Expert Registry MCP Server

# Expert Registry MCP Server - Project Structure **Last Updated: 2025-06-30** ## Overview This project implements a high-performance MCP server for expert discovery, registration, and context injection using FastMCP v2, with integrated vector and graph database support. ## Directory Structure ``` expert-registry-mcp/ ├── src/ │ └── expert_registry_mcp/ │ ├── __init__.py # Package initialization │ ├── server.py # FastMCP server implementation │ ├── models.py # Pydantic data models │ ├── registry.py # Registry management with file watching │ ├── selection.py # Expert selection engine │ ├── context.py # Context loading and injection │ ├── vector_db.py # Vector database integration (ChromaDB) │ ├── graph_db.py # Graph database integration (Neo4j) │ ├── embeddings.py # Embedding generation pipeline │ └── discovery.py # Hybrid discovery algorithms │ ├── tests/ │ ├── test_registry.py # Registry management tests │ ├── test_selection.py # Selection engine tests │ ├── test_vector_db.py # Vector database tests │ ├── test_graph_db.py # Graph database tests │ └── test_integration.py # End-to-end integration tests │ ├── docs/ │ ├── python-code-updates.md # Python implementation examples │ └── typescript-examples/ # Legacy TypeScript reference │ ├── expert-registry-mcp-server-example.ts │ ├── typescript-package-example.json │ └── typescript-readme.md │ ├── scripts/ │ ├── setup_databases.py # Database initialization script │ ├── import_registry.py # Import existing registry │ └── generate_embeddings.py # Pre-generate embeddings │ ├── expert-system/ # Default expert system directory │ ├── registry/ │ │ └── expert-registry.json # Central expert registry │ ├── expert-contexts/ │ │ ├── aws-amplify-gen2.md # Expert context files │ │ ├── aws-cloudscape.md │ │ └── ... │ ├── vector-db/ # ChromaDB storage │ └── performance/ │ └── metrics.json # Performance tracking │ ├── pyproject.toml # Project configuration (uv/pip) ├── README.md # Main documentation ├── expert-registry-mcp-design.md # Architecture design document ├── PROJECT_STRUCTURE.md # This file ├── LICENSE # MIT License └── .gitignore # Git ignore patterns ``` ## Key Components ### Core Server (`src/expert_registry_mcp/`) - **server.py**: Main FastMCP server with all tool definitions - **models.py**: Pydantic models for type safety and validation - **registry.py**: File-based registry with hot reload support - **selection.py**: Technology detection and expert scoring - **context.py**: Expert knowledge management ### Database Integration - **vector_db.py**: ChromaDB integration for semantic search - **graph_db.py**: Neo4j integration for relationship modeling - **embeddings.py**: Sentence transformer pipeline - **discovery.py**: Hybrid search combining vector and graph ### Testing (`tests/`) Comprehensive test suite using pytest and pytest-asyncio for: - Unit tests for each component - Integration tests for database operations - End-to-end workflow tests ### Documentation (`docs/`) - Implementation guides and examples - Legacy TypeScript reference (for historical context) - API documentation ### Scripts (`scripts/`) Utility scripts for: - Database setup and initialization - Registry import from existing systems - Embedding pre-generation for performance ## Data Flow 1. **Registry Loading**: File watcher monitors `expert-registry.json` 2. **Embedding Generation**: Automatic generation on registry updates 3. **Database Sync**: Updates propagated to both vector and graph DBs 4. **Query Processing**: Hybrid search across both databases 5. **Context Injection**: Expert knowledge enhancement of prompts 6. **Performance Tracking**: Analytics stored and used for optimization ## Configuration ### Environment Variables - `EXPERT_SYSTEM_PATH`: Base path for expert system files - `NEO4J_URI`: Neo4j connection URI (default: bolt://localhost:7687) - `NEO4J_PASSWORD`: Neo4j authentication password - `CHROMA_PERSIST_PATH`: ChromaDB persistence directory - `EMBEDDING_MODEL`: Model for embeddings (default: all-MiniLM-L6-v2) ### Configuration Files - `pyproject.toml`: Package dependencies and tool configurations - `expert-registry.json`: Central registry of all experts - Expert context files: Markdown files with expert knowledge ## Development Workflow 1. **Setup Environment**: ```bash uv venv uv pip install -e ".[dev]" ``` 2. **Start Databases**: ```bash docker-compose up -d # Neo4j # ChromaDB is embedded ``` 3. **Run Tests**: ```bash pytest pytest --cov=expert_registry_mcp ``` 4. **Start Server**: ```bash fastmcp run expert-registry-mcp ``` ## Integration The server integrates with: - Claude Desktop via MCP protocol - Multi-agent workflows for expert-enhanced operations - CI/CD pipelines for automated testing - Monitoring systems for performance tracking