Deep Search Semantic Memory
Integrates local semantic search into large codebases to provide Claude with instant access to long-term documentation and source code context.
cm-deep-search is a specialized power-up for the CodyMaster framework that overcomes the limitations of AI context windows in massive projects. By bridging the gap between short-term working memory and long-term document retrieval, it allows Claude to leverage local semantic search (via qmd) to find relevant historical specs, PRDs, and code patterns across thousands of files. It features automated detection of oversized projects, zero-config suggestions, and incremental indexing to ensure the AI's knowledge base remains current without manual overhead.
主な機能
01Five-tier memory architecture linking working memory with external search
02Seamless MCP server support for direct AI-to-index communication
0345 GitHub stars
04Automated codebase size detection and semantic search triggering
05Incremental 'Post-Execution' sync to keep search results fresh
06Local vector and BM25 search integration using tobi/qmd
ユースケース
01Finding specific implementation details or architectural decisions in massive source repositories
02Syncing AI context with historical meeting notes and PRDs during the brainstorming phase
03Navigating and retrieving information from projects with hundreds of markdown documentation files