corpus-rag-setup

To set up local semantic search, run the corpus-rag-setup skill to initialize an index manifest for your project directory using the mcp-local-rag server, filtering out build artifacts and scratch files to index only relevant documentation.

Semantic search indexing intelligently filters out noise like scratch files, archives, and build artifacts, ensuring only relevant content such as project documentation, strategy files, and canonical records are indexed for natural language querying.

Yes, you need npm for package management to configure and maintain the local semantic search index, which integrates directly with the Claude Code MCP framework and the mcp-local-rag server.

Yes, you can query your project's design decisions using natural language. Semantic search enables you to search your project corpus to find specific context, history, and specs from indexed documentation.

Index quality is maintained through a structured reconciliation process that scans project directories, manages an index manifest, and ensures the semantic search index accurately reflects relevant documentation while excluding irrelevant noise.

To manage fragmented project knowledge, use local semantic search to index your documentation and files, allowing the AI to query your entire project history and canonical records using natural language instead of manual searching.