| name | deployment-git |
| description | Kailash deployment + Git — PyPI publish, CI/CD, wheels, version bumps, multi-package. |
SDK Release & Git Workflows
Comprehensive guides for releasing the Kailash Python SDK to PyPI, managing CI/CD pipelines, deploying documentation, and Git workflow best practices.
Overview
SDK release and development infrastructure patterns for:
- PyPI package publishing (TestPyPI + production)
- CI/CD pipelines (GitHub Actions)
- Multi-platform wheel building
- Documentation deployment
- Git workflows and branching strategies
- Multi-package version coordination
Reference Documentation
Deployment Lifecycle
-
deployment-onboarding - SDK release onboarding process
- Codebase analysis (packages, build system, CI, docs)
- Structured questions for human architect
- Research current PyPI/CI best practices
- Create deployment-config.md
-
release-runbook - SDK release runbook (detailed procedures)
- Version locations for all packages (pyproject.toml + init.py)
- SDK dependency pin rules
- Version consistency verification commands
- Pre-release, build, publish, post-release step-by-step procedures
-
deployment-packages - Package release workflow
- PyPI and GitHub release process
- Multi-package coordination and publish order
- Version bumping and changelog
- CI-triggered releases
- TestPyPI validation
- Rollback procedures
-
deployment-ci - CI/CD infrastructure
- GitHub Actions workflows for Python packages
- Multi-platform wheel building
- Test matrix (Python versions x OS)
- Tag-triggered publishing pipeline
- Documentation deployment (ReadTheDocs, GitHub Pages)
- Self-hosted runner management
-
python-version-bump - Python minor-version bump playbook
- When to declare a new CPython release (3.X stable + ≥1 patch)
- 3-step recipe: pyproject classifiers → CI matrices →
uv pip install --dry-run --python 3.X verification
- Concrete file list for all 10 packages and 4 CI matrix files
- Common gotcha: stale literal version assertions in test fixtures (durable cross-surface contract pattern)
- ML stack wheel-lag guidance (torch / transformers / accelerate)
-
multi-package-release-wave - Atomic 7-package release coordination
- Reverse dep-graph publish order (
kailash → dataflow → nexus → kaizen → pact → align → ml)
- Per-package version owner + sole CHANGELOG owner rules (parallel-worktree safe)
- Pre-flight build + twine + TestPyPI dry-run for every package in the wave
- Version consistency verification across 14 version locations
- Rollback decision tree (mid-wave failure handling)
- kailash-ml 1.0.0 M1 atomic wave (2026-04-23, 7 packages, 227 tests)
Docker
- deployment-docker-quick - Docker deployment patterns
- Dockerfile setup for Kailash apps
- Docker Compose configurations
- Multi-stage builds
- Health checks
Kubernetes
Git Workflow
- git-workflow-quick - Git workflow best practices
- Branching strategies
- Commit conventions
- Pull request workflow
- Code review process
- Release management
- Hotfix procedures
GitHub Management
- github-management-patterns - GitHub project and issue management
- Issue templates (User Story, Bug, Technical Task)
- Story points and estimation
- Project board organization
- Label system
Project Management
SDK Release Patterns
Release Flow
Version bump + CHANGELOG → Build wheels → TestPyPI → Verify → PyPI → GitHub Release → Deploy Docs
CI-Triggered Release (Preferred)
git tag v1.2.3 → push tag → CI builds wheels → CI tests → CI publishes TestPyPI → CI publishes PyPI → CI creates GitHub Release
Multi-Package Release Order
kailash (core)
↓
kailash-dataflow (depends on core)
kailash-nexus (depends on core)
kailash-kaizen (depends on core)
Git Workflow Patterns
Branch Strategy
main (production)
↓
develop (integration)
↓
feature/* (new features)
hotfix/* (urgent fixes)
release/* (release prep)
Commit Conventions
feat: Add user authentication workflow
fix: Resolve async runtime threading issue
docs: Update DataFlow integration guide
test: Add cycle workflow test cases
chore: Bump version to 0.9.25
Critical Rules
SDK Release
- Run full test suite before any release
- TestPyPI validation required for major/minor releases
- Wheel-only publishing for proprietary code
- Version consistency across all sub-packages
- Publish in dependency order (core first)
- Security review before every publish
- NEVER commit PyPI tokens to source
- NEVER publish with failing CI
Git
- Use feature branches for development
- Write descriptive commit messages
- Squash commits before merging
- Use pull requests for code review
- Tag releases semantically
- NEVER commit directly to main
- NEVER force push to shared branches
- NEVER commit sensitive data
When to Use This Skill
Use this skill when you need to:
- Run SDK release onboarding for a new project
- Release packages to PyPI or GitHub
- Set up or debug CI/CD pipelines
- Configure GitHub Actions workflows
- Build multi-platform wheels
- Deploy documentation
- Coordinate multi-package releases
- Establish Git workflows
- Manage test matrices
Related Skills
Support
For SDK release help, invoke:
release-specialist - Release onboarding, PyPI publishing, CI management
release-specialist - Git workflows, releases, version management