fix(release): resolve musl C++ compilation by using HuggingFace-only build

- musl builds now use --no-default-features --features huggingface
- Avoids llama.cpp C++ compilation issues with musl toolchain
- Provides lightweight, statically-linked Linux binary
- Other platforms still get full functionality including llama.cpp

This fixes the missing x86_64-linux-musl-g++ compiler error.

Signed-off-by: Michael A. Kuykendall <michaelallenkuykendall@gmail.com>

Author: Michael-A-Kuykendall

.github/workflows/release.yml

@@ -179,10 +179,18 @@ jobs:
 
       - name: Install musl-tools (for musl builds)
         if: matrix.target == 'x86_64-unknown-linux-musl'
-        run: sudo apt-get update && sudo apt-get install -y musl-tools
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y musl-tools musl-dev
 
       - name: Build binary
-        run: cargo build --release --target ${{ matrix.target }}
+        run: |
+          if [[ "${{ matrix.target }}" == "x86_64-unknown-linux-musl" ]]; then
+            # For musl builds, use huggingface-only to avoid llama.cpp C++ compilation issues
+            cargo build --release --target ${{ matrix.target }} --no-default-features --features huggingface
+          else
+            cargo build --release --target ${{ matrix.target }}
+          fi
 
       - name: Upload artifact
         uses: actions/upload-artifact@v4