Changelog¶

All notable changes to the Code Graph Knowledge System will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]¶

[0.9.0] - 2025-11-07¶

Maintenance¶

• Major improvements:
• Major improvements:
• • Update pyproject.toml description from 'AI-powered code intelligence' to 'Enterprise knowledge management platform'
• • Remove invalid ../README.md references from mkdocs.yml (mkdocs can't access files outside docs/)
• • Move frontend documentation to appropriate docs/ directories

[0.8.0] - 2025-11-06¶

Added¶

• Add Memory Store for AI agent project knowledge persistence (v0.6)

Maintenance¶

-¶

-¶

-¶

• Update permission settings
• Claude/refactor directory structure 011 c us sd tp bi mf1 eiq b hx zr m
• Fix import paths for reorganized services subpackage structure
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• CRITICAL FIX (P0): Fixed all internal imports to use codebase_rag.* namespace
• Fixed all dynamic imports inside test functions that were still using
• Updated all test files to use new import paths:
• • Update build-frontend.sh references: ./build-frontend.sh → ./scripts/build-frontend.sh
• • Remove docs/CNAME (use GitHub Pages default URL)
• This reverts commit 393317c.
• Added professional technical documentation to explain:
• Comprehensive documentation update to reflect the new src-layout structure.
• Complete removal of backward compatibility shims and legacy code structure.
• Scripts documentation should be in docs/development/ instead of
• Phase 6: Non-code file organization and cleanup
• Major restructuring to align with Python best practices:
• Updated cache-to configuration to include a TTL of 7 days.
• Fetch Latest Main Branch Updates
• REMOVED from all Dockerfiles:
• Changed from 'version' to 'bun-version' per official docs:
• • uses: oven-sh/setup-bun@v2 (was @v1)
• CRITICAL FIXES:
• Removed temporary documentation files created during development:
• Updated all docker-compose files to support two-port architecture:

• User Feedback¶

• Context¶

• Context¶

• Problem¶

• Problem¶

• Changes¶

• Problem¶

• Problem¶

• These files are now in .gitignore and should not be tracked:
• As TS Type Safety Auditor, eliminated ALL type issues:
• Upgraded frontend to production-ready modern stack:
• This commit fixes multiple issues that prevented Docker build from succeeding:
• The frontend-builder stage in Dockerfile.full and Dockerfile.standard
• Migrate monitor features from NiceGUI to frontend
• Properly integrate the React frontend into the Docker build process to enable
• Backend Docker images don't need frontend code:
• Runtime images don't need README.md or pyproject.toml:

