apple-llm-daemon

A lightweight Swift daemon that exposes Apple’s on-device LLM through a simple local HTTP endpoint. It listens on port 5454 and accepts JSON requests with a text prompt. Built with Network and FoundationModels.

---

🧠 Overview

This project runs a minimal TCP server to bridge your applications with Apple’s native LLM (LanguageModelSession). It converts plain HTTP JSON requests into model responses, enabling easy local AI integration.

---

🚀 Features

• Uses Apple’s built-in LanguageModelSession API.
• Runs locally, no cloud or API key required.
• Lightweight single-file daemon.
• JSON-based request/response format.

---

🧩 Example Request

Send a simple POST request to the daemon:

POST / HTTP/1.1
Host: 127.0.0.1:5454
Content-Type: application/json

{"prompt": "Explain quantum computing in simple terms."}

Response:

HTTP/1.1 200 OK
Content-Type: application/json

{"text": "Quantum computing uses qubits that can represent multiple states at once..."}

---

🛠️ Build & Run

1. Clone the repository:

git clone git@github.com:qvanderlaan/apple-llm-daemon.git
cd apple-llm-daemon

2. Build and run:

swift run

3. The daemon will start at:

http://127.0.0.1:5454

---

🧪 Development Notes

• Requires macOS 26+ (Tahoe or later).
• Swift 6.2 or newer.
• Uses FoundationModels and Network frameworks.

---

📄 License

MIT License. See LICENSE for details.