gmakstutis/graphiti
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
graphiti/mcp_server/PYPI_PUBLISHING.md
Lars Varming eddeda67b3 Fix graphiti-mcp-varming package for PyPI publication 
MCP Server Package Fixes:
- Add build-system configuration to pyproject.toml
- Fix module imports to use relative imports for proper packaging
- Fix TypedDict import for Python 3.10 compatibility
- Remove unsupported MCP SDK parameters (tags, meta)
- Add GitHub Actions workflow for automatic PyPI publishing
- Add PyPI publishing documentation and checklist

Code Quality Improvements:
- Fix code formatting in graphiti_core (line length, whitespace)

This prepares v1.0.0 for publication to PyPI, enabling users to install
with: uvx graphiti-mcp-varming

🤖 Generated with Claude Code (https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-09 19:32:13 +01:00

3.1 KiB
Raw Blame History

PyPI Publishing Setup Instructions

This guide explains how to publish the graphiti-mcp-varming package to PyPI.

One-Time Setup

1. Add PyPI Token to GitHub Secrets

  1. Go to your repository on GitHub: https://github.com/Varming73/graphiti
  2. Click Settings → Secrets and variables → Actions
  3. Click New repository secret
  4. Name: PYPI_API_TOKEN
  5. Value: Paste your PyPI API token (starts with pypi-)
  6. Click Add secret

Publishing a New Version

Option 1: Automatic Publishing (Recommended)

  1. Update the version in mcp_server/pyproject.toml:

    version = "1.0.1"  # Increment version

  2. Commit the change:

    cd mcp_server
git add pyproject.toml
git commit -m "Bump MCP server version to 1.0.1"
git push

  3. Create and push a tag:

    git tag mcp-v1.0.1
git push origin mcp-v1.0.1

  4. GitHub Actions will automatically:

    • Build the package
    • Publish to PyPI
    • Create a GitHub release

  5. Monitor the workflow:

    • Go to Actions tab in GitHub
    • Watch the "Publish MCP Server to PyPI" workflow

Option 2: Manual Publishing

If you prefer to publish manually:

cd mcp_server

# Remove local graphiti-core override
sed -i.bak '/\[tool\.uv\.sources\]/,/graphiti-core/d' pyproject.toml

# Build the package
uv build

# Publish to PyPI
uv publish --token your-pypi-token-here

# Restore the backup if needed for local development
mv pyproject.toml.bak pyproject.toml

After Publishing

Users can install your package with:

# Basic installation (Neo4j support included)
uvx graphiti-mcp-varming

# With FalkorDB support
uvx --with graphiti-mcp-varming[falkordb] graphiti-mcp-varming

# With all LLM providers (Anthropic, Groq, Gemini, Voyage, etc.)
uvx --with graphiti-mcp-varming[providers] graphiti-mcp-varming

# With everything
uvx --with graphiti-mcp-varming[all] graphiti-mcp-varming

Version Numbering

Follow Semantic Versioning:

  • MAJOR (1.0.0 → 2.0.0): Breaking changes
  • MINOR (1.0.0 → 1.1.0): New features, backwards compatible
  • PATCH (1.0.0 → 1.0.1): Bug fixes

Tag Naming Convention

Use mcp-v{VERSION} format for tags:

  • mcp-v1.0.0 - Initial release
  • mcp-v1.0.1 - Patch release
  • mcp-v1.1.0 - Minor release
  • mcp-v2.0.0 - Major release

This distinguishes MCP server releases from graphiti-core releases.

Troubleshooting

Publishing Fails with "File already exists"

You tried to publish a version that already exists on PyPI. Increment the version number in pyproject.toml and try again.

"Invalid or missing authentication token"

The PyPI token in GitHub secrets is incorrect or expired:

  1. Generate a new token at https://pypi.org/manage/account/token/
  2. Update the PYPI_API_TOKEN secret in GitHub

Workflow doesn't trigger

Make sure:

  • Tag matches pattern mcp-v*.*.*
  • Tag is pushed to GitHub: git push origin mcp-v1.0.0
  • Workflow file is on the main branch

Checking Published Package

After publishing, verify at: