feat: add version flag and release packaging workflow

.github/workflows/release.yml

@@ -0,0 +1,77 @@
+name: Release
+
+# Cuts a GitHub release when a v* tag is pushed. The release artifact is the
+# xcodebuild Release product: the mlx-server executable plus its colocated
+# resource bundles (mlx-swift_Cmlx.bundle carries the Metal default.metallib).
+# A lone binary fails at runtime with "Failed to load the default metallib",
+# so the tarball must keep the bundles next to the executable.
+
+on:
+  push:
+    tags:
+      - "v*"
+
+permissions:
+  contents: write
+
+jobs:
+  release:
+    runs-on: macos-15
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Select Xcode
+        uses: maxim-lobanov/setup-xcode@v1
+        with:
+          xcode-version: latest-stable
+
+      - name: Derive version
+        id: version
+        run: echo "version=${GITHUB_REF_NAME#v}" >> "$GITHUB_OUTPUT"
+
+      # mlx-swift's Metal shaders + the MLX C++ sources are expensive to
+      # compile; cache the SwiftPM and Xcode build products across runs.
+      - name: Cache build
+        uses: actions/cache@v4
+        with:
+          path: |
+            .build
+            ~/Library/Caches/org.swift.swiftpm
+          key: ${{ runner.os }}-release-${{ hashFiles('Package.swift') }}
+          restore-keys: ${{ runner.os }}-release-
+
+      # xcodebuild (not `swift build`) compiles the Metal shaders and copies
+      # the resource bundles next to the binary. -skipMacroValidation is
+      # required for the MLXHuggingFace macro plugin in non-interactive runs.
+      - name: Build (xcodebuild, Release, compiles Metal shaders)
+        run: |
+          xcodebuild -scheme mlx-server \
+            -destination 'platform=macOS' \
+            -configuration Release \
+            -derivedDataPath .build/xcode \
+            -skipMacroValidation \
+            build
+
+      - name: Package artifact
+        id: package
+        run: |
+          set -euo pipefail
+          VERSION="${{ steps.version.outputs.version }}"
+          PRODUCTS=.build/xcode/Build/Products/Release
+          STAGE="mlx-server-${VERSION}-macos-arm64"
+          mkdir -p "dist/${STAGE}"
+          cp "${PRODUCTS}/mlx-server" "dist/${STAGE}/"
+          cp -R "${PRODUCTS}"/*.bundle "dist/${STAGE}/"
+          tar -C dist -czf "dist/${STAGE}.tar.gz" "${STAGE}"
+          ( cd dist && shasum -a 256 "${STAGE}.tar.gz" | tee "${STAGE}.tar.gz.sha256" )
+          echo "tarball=dist/${STAGE}.tar.gz" >> "$GITHUB_OUTPUT"
+
+      - name: Create release
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: |
+          gh release create "${GITHUB_REF_NAME}" \
+            --title "${GITHUB_REF_NAME}" \
+            --generate-notes \
+            "${{ steps.package.outputs.tarball }}" \
+            "${{ steps.package.outputs.tarball }}.sha256"

README.md

@@ -22,7 +22,7 @@ The end goal is to be a drop-in replacement for `llama-server` in [LLMKube](http
 - Tool calling and structured outputs
 - Vision-language model support (Qwen-VL, Gemma 4 VL, etc.)
 - Multi-slot concurrency with longest-prefix KV cache reuse
-- Single static binary distribution via Homebrew + GitHub releases
+- One-command install via Homebrew, plus prebuilt GitHub releases
 
 ## Non-goals
 

Sources/MLXServer/MLXServerCommand.swift

@@ -5,7 +5,8 @@ import MLXServerKit
 struct MLXServerCommand: AsyncParsableCommand {
     static let configuration = CommandConfiguration(
         commandName: "mlx-server",
-        abstract: "OpenAI-compatible HTTP server for mlx-swift-lm on Apple Silicon."
+        abstract: "OpenAI-compatible HTTP server for mlx-swift-lm on Apple Silicon.",
+        version: "0.1.0"
     )
 
     @Option(name: .long, help: "Model identifier (HuggingFace ID or local directory path).")