Deep Search & Semantic Memory
Enhances Claude’s codebase navigation with local semantic search and long-term document retrieval for massive projects.
cm-deep-search is an advanced memory extension for the CodyMaster framework that integrates with the tobi/qmd engine to provide 100% local, AI-powered semantic search. Designed for projects that exceed standard context windows, it automatically detects oversized codebases and suggests indexing to enable BM25 and vector-based retrieval. By bridging the gap between active working memory and long-term documentation, it ensures Claude can find specific historical decisions, specs, and code patterns across thousands of files without losing context or accuracy.
주요 기능
0145 GitHub stars
02Zero-config detection of oversized projects based on file count and content triggers
03Tiered memory architecture that complements active working memory with external semantic disk search
04Automated staleness prevention via post-execution indexing and git hook support
05Seamless integration with MCP servers for Claude, Cursor, and Antigravity
06Local semantic search using BM25, Vector embeddings, and LLM re-ranking via qmd
사용 사례
01Retrieving specific historical architectural decisions during the planning and brainstorming phases
02Navigating massive legacy codebases with hundreds of source files and complex dependencies
03Searching through extensive project documentation, meeting notes, and historical PRDs