Ascend Release Manager for vLLM

# Check if gh is installed
gh --version

# If not installed, install gh CLI:
# Ubuntu/Debian
apt install gh -y

# macOS
brew install gh

# OpenEuler
yum install gh -y

# Check authentication status
gh auth status

# If not authenticated, login with:
gh auth login

github.com
  ✓ Logged in to github.com account <username> (keyring)
  - Active account: true
  - Git operations protocol: https
  - Token: gho_****
  - Token scopes: 'gist', 'read:org', 'repo', 'workflow'

┌─────────────────────────────────────────────────────────────────────────────┐
│                         vLLM Ascend Release Process                         │
├─────────────────────────────────────────────────────────────────────────────┤
│                                                                             │
│  Phase 1: Initialization                                                    │
│  ├── Determine version & branch                                             │
│  ├── Create feedback issue                                                  │
│  └── Create release checklist issue                                         │
│                                                                             │
│  Phase 2: Bug Triage                                                        │
│  ├── Scan open bugs                                                         │
│  ├── Identify release-blocking bugs                                         │
│  └── Update checklist with bug list                                         │
│                                                                             │
│  Phase 3: PR Management                                                     │
│  ├── Identify must-merge PRs                                                │
│  └── Update checklist with PR list                                          │
│                                                                             │
│  Phase 4: Test Coverage Analysis                                            │
│  ├── Scan PRs for features/models without tests                             │
│  ├── Check previous feedback issue status                                   │
│  └── Update checklist with items needing manual testing                     │
│                                                                             │
│  Phase 5: Nightly Status                                                    │
│  ├── Get latest Nightly-A3 and Nightly-A2 runs                              │
│  ├── Analyze failures with extract_and_analyze.py                           │
│  └── Update checklist with nightly status table                             │
│                                                                             │
│  Phase 6: Release Notes (invoke existing skill)                             │
│  ├── Generate release notes via vllm-ascend-release-note-writer             │
│  └── Create release notes PR                                                │
│                                                                             │
│  Phase 7: Documentation & Artifacts                                         │
│  └── Update version references (Docker/wheel built by CI automatically)     │
│                                                                             │
│  Phase 8: Release Execution (requires human review)                         │
│  ├── Human review & approval                                                │
│  ├── Merge release notes PR                                                 │
│  ├── Create GitHub release                                                  │
│  └── Verify automated pipelines (PyPI, Docker, ReadTheDocs)                 │
│                                                                             │
│  Phase 9: WeChat Article (微信公众号推文)                                     │
│  ├── Collect release statistics (commits, contributors)                     │
│  ├── Generate WeChat article from template                                  │
│  └── Review and publish to WeChat official account                          │
│                                                                             │
└─────────────────────────────────────────────────────────────────────────────┘

# Get the latest release tag
gh release list --repo vllm-project/vllm-ascend --limit 5

# Or check existing tags
git tag --sort=-creatordate | head -10

gh issue create --repo vllm-project/vllm-ascend \
  --title "[Feedback]: v${VERSION} Release Feedback" \
  --body "$(cat templates/feedback-issue-template.md)" \
  --label "feedback"

# Generate the checklist from template
python scripts/generate_checklist.py \
  --version ${VERSION} \
  --branch ${BRANCH} \
  --date ${DATE} \
  --manager ${MANAGER} \
  --feedback-issue ${FEEDBACK_ISSUE_NUMBER} \
  --output release-checklist.md

# Create the issue
gh issue create --repo vllm-project/vllm-ascend \
  --title "[Release]: Release checklist for ${VERSION}" \
  --body-file release-checklist.md \
  --label "release"

python scripts/scan_release_bugs.py \
  --repo vllm-project/vllm-ascend \
  --since-tag ${LAST_VERSION} \
  --output issue-scan.md

python scripts/update_checklist_section.py \
  --issue-number ${CHECKLIST_ISSUE} \
  --section "Bug need Solve" \
  --content-file bug-list.md

# 1. [Priority] List open PRs/issues in the release milestone
gh pr list --repo vllm-project/vllm-ascend \
  --state open \
  --search "milestone:${VERSION}" \
  --json number,title,url,labels

gh issue list --repo vllm-project/vllm-ascend \
  --state open \
  --search "milestone:${VERSION}" \
  --json number,title,url,labels

