI, DeepL, Together AI, Replicate, ElevenLabs, Fal, Groq, Perplexity, Mistral, EachAI, OpenRouter, DeepSeek, Fireworks AI, Brave.
• Offers robust API key security through a proxied backend with multiple layers of protection.
• Provides extensive examples for various AI provider features, including chat completions, streaming, image generation, transcription, tool use, and multimodal inputs.
• Includes advanced settings for custom client IDs and detailed error handling for Foundation network errors.

Maintenance & Community

The project is actively maintained, with contributions welcomed. Details on contributing and community channels are available in the README.

Licensing & Compatibility

The library is released under the MIT license, allowing for commercial use and integration with closed-source applications.

Limitations & Caveats

The README mentions that some advanced features, like specific Replicate model interactions or tool calls with Llama/TogetherAI, might require further development or have specific limitations. Users should consult the detailed examples for each provider for precise usage and potential caveats.