CodeWiki

CodeWiki is an open-source, AI-powered documentation engine that transforms source code repositories into structured, hierarchical, and navigable technical documentation — automatically.

Built by Flamingo as part of the OpenFrame platform, CodeWiki eliminates the documentation bottleneck in software engineering. It turns undocumented codebases into architecture-aware, LLM-generated documentation with zero manual effort.

---

Features

• Multi-language support — Python, JavaScript, TypeScript, Java, C, C++, C#, PHP
• AI-powered generation — LLM-generated documentation using Claude, GPT, or any OpenAI-compatible endpoint
• AST-based dependency analysis — Accurate call graphs extracted from actual source code
• Hierarchical, bottom-up generation — Leaf-first documentation that respects module structure
• CLI interface — One-command generation via codewiki generate
• Web interface — FastAPI-based web app for team-wide usage without requiring CLI setup
• Git-native workflows — Auto-creates timestamped documentation branches and commits
• GitHub Pages ready — Generates a static index.html viewer deployable to GitHub Pages
• LLM-agnostic — Works with OpenAI, Anthropic, Azure OpenAI, and any OpenAI-compatible proxy
• Context overflow protection — Staged pipeline and synthetic module fallback prevent runaway LLM calls
• Secure configuration — API keys stored in the OS keychain, never written to disk in plaintext

---

Architecture

CodeWiki runs as a multi-stage pipeline with clearly separated layers:

flowchart TD
    User["User (CLI or Web)"] --> CLICore["CLI Core"]
    User --> FECore["Frontend Core (FastAPI)"]

    CLICore --> DocGen["Documentation Generator"]
    FECore --> DocGen

    DocGen --> DepAnalysis["Dependency Analysis"]
    DocGen --> AgentOrch["Agent Orchestration"]

    AgentOrch --> LLMSvc["LLM Services"]
    DepAnalysis --> GraphBuilder["Dependency Graph Builder"]
    GraphBuilder --> DocGen

    DocGen --> Utils["Utils (FileManager)"]
    DocGen --> Output["Generated Documentation"]

    LLMSvc --> Provider["External LLM Providers"]

Loading

Generation Pipeline

flowchart TD
    Start["Start Job"] --> Analyze["1. Dependency Analysis (40%)"]
    Analyze --> Cluster["2. Module Clustering — LLM (20%)"]
    Cluster --> LeafDocs["3. Generate Leaf Module Docs (30%)"]
    LeafDocs --> ParentDocs["4. Generate Parent Module Docs"]
    ParentDocs --> Overview["5. Generate Repository Overview"]
    Overview --> Metadata["6. Write metadata.json (5%)"]
    Metadata --> EndNode["Done"]

Loading

---

Technology Stack

Layer	Technology
Language	Python 3.9+
CLI Framework	Click
Web Framework	FastAPI + Uvicorn
AI Agent Framework	Pydantic AI
AST Parsing	Tree-sitter (8 language grammars)
LLM Providers	OpenAI SDK, Anthropic SDK (provider-agnostic)
Secure Config	keyring (macOS Keychain / Windows Credential Manager / Linux Secret Service)
Git Integration	GitPython
Templating	Jinja2

---

Quick Start

1. Clone and install

git clone https://github.com/flamingo-stack/CodeWiki.git
cd CodeWiki
pip install -e .

2. Configure your LLM provider

codewiki config set \
  --main-model claude-sonnet-4 \
  --main-api-key YOUR_API_KEY \
  --main-base-url https://api.anthropic.com/v1 \
  --cluster-model claude-sonnet-4 \
  --cluster-api-key YOUR_API_KEY \
  --cluster-base-url https://api.anthropic.com/v1 \
  --fallback-model claude-sonnet-4 \
  --fallback-api-key YOUR_API_KEY \
  --fallback-base-url https://api.anthropic.com/v1

API keys are stored securely in your OS keychain — never written to disk in plaintext.

3. Generate documentation

cd /path/to/your-project
codewiki generate

Documentation is written to ./docs/ by default.

4. View the results

ls ./docs/

docs/
├── README.md              ← Repository overview
├── module_tree.json       ← Module structure metadata
├── metadata.json          ← Generation audit metadata
└── <module-name>/
    └── <module-name>.md   ← Per-module documentation

---

Using OpenAI

codewiki config set \
  --main-model gpt-4o \
  --main-api-key sk-... \
  --main-base-url https://api.openai.com/v1 \
  --cluster-model gpt-4o-mini \
  --cluster-api-key sk-... \
  --cluster-base-url https://api.openai.com/v1 \
  --fallback-model gpt-4o-mini \
  --fallback-api-key sk-... \
  --fallback-base-url https://api.openai.com/v1

Web Interface

Start the web UI to generate documentation for any public GitHub repository via a browser:

python -m codewiki.run_web_app

Then open http://localhost:8000, paste a GitHub repository URL, and CodeWiki will handle the rest.

---

Common Options

Flag	Description	Default
--output / -o	Output directory	./docs
--create-branch	Auto-create a Git branch	Off
--github-pages	Generate GitHub Pages index.html	Off
--include	Glob patterns to include	All files
--exclude	Glob patterns to exclude	None
--doc-type	Style: api, architecture, user-guide, developer	Default
--max-depth	Maximum module nesting depth	2

Example: CI/CD pipeline

codewiki generate \
  --create-branch \
  --github-pages \
  --output ./docs

Example: Focus on specific modules

codewiki generate \
  --include "src/core,src/api" \
  --exclude "tests/" \
  --doc-type architecture

---

Requirements

Software	Minimum Version
Python	3.9+
Git	2.x
pip	22+

An API key for at least one LLM provider (Anthropic, OpenAI, Azure OpenAI, or any OpenAI-compatible endpoint) is required.

---

Documentation

📚 See the Documentation for comprehensive guides including getting started tutorials, development setup, and full architecture reference.

---

Community

CodeWiki is part of the OpenMSP open-source community.

• Community Slack: OpenMSP Slack
• Community portal: openmsp.ai
• Flamingo Platform: flamingo.run
• OpenFrame: openframe.ai

---

Built with 💛 by the Flamingo team