• Critical Docker Build Issues Fixed¶

• Phase 3: Remove NiceGUI Dependencies¶

• This commit migrates all task monitoring features from NiceGUI to the React frontend:
• Delete old file
• Fix Docker setup and MCP image building
• Implemented complete changelog automation using Conventional Commits:
• Implemented complete version management automation:
• Removed manual "Table of Contents" sections from 3 documentation files:
• Fixed all 9 broken links causing build failures in strict mode:
• Added comprehensive documentation (13,000+ more lines):
• Added comprehensive user documentation (13,000+ lines):
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Added complete documentation for all three deployment modes:
• • Create logo.svg with code graph visualization
• • Remove non-existent custom_dir (docs/overrides)
• • Remove temporary documentation files (CONFIGURATION_GUIDE, summaries)
• Address PR review feedback: align Python versions and clarify Ollama config
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Updated remaining references from docs.vantagecraft.dev to code-graph.vantagecraft.dev
• Complete step-by-step guide covering:
• Complete documentation of Docker MCP Registry submission:
• Add submission files for Docker MCP Registry with 3 variants:
• BREAKING CHANGE: Default docker-compose.yml now points to minimal mode
• Implement automatic memory extraction features
• Address code review feedback: extract magic numbers and fix variable reassignment bug
• Co-authored-by: Copilot 175728472+Copilot@users.noreply.github.com
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Comprehensive implementation of automatic memory extraction features using LLM analysis.
• Fixed test failures caused by MCP SDK type handling:
• [WIP] Add memory management system for AI agent project
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Addressed two issues from code review:
• Fixed two critical P1 bugs in Memory Store:
• Co-authored-by: Copilot 175728472+Copilot@users.noreply.github.com
• Co-authored-by: Copilot 175728472+Copilot@users.noreply.github.com
• Add complete testing infrastructure with 105+ unit tests and CI/CD workflows
• BREAKING CHANGE: FastMCP v1 removed, Official MCP SDK is now the only version
• Migration Summary:
• Implements official Model Context Protocol SDK alongside existing FastMCP version,
• Implements a comprehensive Memory Management system for AI agents to maintain
• This commit adds a complete modern frontend application to replace the
• Fix get_changed_files() dict access in incremental ingestion
• Fix deduplication silently dropping nodes without ref
• Fix dict access, missing parameters, and key mismatches from PR #6 review
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Review codebase-rag against requirements
• This commit adds multi-language support and production-ready Docker deployment.
• This commit completes v0.5 milestone by adding MCP tools for code graph
• • Added v0.4 status and progress tracking (90% complete)
• Implements major v0.4 features: incremental mode and pack enhancements.
• Implements v0.3 requirement: extract import/dependency relationships between files.
• Added detailed documentation covering:
• This commit addresses critical gaps in the v0.2 specification:
• Add repository ingestion, related search, and context packing APIs
• Per feedback, keeping original Python 3.13 requirement
• • Remove all v02-suffixed files and backend/ directory structure
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Co-authored-by: royisme 350731+royisme@users.noreply.github.com
• Roy/tests fix query bug
• This commit introduces a complete real-time monitoring ecosystem with three distinct approaches:
• ```python
• Here's the corrected SEARCH/REPLACE block:
• 1. Added OpenRouterEmbeddingGenerator implementation for embedding generation
• 1. Added OpenRouter fields in config model info
• 1. Adding OpenRouter configuration settings (API key, base URL, models)
• This reverts commit 0b8fd12.

Other¶

• 1. services/init.py:
• Removed eager imports of all subpackages to avoid triggering heavy dependencies (llama_index, etc.) when tests import services
• Updated documentation with correct import examples
• 1. Remove docs/CNAME file (use GitHub Pages default URL)
• Update site_url to vantagecraft.dev/docs/code-graph/
• Update workflow notification URL
• Add comprehensive Cloudflare setup guide
• • Remove main branch push trigger from docker-build.yml
• Now Docker images are built only when version tags are pushed
• This aligns with the bump-version.sh workflow where tags are created
• Keep PR and manual workflow_dispatch triggers for testing
• Avoid conflicts with user's main GitHub Pages site (royisme.github.io) and allow multiple project-specific subdomains.

Planned Features¶

• API authentication with JWT
• Web-based configuration UI
• Multi-provider LLM support (simultaneous)
• Advanced code refactoring suggestions
• Rust, C++, C# language support
• Real-time collaboration features
• Plugin system for custom extensions

[0.7.0] - 2025-01-15¶

Added - Automatic Memory Extraction¶

• Conversation Analysis: Extract memories from AI conversation history
• LLM-powered decision and experience detection
• Confidence scoring for automatic saving
• Configurable auto-save threshold
• Git Commit Analysis: Analyze git commits for architectural decisions
• Parse commit messages and changed files
• Extract decisions, experiences, and conventions
• Link memories to specific commits
• Code Comment Mining: Extract TODO, FIXME, NOTE, DECISION markers
• Automatic scanning of code comments
• Convert markers to structured memories
• Track technical debt and action items
• Query-based Memory Suggestions: Suggest important memories from Q&A
• Analyze knowledge base queries and answers
• Identify information worth remembering
• Suggest memory creation with auto-populated fields
• Batch Repository Extraction: Comprehensive codebase analysis
• Extract from git history (configurable commit limit)
• Mine code comments across file patterns
• Bulk memory creation from repository insights

Added - New MCP Tools (5 tools)¶

• extract_from_conversation: Extract memories from conversation history
• extract_from_git_commit: Analyze git commits for memories
• extract_from_code_comments: Mine code comments for action items
• suggest_memory_from_query: Suggest memories from Q&A sessions
• batch_extract_from_repository: Full repository analysis

