code-knowledge-graph-hv 1.0.0

A lightweight knowledge graph engine for codebases with CLI, API, and MCP support

Code Knowledge Graph Engine

A lightweight, production-grade knowledge graph engine for codebases. Build and maintain an always-updated graph of code relationships (functions, classes, imports, calls) and query it via CLI, API, or MCP (Model Context Protocol) for AI-assisted development.

🎯 Goals

• Reduce LLM token usage: Retrieve only relevant code relationships instead of entire files
• Always-updated graph: Auto-update on file changes with built-in watcher
• Plug-and-play: Easy integration into any project with a single command
• Multiple interfaces: CLI, REST API, and MCP server for AI agents
• Extensible: Simple architecture for enterprise customization

🏗️ Architecture

┌─────────────────────────────────────────────────────────────┐
│                     Code Knowledge Graph                     │
├─────────────────────────────────────────────────────────────┤
│                                                               │
│  ┌─────────────┐    ┌─────────────┐    ┌─────────────┐      │
│  │     CLI     │    │  FastAPI    │    │  MCP Server │      │
│  │  (typer)    │    │   Server    │    │  (stdio)    │      │
│  └──────┬──────┘    └──────┬──────┘    └──────┬──────┘      │
│         │                  │                  │              │
│         └──────────────────┼──────────────────┘              │
│                            │                                 │
│                   ┌─────────▼─────────┐                      │
│                   │   Graph Retriever  │                      │
│                   │   (query engine)   │                      │
│                   └─────────┬─────────┘                      │
│                             │                                 │
│                   ┌─────────▼─────────┐                      │
│                   │  Knowledge Graph   │                      │
│                   │  (nodes + edges)   │                      │
│                   └─────────┬─────────┘                      │
│                             │                                 │
│                   ┌─────────▼─────────┐                      │
│                   │  Graph Builder    │                      │
│                   └─────────┬─────────┘                      │
│                             │                                 │
│                   ┌─────────▼─────────┐                      │
│                   │  Python Parser    │                      │
│                   │    (AST-based)    │                      │
│                   └─────────┬─────────┘                      │
│                             │                                 │
│                   ┌─────────▼─────────┐                      │
│                   │  File Watcher     │                      │
│                   │   (watchdog)      │                      │
│                   └───────────────────┘                      │
│                             │                                 │
│                    ┌────────▼────────┐                       │
│                    │  Python Files   │                       │
│                    └─────────────────┘                       │
│                                                               │
└─────────────────────────────────────────────────────────────┘

📦 Installation

From source

# Clone the repository
git clone https://github.com/yourusername/code-knowledge-graph.git
cd code-knowledge-graph

# Install with pip
pip install .

Development installation

pip install -e ".[dev]"

🚀 Quick Start

1. Build the knowledge graph

# Build graph for current directory
kg build

# Build graph for specific directory
kg build /path/to/project

# Exclude certain directories
kg build . --exclude "venv,node_modules,.git"

2. Query dependencies

# Query dependencies for a function
kg query my_function

# Query with custom depth
kg query my_function --depth 3

# Save output to file
kg query my_function --output deps.json

3. Start the watcher (auto-update)

# Watch current directory
kg watch

# Watch specific directory
kg watch /path/to/project

4. Check status

kg status

5. Search nodes

# Search for nodes
kg search "user"

# Filter by type
kg search "user" --node-type function

🌐 API Server

Start the server

kg serve --host 0.0.0.0 --port 8000

API Endpoints

Health Check

GET /health

Get Dependencies

GET /dependencies?target=function_name&depth=5

Get Full Graph

GET /graph

Get Statistics

GET /stats

Search Nodes

GET /search?query=search_term&node_type=function

Get Function Calls

GET /functions/{function_name}/calls

Get File Imports

GET /files/{file_path}/imports

Example using curl

# Get dependencies
curl "http://localhost:8000/dependencies?target=my_function"

# Get stats
curl http://localhost:8000/stats

# Search
curl "http://localhost:8000/search?query=user"

🔌 MCP Server (Model Context Protocol)

The MCP server allows AI agents (like Windsurf, Claude Desktop) to query the knowledge graph directly.

Start MCP server

kg-mcp --graph-path storage/graph.json

MCP Methods

dependencies

Get upstream and downstream dependencies for a target.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "dependencies",
  "params": {
    "target": "my_function",
    "depth": 5
  }
}

search

Search for nodes in the graph.

{
  "jsonrpc": "2.0",
  "id": 2,
  "method": "search",
  "params": {
    "query": "user",
    "node_type": "function"
  }
}

stats

Get graph statistics.

{
  "jsonrpc": "2.0",
  "id": 3,
  "method": "stats"
}

graph

Get the complete graph.

{
  "jsonrpc": "2.0",
  "id": 4,
  "method": "graph"
}

Integration with AI Tools

Windsurf Configuration

Add to your Windsurf config:

{
  "mcpServers": {
    "knowledge-graph": {
      "command": "kg-mcp",
      "args": ["--graph-path", "storage/graph.json"]
    }
  }
}

📊 Graph Schema

Node Types

• file: Represents a Python source file
• function: Represents a function or method
• class: Represents a class

Edge Types

• calls: Function A calls Function B
• imports: File A imports File B
• defines: File defines Function/Class

Example Graph

{
  "nodes": {
    "file:main.py": {
      "id": "file:main.py",
      "type": "file",
      "name": "main.py",
      "file_path": "main.py",
      "line_number": 1
    },
    "function:process_data:main.py": {
      "id": "function:process_data:main.py",
      "type": "function",
      "name": "process_data",
      "file_path": "main.py",
      "line_number": 10
    }
  },
  "edges": [
    {
      "source": "file:main.py",
      "target": "function:process_data:main.py",
      "type": "defines"
    }
  ]
}

🧪 Testing

# Run tests
pytest

# Run with coverage
pytest --cov=core --cov=api --cov=cli --cov=mcp

🔧 Configuration

Storage

The graph is stored in JSON format at storage/graph.json by default. You can customize this path:

kg build . --output /custom/path/graph.json
kg query my_function --graph-path /custom/path/graph.json

Exclude Directories

When building the graph, certain directories are excluded by default:

• venv, env, .git, __pycache__, .pytest_cache, node_modules

You can customize this:

kg build . --exclude "custom_dir,another_dir"

🎨 Use Cases

1. Code Navigation

Quickly find what functions a specific function calls, or what functions call it.

kg query authenticate_user

2. Impact Analysis

Before refactoring, understand the downstream impact of changing a function.

kg query process_payment --depth 10

3. Code Review

Understand the relationships in a new codebase.

kg build /path/to/new/project
kg status

4. AI-Assisted Development

Provide AI agents with structured context about code relationships instead of raw files.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "dependencies",
  "params": {"target": "main"}
}

🔒 Security

• The graph stores file paths and code structure, not actual code content
• No external network calls
• All operations are local

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

MIT License - see LICENSE file for details

🙏 Acknowledgments

• Built with Python AST for reliable parsing
• Uses watchdog for file system monitoring
• FastAPI for the REST API
• Typer for the CLI
• MCP protocol for AI agent integration

📞 Support

For issues, questions, or suggestions, please open an issue on GitHub.