ChunkHound
Transforms codebases into searchable knowledge bases for AI assistants, providing both semantic and regex search capabilities via the Model Context Protocol.
소개
ChunkHound offers a modern Retrieval-Augmented Generation (RAG) solution for codebases, enabling developers and AI assistants to semantically and regex search through code. Built upon the research-backed cAST algorithm from Carnegie Mellon University, it intelligently chunks code to preserve meaning, leading to improved retrieval and generation metrics. Its local-first architecture ensures code privacy and offline functionality, supporting over 22 programming and configuration languages through Tree-sitter and custom parsers. With features like multi-hop semantic search and real-time indexing, ChunkHound facilitates intelligent code discovery and integration with various AI development environments.
주요 기능
- Research-backed cAST Algorithm for semantic code chunking
- Multi-Hop Semantic Search to discover interconnected code relationships
- Semantic and regex search for natural language and pattern queries
- Local-first architecture ensuring code privacy and offline functionality
- Real-time indexing with automatic updates, smart diffs, and branch switching detection
- 92 GitHub stars
사용 사례
- Transforming codebases into searchable knowledge bases for AI assistants.
- Performing intelligent code discovery, such as finding complete feature patterns (e.g., all components of "authentication").
- Creating dynamic knowledge bases by indexing real-time updated documentation and notes alongside code.