feat: update mlx-swift-lm to SharpAI fork main branch for Qwen3.5 support

Author: simba

Package.resolved

@@ -5,12 +5,13 @@ let package = Package(
     name: "mlx-server",
     platforms: [.macOS(.v14)],
     dependencies: [
-        // Apple MLX Swift — core inference engine
-        .package(url: "https://github.com/ml-explore/mlx-swift", .upToNextMinor(from: "0.30.3")),
-        // Apple's LLM library built on MLX Swift (Qwen, Llama, Mistral, Gemma etc.)
-        .package(url: "https://github.com/ml-explore/mlx-swift-lm", from: "2.0.0"),
+        // Apple MLX Swift — core inference engine (Apple-maintained, tagged releases)
+        .package(url: "https://github.com/ml-explore/mlx-swift", .upToNextMinor(from: "0.30.6")),
+        // Apple's LLM library built on MLX Swift (SharpAI fork)
+        // Pinned to main branch for Qwen3.5 support (PRs #97, #120, #129, #133, #135 — not yet in a release tag)
+        .package(url: "https://github.com/SharpAI/mlx-swift-lm", branch: "main"),
         // HuggingFace tokenizers + model download
-        .package(url: "https://github.com/huggingface/swift-transformers", .upToNextMinor(from: "1.1.0")),
+        .package(url: "https://github.com/huggingface/swift-transformers", .upToNextMinor(from: "1.2.0")),
         // Lightweight HTTP server (Apple-backed Swift server project)
         .package(url: "https://github.com/hummingbird-project/hummingbird", from: "2.0.0"),
         // Async argument parser (for CLI flags: --model, --port)