Japanese Full-Text Search

About

This Model Context Protocol (MCP) server delivers advanced Japanese full-text search capabilities, significantly outperforming traditional `grep` for AI applications. It meticulously indexes local text files, employing SudachiPy for precise Japanese tokenization and SQLite FTS5 for lightning-fast, highly relevant search results. Key features include accurate line numbering even within tokenized text, automatic indexing of directory changes, `.gitignore` support, and relevance scoring with contextual snippets, making it an indispensable tool for LLMs requiring nuanced understanding and rapid retrieval of Japanese content.

Key Features

• High-accuracy Japanese full-text search using SudachiPy for precise tokenization.
• Indexes local text files, supporting recursive directory scanning, `.gitignore` filtering, and extension-based targeting.
• Automates index cleanup by removing entries for deleted files during re-indexing.
• Provides precise line numbering for search results within tokenized Japanese text using a token map strategy.
• 0 GitHub stars
• Integrates with FastMCP to expose directory indexing and document search as accessible MCP tools.

Use Cases

• Empowering LLMs to perform high-precision, context-aware Japanese document searches in local repositories.
• Rapidly indexing and searching extensive local documentation or codebases written in Japanese.
• Providing AI agents with an efficient and accurate alternative to `grep` for retrieving Japanese text with linguistic understanding and relevance ranking.