Vault Semantic

Vault Semantic is a local semantic search server designed as a sidecar for MCP-enabled agents like OpenClaw. It indexes Markdown files within a specified vault root, chunking content by headings and paragraphs, then embedding these chunks using OpenAI's `text-embedding-3-small` model. The tool offers a powerful hybrid search mechanism, combining FTS5 keyword search with cosine-similarity semantic search and folder-based ranking. Crucially, the original Markdown files remain the source of truth, with the SQLite database serving solely as a derived search index that can be rebuilt at any time. All functionalities, including search, retrieval, related notes, and reindexing, are exposed to agents via standard I/O (stdio) as MCP tools.

주요 기능

사용 사례