Turn source code into a living technical wiki.

About

Developers spend nearly 58% of their time understanding codebases, yet maintaining comprehensive documentation remains challenging. While recent Large Language Models (LLMs) show promise for function-level documentation, they fail at the repository level, where capturing architectural patterns and cross-module interactions is essential.

CodeWiki is an open-source framework for holistic repository-level documentation across eleven programming languages. It builds the dependency graph, decomposes it into a module hierarchy, and has recursive agents write text and diagrams. Version 2.0 makes the graph more complete, documents build, CI, container, and configuration files alongside code, and keeps documentation current with component-level incremental updates.

Framework Architecture

Figure 1: CodeWiki Framework operates in three main phases: (1) Repository analysis and hierarchical decomposition, (2) Recursive documentation generation with dynamic delegation, (3) Hierarchical assembly and synthesis

Key Innovations

Multilingual Support

Benchmark Results

Evaluated on CodeWikiBench: seven repositories, 486 rubric requirements, the same LLM judge for every system. Scores are weighted rubric scores from 0 to 100. DeepWiki and CodeWiki 1.0 were re-run under the same setting as 2.0 (September 2026), so they differ from the numbers in the paper. README files and docs/ were excluded from CodeWiki's inputs.

Score by Repository

Score by Topic

Figure 2: (a) how much the 2.0 code graph grew over 1.0; (b) nodes and edges added by artifacts; (c) score by stage. "Conference version" is CodeWiki 1.0, "C1 + C2" is CodeWiki 2.0.

The 21-repository study of CodeWiki 1.0 is in the paper.

Demo Video

Watch CodeWiki in action as it generates comprehensive documentation for a real repository:

CLI Usage Example: Generating documentation with CodeWiki

Installation & Quick Start

Prerequisites

• Python 3.12+
• Node.js and npm at install time
• Git
• An LLM API key, or a Claude Code / Codex subscription, or an MCP-capable AI IDE

Installation

# Install from source
pip install git+https://github.com/FSoft-AI4Code/CodeWiki.git

# Verify installation
codewiki --version

Quick Start

1. Configure CodeWiki:

# Any OpenAI-compatible endpoint
codewiki config set \
  --provider openai-compatible \
  --api-key YOUR_API_KEY \
  --base-url https://api.example.com/v1 \
  --main-model claude-sonnet-4 \
  --cluster-model claude-sonnet-4

# Or run on your Claude Code subscription, no API key
codewiki config set --provider claude-code \
  --main-model claude-sonnet-4-6 --cluster-model claude-sonnet-4-6

# Verify configuration
codewiki config show
codewiki config validate

2. Generate Documentation:

# Navigate to your project
cd /path/to/your/project

# Generate documentation (saved to ./docs/)
codewiki generate

# Generate with GitHub Pages HTML viewer
codewiki generate --github-pages

# Later: refresh only what changed
codewiki generate --update

# Full-featured generation
codewiki generate --create-branch --github-pages --verbose

Output Structure

./docs/
├── overview.md              # Repository overview (start here!)
├── module1.md               # Module documentation
├── module2.md               # Additional modules...
├── module_tree.json         # Hierarchical module structure
├── first_module_tree.json   # Initial clustering result
├── metadata.json            # Generation metadata
├── update_record.json       # Decisions of the last --update run
├── temp/                    # Saved dependency graph and artifact index
└── index.html               # Interactive viewer (with --github-pages)

Citation

If you use CodeWiki in your research, please cite our paper:

@inproceedings{hoang-etal-2026-codewiki,
    title = "{C}ode{W}iki: Evaluating {AI}{'}s Ability to Generate Holistic Documentation for Large-Scale Codebases",
    author = "Hoang, Anh Nguyen and Le-Anh, Minh and Le, Bach and Bui, Nghi D. Q.",
    booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2026.findings-acl.288/",
    doi = "10.18653/v1/2026.findings-acl.288",
    pages = "5812--5827",
}

Resources