fix release publish workflow failures (#1491)

(cherry picked from commit 5f51ea4)

Signed-off-by: xunzhuo <xunzhuo@vllm-semantic-router.ai>

Author: Xunzhuo

.github/workflows/publish-crate.yml

@@ -14,6 +14,8 @@ jobs:
   publish:
     if: github.repository == 'vllm-project/semantic-router'
     runs-on: ubuntu-latest
+    permissions:
+      contents: write
 
     steps:
     - name: Check out the repo
@@ -42,7 +44,7 @@ jobs:
     - name: Set up Rust
       uses: dtolnay/rust-toolchain@stable
       with:
-        toolchain: 1.85
+        toolchain: 1.90
 
     - name: Install system dependencies
       run: |
@@ -76,9 +78,13 @@ jobs:
           exit 1
         fi
 
-    - name: Run tests (CPU-only, no CUDA)
+    - name: Run API smoke tests (CPU-only, no CUDA)
       working-directory: candle-binding
-      run: cargo test --no-default-features --verbose
+      run: |
+        cargo test --no-default-features \
+          --test qwen3_classification_api_test \
+          --test classification_output_demo \
+          --verbose
 
     - name: Check crate (CPU-only, no CUDA)
       working-directory: candle-binding
@@ -90,11 +96,11 @@ jobs:
 
     - name: Dry run publish
       working-directory: candle-binding
-      run: cargo publish --dry-run --verbose
+      run: cargo publish --dry-run --no-default-features --verbose
 
     - name: Publish to crates.io
       working-directory: candle-binding
-      run: cargo publish --verbose
+      run: cargo publish --no-default-features --verbose
       env:
         CARGO_REGISTRY_TOKEN: ${{ secrets.CRATES_TOKEN }}
 

.github/workflows/pypi-publish.yml

@@ -22,7 +22,7 @@ jobs:
     if: github.repository == 'vllm-project/semantic-router'
     runs-on: ubuntu-latest
     permissions:
-      contents: read
+      contents: write
       id-token: write  # Required for trusted publishing to PyPI
 
     steps:
@@ -174,7 +174,7 @@ jobs:
           TWINE_PASSWORD: ${{ secrets.PYPI_API_TOKEN }}
         run: |
           echo "Publishing to PyPI..."
-          twine upload dist/* --verbose
+          twine upload dist/* --skip-existing --verbose
 
       - name: Create GitHub Release
         if: github.event_name == 'push' && startsWith(github.ref, 'refs/tags/')
@@ -235,4 +235,3 @@ jobs:
           if [ "${{ steps.extract_version.outputs.is_dev }}" = "false" ]; then
             echo "- [GitHub Release](https://github.com/${{ github.repository }}/releases/tag/${{ steps.extract_version.outputs.tag }})" >> $GITHUB_STEP_SUMMARY
           fi
-

candle-binding/Cargo.toml

@@ -28,13 +28,13 @@ accelerate = ["candle-core/accelerate", "candle-nn/accelerate", "candle-transfor
 
 [dependencies]
 anyhow = { version = "1", features = ["backtrace"] }
-candle-core = { git = "https://github.com/huggingface/candle", tag = "0.9.2-alpha.1" }
-candle-nn = { git = "https://github.com/huggingface/candle", tag = "0.9.2-alpha.1" }
-candle-transformers = { git = "https://github.com/huggingface/candle", tag = "0.9.2-alpha.1" }
+candle-core = { version = "0.9.2-alpha.1", git = "https://github.com/huggingface/candle", tag = "0.9.2-alpha.1" }
+candle-nn = { version = "0.9.2-alpha.1", git = "https://github.com/huggingface/candle", tag = "0.9.2-alpha.1" }
+candle-transformers = { version = "0.9.2-alpha.1", git = "https://github.com/huggingface/candle", tag = "0.9.2-alpha.1" }
 # Flash Attention 2 (optional, requires CUDA)
 # Reference: https://github.com/huggingface/candle/tree/main/candle-flash-attn
 # Using git dependency to automatically fetch CUTLASS submodule
-candle-flash-attn = { git = "https://github.com/huggingface/candle", tag = "0.9.2-alpha.1", optional = true }
+candle-flash-attn = { version = "0.9.2-alpha.1", git = "https://github.com/huggingface/candle", tag = "0.9.2-alpha.1", optional = true }
 tokenizers = { version = "0.21.0", default-features = false, features = ["http", "onig"] }
 hf-hub = { version = "0.4.1", default-features = false, features = ["rustls-tls"] }
 safetensors = "0.4.1"