exploring-codebases

Searching a multi-language codebase for complete functions combines fast text search with structural AST expansion to return the entire syntactic block containing each match. This hybrid approach de-duplicates results semantically, providing complete code blocks ready for review.

Finding classes across large repositories can use pre-generated MAP.md files to enable context expansion without tree-sitter. This map-based mode speeds up searches by relying on existing maps to return full implementations instead of parsing the AST on the fly.

Ripgrep provides fast text search to locate matches, while tree-sitter adds AST-aware context to expand those matches into full code blocks. This combination ensures that search results include complete function or class implementations rather than fragmented lines.

Yes, progressive disclosure returns only function signatures by default to reduce overhead. You can then selectively expand specific matches to view their full implementations as needed during the code search process.

Yes, hybrid codebase search applies to multi-language repositories. It locates functions, classes, and interfaces across different languages by combining ripgrep speed with structural awareness from either tree-sitter or pre-generated maps.

Use pre-generated maps for context expansion when searching very large codebases where performance is critical. Maps enable fast context expansion without the overhead of tree-sitter parsing, returning complete code blocks quickly.