# 2. List open PRs with release-related labels
gh pr list --repo vllm-project/vllm-ascend \
  --state open \
  --label "release-blocker" \
  --json number,title,url

# 3. List PRs merged since last release
gh pr list --repo vllm-project/vllm-ascend \
  --state merged \
  --search "merged:>${LAST_RELEASE_DATE}" \
  --json number,title,mergedAt

python scripts/update_checklist_section.py \
  --issue-number ${CHECKLIST_ISSUE} \
  --section "PR need Merge" \
  --content-file pr-list.md

python scripts/scan_test_coverage.py \
  --repo vllm-project/vllm-ascend \
  --since-tag ${LAST_VERSION} \
  --feedback-issue ${PREVIOUS_FEEDBACK_ISSUE} \
  --output test-coverage-analysis.md

#### Manual Testing Required

- [ ] Model: Kimi K2.5 - Basic inference works
- [ ] Model: GLM-5 - Multimodal features work
- [ ] Feature: Expert parallel with 8 GPUs
- [ ] Feedback: User reported slow startup (verify fixed)

python scripts/update_checklist_section.py \
  --issue-number ${CHECKLIST_ISSUE} \
  --section "Functional Test" \
  --content-file test-results.md

python scripts/scan_nightly_status.py \
  --repo vllm-project/vllm-ascend \
  --output nightly-status.md

python scripts/update_checklist_section.py \
  --issue-number ${CHECKLIST_ISSUE} \
  --section "Nightly Status" \
  --content-file nightly-status.md

# Create output directory
mkdir -p output/${VERSION}

# Fetch commits with contributor statistics
uv run python scripts/fetch_commits.py \
  --owner vllm-project \
  --repo vllm-ascend \
  --base-tag ${LAST_VERSION} \
  --head-tag ${NEW_VERSION} \
  --stats \
  --output output/${VERSION}/0-current-raw-commits.md \
  --stats-output output/${VERSION}/0-contributor-stats.md

# Create analysis workspace
touch output/${VERSION}/1-commit-analysis-draft.csv

## v${VERSION} - ${DATE}

This is the first release candidate of v${VERSION} for vLLM Ascend.
Please follow the [official doc](https://docs.vllm.ai/projects/ascend/en/latest) to get started.

### Highlights
(Top 3-5 most impactful changes)

### Features
(New functionality)

### Hardware and Operator Support
(New hardware/operators)

### Performance
(Performance improvements)

### Dependencies
(Version upgrades)

### Deprecation & Breaking Changes
(Breaking changes)

### Documentation
(Doc updates)

### Others
(Bug fixes, misc)

### Known Issue
(Known limitations)

# Create branch
git checkout -b release/${VERSION}

# Make changes (see Phase 6 for full list)
# ...

# Create PR
gh pr create --repo vllm-project/vllm-ascend \
  --title "Release ${VERSION}" \
  --body "Release notes and version updates for ${VERSION}" \
  --label "release"

python scripts/update_version_references.py \
  --version ${VERSION} \
  --vllm-version ${VLLM_VERSION} \
  --feedback-issue ${FEEDBACK_ISSUE_URL}

# 1. Merge release notes PR (after human review)
gh pr merge ${RELEASE_PR_NUMBER} --repo vllm-project/vllm-ascend --squash

# 2. Create GitHub release
gh release create ${VERSION} \
  --repo vllm-project/vllm-ascend \
  --title "vLLM Ascend ${VERSION}" \
  --notes-file release-notes.md \
  --target main

# 3. Verify automated pipelines (no action needed - CI handles these)
# - Docker image: quay.io/ascend/vllm-ascend:${VERSION}
# - PyPI package: https://pypi.org/project/vllm-ascend/${VERSION}
# - ReadTheDocs: https://app.readthedocs.org/dashboard/

# 4. Upload 310P wheel if applicable
gh release upload ${VERSION} \
  --repo vllm-project/vllm-ascend \
  vllm_ascend-${VERSION}-310p-*.whl

# 1. Broadcast release (prepare announcement)
python scripts/generate_announcement.py \
  --version ${VERSION} \
  --release-notes release-notes.md \
  --output announcement.md

# 2. Close release checklist issue
gh issue close ${CHECKLIST_ISSUE} \
  --repo vllm-project/vllm-ascend \
  --comment "Release ${VERSION} completed successfully!"