Added - New API Endpoints (5 endpoints)¶

• POST /api/v1/memory/extract/conversation: Extract from conversations
• POST /api/v1/memory/extract/commit: Extract from git commits
• POST /api/v1/memory/extract/comments: Extract from code comments
• POST /api/v1/memory/suggest: Suggest memory from query/answer
• POST /api/v1/memory/extract/batch: Batch repository extraction

Changed¶

• Enhanced memory extraction service with LLM-powered analysis
• Improved error messages for memory operations
• Updated MCP handler architecture documentation
• Enhanced memory search relevance scoring

Fixed¶

• Neo4j connection timeout in Docker environments
• Memory search not finding recently added memories
• Environment variable handling in Docker deployment
• Race condition in concurrent memory additions

Documentation¶

• Added comprehensive memory extraction guide
• Updated API documentation with extraction endpoints
• New examples for automatic memory extraction
• Enhanced troubleshooting guide

[0.6.0] - 2024-12-20¶

Added - Memory Store for AI Agents¶

• Memory Management System: Long-term project knowledge persistence
• Decision memory type: Architecture and technical choices
• Preference memory type: Coding styles and conventions
• Experience memory type: Problems encountered and solutions
• Convention memory type: Team rules and standards
• Plan memory type: Future improvements and TODOs
• Note memory type: General project information
• Memory Operations: Full CRUD operations for memories
• Add memory with importance scoring
• Search memories with semantic search
• Update existing memories
• Delete memories (soft delete)
• Supersede memories (version history)
• Project memory summaries
• Memory Relationships: Graph-based memory connections
• BELONGS_TO: Memory to project relationships
• RELATES_TO: Inter-memory relationships
• SUPERSEDES: Memory version history

Added - Multi-Language Support (3 languages)¶

• Java Support: Complete Java code analysis
• Import statement parsing (standard and static)
• Class inheritance and interface tracking
• Method visibility detection (public/protected/private)
• Package dependency mapping
• PHP Support: PHP code analysis
• Use statement parsing (class, function, const)
• Require/include dependency tracking
• Class extends and implements relationships
• Function type hint extraction
• Go Support: Golang code analysis
• Package import parsing (single and blocks)
• Struct and interface detection
• Function and method extraction (with receivers)
• Package alias tracking

Added - Docker Multi-Mode Deployment¶

• Three deployment modes:
• Minimal: Code Graph only (~800MB)
• Standard: Code Graph + Memory (~1.2GB)
• Full: All features (~1.5GB)
• Docker Compose configurations:
• docker-compose.minimal.yml
• docker-compose.standard.yml
• docker-compose.full.yml
• Multi-platform support: amd64, arm64
• Helper scripts: Simplified deployment commands

Added - MCP Tools (7 memory tools)¶

• add_memory: Save new project knowledge
• search_memories: Find relevant memories
• get_memory: Retrieve specific memory
• update_memory: Modify existing memory
• delete_memory: Remove memory
• supersede_memory: Create new memory that replaces old one
• get_project_summary: Get project memory overview

Added - API Endpoints (7 memory endpoints)¶

• POST /api/v1/memory/add: Add new memory
• POST /api/v1/memory/search: Search memories
• GET /api/v1/memory/{memory_id}: Get specific memory
• PUT /api/v1/memory/{memory_id}: Update memory
• DELETE /api/v1/memory/{memory_id}: Delete memory
• POST /api/v1/memory/supersede: Supersede old memory
• GET /api/v1/memory/project/{project_id}/summary: Project summary

Changed¶

• Updated file patterns to include Java, PHP, Go files
• Enhanced code graph to support new language relationships
• Improved transformer architecture for multi-language support

Documentation¶

• Added Memory Store user guide
• Added memory API documentation
• Updated examples with memory usage
• Enhanced CLAUDE.md with memory workflows

