This Swift-based web API exposes Apple's on-device Foundation Models through an OpenAI-compatible HTTP interface. Built with Vapor, it targets developers seeking to integrate Apple Intelligence capabilities into applications, offering a familiar API structure for streamlined adoption and leveraging powerful on-device AI without external dependencies.

How It Works

The project utilizes the Vapor web framework to create an HTTP server that interfaces with Apple's on-device Foundation Models. It translates standard API requests into calls to these local models, supporting both single-prompt completions and multi-turn chat conversations. Responses can be streamed in real-time using Server-Sent Events, mirroring the behavior of popular LLM APIs.

Quick Start & Requirements

• Requirements: An Apple Intelligence-enabled device and Swift 6.0+.
• Build: swift build
• Run: swift run AppleIntelligenceApi serve [--port, -p] (defaults to http://localhost:8080).
• Setup: Requires enabling Apple Intelligence in device settings.
• Docs: Refers to OpenAI/OpenRouter documentation for API usage details.

Highlighted Details

• OpenAI-compatible API: Works seamlessly with existing OpenAI/OpenRouter client libraries.
• Supports chat completions (multi-turn) and non-chat completions (single-prompt).
• Real-time streaming responses via Server-Sent Events.
• Offers two model variants: base (default guardrails) and permissive (relaxed filtering).

Maintenance & Community

• Tests: Currently lacks automated tests; contributions are welcomed.
• Contributing: Standard pull request workflow is outlined.
• Community: No specific community channels (e.g., Discord, Slack) are listed.

Licensing & Compatibility

• License: Not explicitly stated in the README.
• Compatibility: Designed for on-device execution on Apple Intelligence-enabled hardware. Local network deployment is supported.

Limitations & Caveats

Features such as Authentication and Tool/Function calling are not yet implemented. The project is an unofficial API wrapper and is not affiliated with Apple Inc. No automated tests are currently available.