<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> Codanna equips AI coding assistants with deep codebase understanding, enabling precise function navigation, relationship tracing, and implementation discovery. It targets engineers and AI developers seeking to boost productivity by minimizing manual code exploration and enhancing AI-driven code generation through context-first interactions. The core benefit is smarter, faster engineering with reduced wasted effort.

How It Works

Codanna leverages Tree-sitter for rapid AST parsing across supported languages (Rust, Python, TypeScript, Go, PHP). It extracts functions, traits, and call graphs, then builds semantic vectors from documentation comments. Indexing uses Tantivy with a memory-mapped symbol cache for sub-10ms lookups. Data is served via the MCP protocol for seamless AI assistant integration, prioritizing context and efficiency over imprecise searches.

Quick Start & Requirements

Install via cargo install codanna --all-features. Setup involves codanna init and codanna index src --progress. Development requires Rust 1.75+. Linux systems need pkg-config and libssl-dev/openssl-devel; macOS/Windows have no extra system dependencies. Model storage requires ~150MB.

Highlighted Details

• Performance: Sub-0.16s semantic query responses. Rust parsing exceeds 91k symbols/sec. Memory usage is ~100 bytes/symbol.
• AI Integration: Native MCP protocol support for AI assistants, including Claude slash commands (/find, /deps) and agent steering via system messages.
• Unix-Native CLI: Robust CLI integrates with shell pipelines using --json output for structured data.
• Indexing: Efficiently indexes codebases respecting .gitignore, using single-pass AST traversal for symbol, relationship, and embedding extraction.

Maintenance & Community

Developed by "devs throttled by tools that 'understand' code only in theory." Specific contributor details, community channels, or sponsorship information are not provided in the excerpt. A roadmap is mentioned but not detailed.

Licensing & Compatibility

Licensed under the Apache License, Version 2.0. Attribution is required via the NOTICE file.

Limitations & Caveats

Current language support includes Rust, Python, TypeScript, Go, PHP, C, and C++; JavaScript, C#, and Java are planned. Semantic search requires English documentation/comments. Windows support is experimental.