[0.5.0] - 2024-11-15¶

Added - MCP Protocol Support¶

• Official MCP SDK Integration: Model Context Protocol v1.1.0+
• Modular Architecture: Handler-based design (310-line main server)
• Knowledge handlers: Query, search, document management
• Code graph handlers: Ingestion, analysis, statistics
• System handlers: Health checks, configuration
• Task handlers: Background processing, monitoring
• 30 MCP Tools: Comprehensive AI assistant integration
• 8 knowledge tools
• 10 code graph tools
• 4 system tools
• 8 task monitoring tools
• Advanced Features:
• Session management framework
• Streaming support (SSE)
• Multi-transport capability (stdio, SSE, WebSocket)

Added - Prometheus Metrics¶

• 15+ metrics for monitoring:
• Request counters (total, by endpoint, by status)
• Request duration histograms
• Active request gauges
• Neo4j operation metrics
• Document processing metrics
• Error rate tracking
• Metrics endpoint: GET /api/v1/metrics
• Grafana dashboard configuration (optional)

Added - Neo4j Health Monitoring¶

• Connection status tracking
• Query performance metrics
• Database size monitoring
• Index usage statistics

Changed¶

• Refactored MCP server from 1400 lines to 310 lines (78% reduction)
• Extracted handlers into mcp_tools/ package
• Improved error handling and logging
• Enhanced code organization and maintainability

Documentation¶

• Added MCP v2 modularization guide
• Updated MCP integration documentation
• Added Prometheus metrics documentation
• Enhanced deployment guides

[0.4.0] - 2024-10-20¶

Added - Real-time Task Monitoring¶

• Web UI Monitoring: NiceGUI-based monitoring interface
• Real-time task status updates via WebSocket
• File upload functionality (50KB size limit)
• Directory batch processing
• Task progress visualization
• Accessible at /ui/monitor when ENABLE_MONITORING=true
• Server-Sent Events (SSE): Streaming APIs for real-time updates
• /api/v1/sse/task/{task_id}: Monitor single task
• /api/v1/sse/tasks: Monitor all tasks with filtering
• /api/v1/sse/stats: SSE connection statistics
• Task Queue System: Background processing with monitoring
• Async task execution
• Progress tracking
• Error handling and retry logic
• Task history and logs

Added - Large File Handling¶

• Multi-strategy approach:
• Direct processing: Files < 10KB
• Temporary file strategy: Files 10-50KB
• Directory processing prompt: Files > 50KB
• MCP automatic temp files: All sizes
• Configurable limits: Size thresholds via environment variables

Added - Client Examples¶

• examples/pure_mcp_client.py: Pure MCP client with watch tools
• examples/hybrid_http_sse_client.py: HTTP + SSE hybrid approach
• Real-time monitoring demonstrations

Changed¶

• Enhanced file upload handling with size validation
• Improved error messages for large file uploads
• Better timeout handling for large documents

Fixed¶

• Memory leaks in long-running tasks
• SSE connection stability issues
• File upload timeout for large files

[0.3.0] - 2024-09-15¶

Added - Universal SQL Schema Parser¶

• Multi-dialect support: Oracle, MySQL, PostgreSQL, SQL Server
• Configurable business domain classification: YAML/JSON configuration
• Pre-built industry templates:
• Insurance: Policies, claims, underwriting
• E-commerce: Products, orders, customers
• Banking: Accounts, transactions, loans
• Healthcare: Patients, diagnoses, treatments
• Comprehensive parsing:
• Table and column extraction
• Foreign key relationships
• Index definitions
• Business domain classification
• Professional documentation generation: Markdown output
• Real-world tested: 356-table Oracle database (4,511 columns)

Added - SQL API Endpoints¶

• POST /api/v1/sql/parse: Parse SQL schema files
• POST /api/v1/sql/analyze: Analyze database structure
• GET /api/v1/sql/templates: List available templates

Changed¶

• Enhanced schema parsing with configurable templates
• Improved relationship detection
• Better error handling for malformed SQL

Documentation¶

• Added SQL parsing user guide
• Industry template documentation
• Configuration examples

[0.2.0] - 2024-08-01¶

Added - Multi-Provider LLM Support¶

• Ollama integration: Local LLM hosting (default)
• OpenAI integration: GPT models and embeddings
• Google Gemini integration: Gemini models and embeddings
• OpenRouter integration: Multi-provider access
• HuggingFace embeddings: Local embedding models
• Provider configuration: Via .env file with flexible switching

