CodeWiki is an open-source framework designed to automate the generation of holistic, architecture-aware documentation for large-scale, multilingual codebases. It addresses the challenge of understanding complex software systems by capturing not just individual functions but also their interdependencies and system-level interactions, benefiting developers and researchers by providing a comprehensive, structured view of code.

How It Works

CodeWiki employs a three-stage process: Hierarchical Decomposition partitions codebases into modules using dynamic programming-inspired algorithms while preserving architectural context. A Recursive Multi-Agent System then processes these modules adaptively with dynamic task delegation, enabling scalability to large repositories. Finally, Multi-Modal Synthesis integrates textual descriptions with visual artifacts like architecture diagrams, data flows, and sequence diagrams for a comprehensive understanding. This approach is advantageous for handling arbitrary codebase sizes and maintaining quality across repository-level scope.

Quick Start & Requirements

• Install: pip install git+https://github.com/FSoft-AI4Code/CodeWiki.git
• Prerequisites: Python 3.12+, Node.js (for Mermaid diagram validation), LLM API access (e.g., Anthropic Claude, OpenAI), Git.
• Resources: Documentation & Guides, Docker Deployment, Development Guide, CodeWikiBench, Live Demo, Academic Paper.

Highlighted Details

• Supports seven programming languages: Python, Java, JavaScript, TypeScript, C, C++, C#.
• Generates comprehensive textual documentation and visual artifacts including system architecture diagrams, data flows, and sequence diagrams.
• Evaluated on CodeWikiBench, achieving an overall average of 68.79% documentation quality, a +4.73% improvement over DeepWiki, with notable gains in high-level languages.
• Handles large codebases (tested up to 1.4M LOC) through its hierarchical decomposition and recursive agentic system.

Licensing & Compatibility

Licensed under the MIT License, which generally permits commercial use and integration into closed-source projects.

Limitations & Caveats

Performance metrics indicate a slight decrease in documentation quality for Systems languages (C, C++) compared to a baseline, suggesting potential areas for further optimization in these categories.