Facilitates finding code by its meaning and context rather than relying solely on text-based keyword matching.
Transforms your codebase into a semantically searchable index, allowing you to discover functions, classes, methods, and markdown sections based on their meaning. It leverages tree-sitter for AST-based chunking, sentence-transformers for encoding code chunks into vectors, and LanceDB for efficient, embedded storage and retrieval. This intelligent search capability significantly enhances developer productivity by moving beyond traditional grep/glob methods, offering ranked results by semantic relevance and supporting incremental indexing.
Key Features
01Incremental indexing based on file modification times
021 GitHub stars
03Semantic code search with natural language queries
04Hybrid semantic + keyword search with recency boosting
05Support for Python, Rust, and Markdown (extensible)
06Embedded vector store (LanceDB) for local operation
Use Cases
01Integrate advanced, meaning-based code search into AI agents and developer workflows
02Quickly locate relevant code snippets or functions using natural language descriptions
03Efficiently explore and understand large, unfamiliar codebases
