Code Graph Knowledge System

Enterprise Knowledge Management Platform

Neo4j-powered graph database with multi-interface architecture (MCP/Web/REST) and intelligent code analysis

What is Code Graph Knowledge System?¶

Code Graph Knowledge System is an enterprise-grade solution that transforms unstructured development documentation and code into a structured, queryable knowledge graph. By combining vector search, graph database technology, and large language models, it provides intelligent code analysis, documentation management, and development assistance capabilities.

✨ Key Features¶

🎯 Deployment Modes¶

Choose the right deployment mode for your needs:

Feature	Minimal	Standard	Full
Code Graph
└ Repository ingestion	✅	✅	✅
└ File relationship search	✅	✅	✅
└ Impact analysis	✅	✅	✅
└ Context packing	✅	✅	✅
Memory Store
└ Manual management	❌	✅	✅
└ Vector search	❌	✅	✅
└ Auto extraction (Git)	❌	❌	✅
└ Auto extraction (Conversations)	❌	❌	✅
Knowledge RAG
└ Document vectorization	❌	❌	✅
└ Intelligent Q&A	❌	❌	✅
Requirements
└ Neo4j	✅	✅	✅
└ Embedding Model	❌	✅	✅
└ LLM	❌	❌	✅
Image Size	~500MB	~600MB	~800MB
Startup Time	~5s	~8s	~15s

🚀 Core Capabilities¶

Code GraphMemory StoreKnowledge RAG

No LLM Required - Pure graph-based code intelligence

Repository Ingestion: Parse and index entire codebases
Relationship Discovery: Find file dependencies and imports
Impact Analysis: Understand the blast radius of changes
Context Packing: Generate AI-friendly context bundles

# Start minimal deployment
make docker-minimal

Embedding Required - Long-term project knowledge

Manual Memory Management: Add, search, update memories
Vector Search: Find relevant project decisions
Auto Extraction: Extract from git commits and conversations
Knowledge Evolution: Track decision changes over time

# Start standard deployment
make docker-standard

LLM + Embedding Required - Full AI capabilities

Document Processing: Index documentation and code
Intelligent Q&A: Answer questions about your codebase
Multi-format Support: Markdown, PDF, code files
Hybrid Search: Combine vector and graph traversal

# Start full deployment
make docker-full

🎯 Quick Start¶

1. Choose Your Deployment Mode¶

# Minimal - Code Graph only (No LLM needed)
make docker-minimal

# Standard - Code Graph + Memory (Embedding needed)
make docker-standard

# Full - All features (LLM + Embedding needed)
make docker-full

2. Access the System¶

API Documentation: http://localhost:8000/docs
Neo4j Browser: http://localhost:7474
Health Check: http://localhost:8000/api/v1/health

3. Ingest Your Repository¶

MCP (Claude Desktop)REST API

// In Claude Desktop, use MCP tools:
code_graph_ingest_repo({
  "local_path": "/path/to/your/repo",
  "mode": "incremental"
})

curl -X POST http://localhost:8000/api/v1/code-graph/ingest \
  -H "Content-Type: application/json" \
  -d '{
    "local_path": "/path/to/your/repo",
    "mode": "incremental"
  }'

🎨 Use Cases¶

For Individual Developers¶

Learn Large Codebases: Quickly understand unfamiliar projects
Code Navigation: Find relationships and dependencies
Impact Assessment: See what breaks before making changes

For Development Teams¶

Project Knowledge Base: Preserve team decisions and context
Onboarding: Help new team members get up to speed
Documentation: Auto-generate context for AI coding assistants

For AI-Assisted Development¶

Claude Desktop Integration: Use as MCP server for enhanced code understanding
VS Code Integration: Access code graph directly in your editor
Context Generation: Create optimal context for LLM queries

📚 Documentation¶

Quick Start Guide - Get running in 5 minutes
Deployment Overview - Choose the right mode
Code Graph Guide - Learn code intelligence features
MCP Integration - Use with Claude Desktop
API Reference - Complete tool documentation

🌟 Why Code Graph Knowledge System?¶

🎯 Flexible Architecture¶

No Vendor Lock-in: Use Ollama, OpenAI, Gemini, or any LLM
Scalable: From single developer to enterprise teams
Modular: Only use what you need

🚀 Performance Optimized¶

Incremental Updates: 60x faster than full re-indexing
Smart Caching: Reduce redundant processing
Efficient Storage: Neo4j native vector indexes

🔒 Privacy Focused¶

Self-Hosted: Keep your code on your infrastructure
No Data Leaks: Optional local LLM support
Secure: Enterprise-grade Neo4j backend

🛠️ Technology Stack¶

Backend: Python 3.13, FastAPI
Database: Neo4j 5.15+ with APOC
AI: LlamaIndex, Multiple LLM providers
Protocol: Model Context Protocol (MCP)
Deployment: Docker, Docker Compose

GitHub: royisme/codebase-rag
Docker Hub: royisme/codebase-rag
Issues: Report bugs
Discussions: Community forum

📝 License¶

This project is licensed under the MIT License - see the LICENSE file for details.

**Ready to get started?** → [Quick Start Guide](getting-started/quickstart.md)