[CI] Publish nightly sglang wheel under both cu129 and cu130 indexes (#24176)

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.github/workflows/release-pypi-nightly.yml

@@ -14,11 +14,6 @@ on:
         description: 'Specific commit SHA to build (leave empty for latest)'
         required: false
         type: string
-      cuda_version:
-        description: 'CUDA version (e.g., 129 or 130)'
-        required: false
-        default: '129'
-        type: string
 
 concurrency:
   group: release-pypi-nightly-${{ github.ref }}
@@ -112,6 +107,15 @@ jobs:
     needs: build-nightly-wheel
     runs-on: ubuntu-latest
     environment: 'prod'
+    strategy:
+      fail-fast: false
+      # The wheel is CUDA-agnostic and built once — we just register the same
+      # artifact under cu129/sglang/ and cu130/sglang/ wheel indexes so users
+      # can install via either --extra-index-url. Serialize because both matrix
+      # runs clone and push to the same sgl-whl branch.
+      max-parallel: 1
+      matrix:
+        cuda_version: ['129', '130']
     steps:
       - uses: actions/checkout@v4
 
@@ -160,12 +164,12 @@ jobs:
           python3 scripts/update_nightly_whl_index.py \
             --commit-hash ${{ needs.build-nightly-wheel.outputs.commit_hash }} \
             --nightly-version ${{ needs.build-nightly-wheel.outputs.nightly_version }} \
-            --cuda-version ${{ inputs.cuda_version || '129' }} \
+            --cuda-version ${{ matrix.cuda_version }} \
             --build-date ${{ needs.build-nightly-wheel.outputs.build_date }}
 
       - name: Push wheel index
         run: |
           cd sgl-whl
           git add -A
-          git diff --staged --quiet || git commit -m "Update nightly wheel index for commit ${{ needs.build-nightly-wheel.outputs.commit_hash }}"
+          git diff --staged --quiet || git commit -m "Update cu${{ matrix.cuda_version }} nightly wheel index for commit ${{ needs.build-nightly-wheel.outputs.commit_hash }}"
           git push

scripts/update_nightly_whl_index.py

@@ -42,8 +42,10 @@ def update_wheel_index(
     whl_repo_dir = pathlib.Path("sgl-whl")
 
     if not dist_dir.exists():
-        print(f"Warning: {dist_dir} does not exist, skipping index update")
-        return
+        raise FileNotFoundError(
+            f"{dist_dir} does not exist — the download-artifact step did not "
+            f"populate it; refusing to silently no-op the index update"
+        )
 
     # Format CUDA version with 'cu' prefix if not already present
     if not cuda_version.startswith("cu"):
@@ -87,7 +89,6 @@ def update_wheel_index(
     # Generate new links for current wheels
     new_links = []
     for wheel_path in sorted(dist_dir.glob("*.whl")):
-        try:
         filename = wheel_path.name
         sha256 = compute_sha256(wheel_path)
 
@@ -97,13 +98,12 @@ def update_wheel_index(
 
         new_links.append(link)
         print(f"  Added: {filename}")
-        except Exception as e:
-            print(f"  Error processing {wheel_path.name}: {e}")
-            continue
 
     if not new_links:
-        print("  No new wheels to add")
-        return
+        raise RuntimeError(
+            f"No wheels found in {dist_dir} — index update for {cuda_version} "
+            f"would be a no-op; failing loudly instead of pushing an empty change"
+        )
 
     # Combine existing and new links (new links first for latest)
     all_links = new_links + existing_links