search-hierarchy

The best way to choose a code search tool for token efficiency is to classify your query as structural, semantic, literal, or full context. This decision tree then routes you to AST-grep, LEANN, Grep, or Read to prevent wasted tokens.

To find structural code patterns using AST-grep, you identify your query as structural rather than literal. The search hierarchy recommends AST-grep for matching code patterns, reserving standard Grep only for exact identifiers.

You should use semantic search over Grep for conceptual code questions when your query requires understanding meaning rather than exact matches. The search hierarchy classifies these as semantic and recommends LEANN to find conceptual logic.

The search hierarchy requires no external dependencies to enforce tool selection. It integrates internally with a hook, specifically the grep-to-leann.sh script, to guide user behavior and ensure optimal tool usage without extra installations.

Using Read as a search method wastes tokens because it loads full context unnecessarily. The search hierarchy designates Read as a last resort, prioritizing AST-grep, LEANN, or Grep to improve search accuracy and minimize token consumption.