Knowledge Retrieval Server

# CLAUDE.md This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository. ## Project Overview This is an MCP (Model Context Protocol) server for knowledge retrieval using BM25 algorithm. The server provides document search and retrieval capabilities from organized knowledge domains. ### Core Architecture - **MCP Server**: Built on `@modelcontextprotocol/sdk`, provides tools for document search and retrieval - **Document Processing**: Markdown documents are chunked and indexed using BM25 algorithm - **Domain Organization**: Documents are organized by domains (company, customer, product, technical) - **Search Engine**: BM25-based semantic search with configurable parameters ### Key Components - `src/index.ts`: Main MCP server entry point with tool handlers - `src/services/DocumentRepository.ts`: Core search and document management using BM25 - `src/services/DocumentLoader.ts`: Document loading from local/remote sources - `src/models/Document.ts`: Document and chunk data structures - `src/utils/bm25.ts`: BM25 algorithm implementation - `src/config/config.ts`: Configuration management with environment variable support ## Development Commands ```bash # Build the project npm run build # Start production server npm start # Development mode with hot reload npm run dev # Run tests npm test # Lint code npm run lint ``` ## Configuration The server uses `config.json` for configuration (see `config.example.json`). Key settings: - Document source paths and domains - BM25 parameters (k1, b) - Chunk settings (minWords, contextWindowSize) Environment variables can override config file settings (see `src/config/config.ts:26-42`). ## Document Structure Documents are organized in domains under the configured base path: - Each domain has a name, path, and category - Supports both local files and remote sources (via llms.txt) - Documents are automatically chunked for optimal search performance ## MCP Tools Available 1. `search-documents`: BM25-based document search with keyword arrays 2. `get-document-by-id`: Retrieve full document by ID 3. `list-domains`: Get available domains and document counts 4. `get-chunk-with-context`: Get specific chunks with surrounding context ## Testing and Quality - TypeScript with strict mode enabled - ESLint configuration for code quality - Jest for testing framework - All code includes Korean comments for documentation