Added - Enhanced Configuration¶

• Environment-based configuration system
• Support for multiple embedding providers
• Configurable timeouts and limits
• Feature flags (monitoring, Prometheus)

Changed¶

• Refactored service initialization for multi-provider support
• Improved LLM provider abstraction layer
• Enhanced error messages for provider issues

Fixed¶

• OpenAI API compatibility issues
• Gemini embedding dimension mismatches
• Provider-specific timeout handling

[0.1.0] - 2024-07-01¶

Added - Initial Release¶

• Core Features:
• Neo4j GraphRAG integration
• Vector search with LlamaIndex
• Document processing (text, markdown, code)
• Knowledge graph construction
• Intelligent query engine
• RESTful API
• Code Analysis:
• Python code parsing
• TypeScript/JavaScript parsing
• Import relationship mapping
• Basic code graph visualization
• Document Management:
• Multi-format support
• Asynchronous processing
• Chunk-based indexing
• Vector similarity search
• API Endpoints:
• /api/v1/health: Health check
• /api/v1/knowledge/query: Query knowledge base
• /api/v1/knowledge/search: Vector search
• /api/v1/documents/upload: Upload documents
• /api/v1/documents/list: List documents
• Infrastructure:
• FastAPI backend
• Neo4j database
• Docker support
• Basic logging and error handling

Documentation¶

• Initial README
• API documentation
• Basic deployment guide
• Example scripts

Version History Summary¶

Upgrade Guides¶

Upgrading from 0.6.x to 0.7.0¶

No breaking changes. Simply pull new Docker image:

docker pull royisme/codebase-rag:0.7.0-full
docker-compose restart

New Features Available: - Memory extraction endpoints and MCP tools - Automatic memory mining from git and code

Upgrading from 0.5.x to 0.6.0¶

Breaking Changes: - None, fully backward compatible

New Configuration Options:

# Optional: Enable memory features (included in standard/full modes)
ENABLE_MEMORY_STORE=true

Data Migration: - No migration needed - Memory Store creates new nodes in existing Neo4j database

Upgrading from 0.4.x to 0.5.0¶

Breaking Changes: - MCP server entry point changed from mcp_server.py to start_mcp.py

Configuration Update:

// claude_desktop_config.json
{
  "mcpServers": {
    "code-graph": {
      "command": "python",
      "args": ["/path/to/start_mcp.py"]  // Changed from mcp_server.py
    }
  }
}

Data Migration: - No database changes - MCP protocol fully backward compatible

Upgrading from 0.3.x to 0.4.0¶

No breaking changes. New features are opt-in:

# Enable monitoring UI
ENABLE_MONITORING=true

# Enable Prometheus metrics
ENABLE_PROMETHEUS=true

Migration Notes¶

Python Version Upgrade¶

As of v0.6.0, Python 3.13+ is required. If upgrading from older versions:

# Update Python
python3.13 -m venv .venv
source .venv/bin/activate

# Reinstall dependencies
pip install --upgrade pip
pip install -e .

Neo4j Version Compatibility¶

All versions support Neo4j 5.0+. No database migration needed between versions.

Environment Variables¶

Check .env.example for new configuration options in each version.

Deprecation Notices¶

Deprecated in 0.7.0¶

• None

Deprecated in 0.6.0¶

• None

Deprecated in 0.5.0¶

• Old MCP server entry point (mcp_server.py): Use start_mcp.py instead
• Will be removed in: v1.0.0

Removed in 0.5.0¶

• None

Contributing¶

See CONTRIBUTING.md for guidelines on contributing to this project.

Support¶

• Documentation: https://code-graph.vantagecraft.dev
• Issues: https://github.com/royisme/codebase-rag/issues
• Discussions: https://github.com/royisme/codebase-rag/discussions

Links¶

• Homepage
• GitHub Repository
• Docker Hub
• Issue Tracker

Note: Dates in this changelog are illustrative. Check GitHub Releases for actual release dates.