Show HN: Semantic grep for Claude Code (RUST) (local embeddings) (github.com)

A new Rust-based tool called ck offers a semantic grep experience by allowing developers to search code by meaning rather than just keywords. Acting as a drop-in replacement for traditional grep, ck can locate code related to concepts like "error handling" or "user authentication" by understanding synonyms and contextual similarity, enabling it to find relevant code even when exact search terms aren’t present. This semantic search is powered by local embeddings and integrates fast indexing with tree-sitter parsing for languages including Python, JavaScript, TypeScript, and Haskell, making it particularly useful for complex codebases. Significantly, ck runs entirely offline, ensuring privacy and security, and produces JSON output ideal for AI agents and automation, enabling structured code analysis, documentation generation, and refactoring workflows. It supports hybrid searches that combine regex precision with semantic relevance using advanced ranking methods like Reciprocal Rank Fusion. The tool also preserves all familiar grep flags and behaviors, offering incremental indexing, intelligent file exclusion, and complete-function retrieval for deep insights. With sub-second query speeds on large projects and a modular architecture for embedding, indexing, and search engines, ck presents a powerful, privacy-conscious option for AI/ML developers and teams seeking more intuitive code search tailored to semantic meaning rather than raw text.