Project Quality Hub

PyPI 最新版本：project-quality-hub 0.1.0（发布日期：2025-11-01）

Model Context Protocol tooling that gives AI assistants a trustworthy project graph, branch intelligence, smart incremental updates, and explainable quality scores.

Looking for the Chinese overview? Jump to 中文简介.

Quickstart

# 1. Install the package
pip install project-quality-hub

# 2. Analyse a repository and optionally enable live monitoring
project-quality-hub analyze ./demo --monitor

# 3. Retrieve the knowledge graph summary
project-quality-hub summary ./demo

# 4. Score a project or a single file
project-quality-hub score ./demo --file src/example.py --max-files 50

Prefer editable installs while developing?

pip install -e .[dev]
pytest
ruff check src

Why Project Quality Hub?

• Project graph intelligence – Build a knowledge graph with entity-level insights, dependency edges, and risk scoring so assistants can reason about codebases.
• Branch-aware memories – Cache per-branch analyses, switch between them, and compare git branches without rebuilding from scratch.
• Smart incremental updates – Watch file changes with watchdog to refresh analysis results in the background.
• Quality scoring built for AI – Blend metrics, static-analysis findings, and heuristics into transparent 0‑100 scores with actionable recommendations.
• MCP-native experience – Ship the same capabilities via CLI commands or an MCP stdio server for Claude, Cursor, and other compatible clients.

CLI Essentials

# Analyse and cache a project (monitoring is opt-in)
project-quality-hub analyze /path/to/project [--force] [--monitor]

# Retrieve a summary of the analysed project
project-quality-hub summary /path/to/project

# Run quality scoring across the repo or a single file
project-quality-hub score /path/to/project [--file relative/path.py] [--max-files N]

# Control background monitoring
project-quality-hub monitor /path/to/project start|stop|status

# Launch the MCP stdio server
project-quality-hub server
project-quality-hub-server  # dedicated entry point

See project-quality-hub --help for the full command list.

MCP Client Integration

1. Install the package on the machine hosting your MCP server.
2. Point your client at the stdio transport. Claude Desktop example:

   {
     "endpoints": [
       {
         "name": "project-quality-hub",
         "command": ["project-quality-hub-server"],
         "transport": { "type": "stdio" }
       }
     ]
   }
   

3. Restart the client. The tools listed above become available instantly.
4. Need more detail? Check the full walkthrough in docs/integration.md.

Project Layout

• src/project_quality_hub/core: knowledge-graph modelling, multi-branch management, incremental updates.
• src/project_quality_hub/quality: AST inspection, static-analysis adapters, scoring heuristics.
• src/project_quality_hub/server: MCP stdio adapter, task orchestration, utilities.
• src/project_quality_hub/cli.py: CLI entry point mirroring the MCP toolset.
• tests/: import safety plus behavioural tests for scoring and parsing.
• docs/: design notes, client integration, contributing guide.

Testing & Development

• Run unit tests with pytest.
• Lint with ruff check src; format with black src.
• Export WATCHDOG_FORCE_POLLING=1 in sandboxed environments to guarantee deterministic monitoring.
• Clean build artefacts (dist/, build/) before running quality scoring for the most accurate results.

Releases

1. Bump the version in pyproject.toml and project_quality_hub/__init__.py.
2. Build artefacts via python -m build.
3. Publish to PyPI using twine upload dist/*.
4. Tag the release, open a GitHub Release, and capture the highlights in CHANGELOG.md.

Contributing

We welcome issues and pull requests! Review the contributing guide for coding standards, workflow, and communication expectations. A behaviour code and PR templates will keep contributions friendly and consistent.

License

Distributed under the MIT License.

中文简介

project-quality-hub 将项目图谱、分支管理、智能增量更新和质量评分能力封装为 Model Context Protocol (MCP) 服务，方便 Claude、Cursor 等客户端直接调用。

• 项目图谱分析：构建实体级知识图谱，输出依赖关系和风险评估。
• 多分支记忆：缓存并比较不同 Git 分支的分析结果，快速切换。
• 实时增量更新：结合 watchdog 监听文件变化，自动刷新结果。
• 质量评分：综合指标与静态分析，提供 0-100 分的评分和优化建议。

仓库提供 CLI 与 MCP 双入口。更多集成细节请参阅 docs/integration.md，贡献准则见 docs/CONTRIBUTING.md。

Historical versions of this README remain available in README_EN.md.