OpenDeepWiki Documentation

The backend is an ASP.NET Core application (targeting .NET 10) located in src/OpenDeepWiki/. It utilizes Minimal APIs for routing and Entity Framework Core for data access. Key service responsibilities include:

• IWikiGenerator: Orchestrates AI-driven catalog and content generation.
• IRepositoryAnalyzer: Handles Git operations (clone/pull) via LibGit2Sharp.
• IIncrementalUpdateService: Manages incremental update tasks.
• IChatAssistantService: Handles SSE streaming chat and document context.
• ITranslationService: Manages multi-language document translation.
• IEmbedService: Handles authentication and chat for embeddable widgets.

OpenDeepWiki uses a decoupled architecture consisting of a Next.js frontend and an ASP.NET Core backend.

• Frontend (web/): A Next.js application using the App Router, React components, custom hooks, and next-intl for internationalization. It uses an API proxy at /api/[...path] to communicate with the backend.
• Backend (src/OpenDeepWiki/): An ASP.NET Core service utilizing Minimal APIs. It handles Git repository analysis, AI-driven documentation generation, and data persistence.
• Data Layer: Powered by Entity Framework Core, supporting both SQLite and PostgreSQL.
• External Integrations: Connects to Git providers (GitHub, GitLab, Gitee) and AI providers (OpenAI, Azure, Anthropic).
• Background Processing: Long-running tasks like repository processing, translation, and mind map generation are handled by IHostedService background workers.

The OpenDeepWiki frontend is a Next.js 16 application located in the web/ directory. It utilizes the App Router to provide a user interface for repository browsing, documentation reading, AI chat interaction, and system management. The architecture is organized into several key layers:

• Routing: Uses Next.js App Router with nested layouts (Root, Main, Admin, and Dynamic Repository routes).
• Components: Organized by domain (UI primitives, Repository, Chat, Apps, Admin, and Animations).
• State & Logic: Managed via custom React hooks and React Context providers.
• Internationalization: Powered by next-intl supporting zh (default), en, ko, and ja.
• Styling: A utility-first approach using Tailwind CSS, Radix UI for accessible primitives, and Motion for animations.

The catalog-generator.md file is a system prompt designed for a senior code repository analyst AI. Its purpose is to guide an AI agent through the process of analyzing a codebase and generating a structured, 'DeepWiki-style' documentation catalog in JSON format.

Key objectives of this prompt include:

• Building a navigable wiki organized by topic domains (e.g., Architecture, Product Workflows, AI/Integrations) rather than file structures.
• Ensuring 'right-sized' coverage: avoiding both massive catch-all chapters and overly granular file-by-file pages.
• Creating a hierarchy where parent nodes serve as navigation domains and leaf nodes contain the actual documentation content.
• Verifying all catalog items against actual source code to prevent fabrication.

The Admin API provides system management capabilities including repository management, user management, role management, tool configuration, and statistics.

Base Path: /api/admin

Authentication Requirement: All Admin API endpoints require administrator privileges. Requests must include a valid JWT token containing an administrator role.

OpenDeepWiki follows a four-step process to transform code into a knowledge base:

1. Clone: The target repository is cloned locally.
2. Scan: The system recursively scans directories, respecting .gitignore files. It offers an option to use AI to filter large repositories.
3. Analyze: AI generates the directory structure, project overviews, and detailed documentation for every module.
4. Serve: Documentation is stored in a database and served via a Next.js frontend, supporting full-text search, MCP, and conversational AI.

To start the local development server for the web application, use one of the following package manager commands. Once running, the application will be accessible at http://localhost:3000. You can modify app/page.tsx to begin editing the application's main page, and changes will auto-update in the browser.

npm run dev
# or
yarn dev
# or
pnpm dev
# or
bun dev

To generate a valid mind map, the agent follows these steps:

1. Analyze Project Structure: Read entry point files to understand bootstrap processes, module organization, and dependencies.
2. Identify Core Components: Use discovery tools:
   • ListFiles("src/**/*", maxResults=50) to find modules.
   • ListFiles("**/config*", maxResults=20) to find configuration.
   • Grep("main|bootstrap|app", "**/*.{ts,js,cs,py,go}") to find entry points.
3. Build Architecture Map: Organize findings into: Entry Points, Core Business Logic, Data Layer, Infrastructure, and External Integrations.
4. Generate Mind Map: Create the hierarchical representation and call the WriteMindMap tool.