Summary

ck is a semantic grep tool for developers and AI agents, enabling code search by meaning, not just keywords. It enhances code discovery accuracy and facilitates automated code analysis by understanding conceptual similarity, offering a powerful alternative to traditional keyword-based search.

How It Works

ck employs multi-modal search: classic regex, pure semantic search via local embeddings, and a hybrid approach combining both. It uses Tree-sitter for language-aware parsing and chunking, allowing users to find code by concept while maintaining grep compatibility.

Quick Start & Requirements

Install via cargo install ck-search or build from source. Requires Rust/Cargo. Indexing is a one-time cost; searches are sub-500ms. Index size is ~2x source code. See GitHub for details.

Highlighted Details

• Semantic Search: Finds code by concept, synonyms, and related terms.
• Grep Compatibility: Supports familiar flags and output formats.
• Hybrid Search: Combines keyword precision with semantic relevance.
• Agent-Friendly Output: --json flag for AI agents and automation.
• Smart File Filtering: Excludes common directories and respects .gitignore.
• Full Section Extraction: --full-section retrieves entire functions/classes.
• File Listing: -l and -L flags list files with/without matches.

Maintenance & Community

Actively developed with a roadmap for future features. Welcomes contributions. No specific community channels or notable contributors are detailed.

Licensing & Compatibility

Dual-licensed under Apache 2.0 or MIT. All processing is local, ensuring offline functionality and data privacy.

Limitations & Caveats

Package manager distributions and configuration file support are planned features. Tree-sitter language coverage may vary for specific constructs.