ccg-rag

Semantic code search lets you describe what code does rather than search for exact text. This Skill uses embeddings and language-aware chunking with tree-sitter to match queries like "authentication middleware" to relevant functions and files across your codebase, returning results by meaning instead of keyword overlap.

Yes. This Skill builds a knowledge graph of function calls and dependencies, then maps your natural language queries to relevant code patterns and documentation. Joining a new project, you can query "user authentication login session" and instantly retrieve related files, documentation, and call hierarchies instead of manually exploring the repository.

Keyword search matches exact terms; semantic search understands intent and functionality. This Skill implements hybrid search combining both approaches with embeddings, so a query about "validation logic" surfaces similar implementations across your codebase even if they use different function names or variable conventions.

It chunks code intelligently using tree-sitter for language awareness, indexes embeddings for semantic retrieval, and constructs a knowledge graph of function relationships and imports. This approach scales across large codebases by structuring search and dependency traversal hierarchically rather than linearly scanning all files.

Yes. Pattern search discovers similar implementations of functions or logic throughout your repository. Query "functions similar to validateUser" to locate related validation patterns, promoting code consistency and enabling reuse of proven approaches across teams and modules.

Yes. This Skill performs semantic search across both code files and documentation. Query "API documentation for payments" to retrieve relevant technical docs and code examples together, centralizing knowledge access and reducing time spent switching between repositories and documentation sites.