OpenDeepWiki

中文 | English

AI-Driven Code Knowledge Base

---

enterprise service

Pricing of enterprise services

Our enterprise service offers comprehensive support and flexibility for businesses seeking professional AI solutions.

---

Features

• Quick Conversion: Supports converting all GitHub, GitLab, Gitee, Gitea and other code repositories into knowledge bases within minutes.
• Multi-language Support: Supports code analysis and documentation generation for all programming languages.
• Code Structure Diagrams: Automatically generates Mermaid diagrams to help understand code structure.
• Custom Model Support: Supports custom models and custom APIs for flexible extension.
• AI Intelligent Analysis: AI-based code analysis and code relationship understanding.
• SEO Friendly: Generates SEO-friendly documentation and knowledge bases based on Next.js for easy search engine crawling.
• Conversational Interaction: Supports conversations with AI to obtain detailed code information and usage methods for deep code understanding.

---

Feature List

• Support multiple code repositories (GitHub, GitLab, Gitee, Gitea, etc.)
• Support multiple programming languages (Python, Java, C#, JavaScript, etc.)
• Support repository management (CRUD operations on repositories)
• Support multiple AI providers (OpenAI, AzureOpenAI, Anthropic, etc.)
• Support multiple databases (SQLite, PostgreSQL, SqlServer, MySQL, etc.)
• Support multiple languages (Chinese, English, French, etc.)
• Support uploading ZIP files and local files
• Provide data fine-tuning platform to generate fine-tuning datasets
• Support directory-level repository management with dynamic directory and document generation
• Support repository directory modification management
• Support user management (CRUD operations on users)
• Support user permission management
• Support repository-level generation of different fine-tuning framework datasets

---

Project Introduction

OpenDeepWiki is an open-source project inspired by DeepWiki, developed based on .NET 9 and Semantic Kernel. It aims to help developers better understand and utilize code repositories, providing features such as code analysis, documentation generation, and knowledge graph construction.

Main Features:

• Analyze code structure
• Understand repository core concepts
• Generate code documentation
• Automatically generate README.md for code
• Support MCP (Model Context Protocol)

---

MCP Support

OpenDeepWiki supports the MCP protocol:

• Can serve as a single repository MCPServer for repository analysis.

Example configuration:

{
  "mcpServers": {
    "OpenDeepWiki":{
      "url": "http://Your OpenDeepWiki service IP:port/api/mcp?owner=AIDotNet&name=OpenDeepWiki"
    }
  }
}

If mcp streamable http is not supported, use the following format:

{
  "mcpServers": {
    "OpenDeepWiki":{
      "url": "http://Your OpenDeepWiki service IP:port/api/mcp/sse?owner=AIDotNet&name=OpenDeepWiki"
    }
  }
}

MCP Streamable Configuration

You can configure MCP streamable support for specific services using the MCP_STREAMABLE environment variable:

environment:
  # Format: serviceName1=streamableUrl1,serviceName2=streamableUrl2
  - MCP_STREAMABLE=claude=http://localhost:8080/api/mcp,windsurf=http://localhost:8080/api/mcp

This allows you to specify which services should use streamable HTTP endpoints and their corresponding URLs.

• owner: Repository organization or owner name
• name: Repository name

After adding the repository, you can test by asking questions like "What is OpenDeepWiki?", with effects as shown below:

This way, OpenDeepWiki can serve as an MCPServer for other AI models to call, facilitating analysis and understanding of open-source projects.

---

🚀 Quick Start

1. Clone the repository

git clone https://github.com/AIDotNet/OpenDeepWiki.git
cd OpenDeepWiki

2. Modify environment variable configuration in docker-compose.yml:

• OpenAI example:

services:
  koalawiki:
    environment:
      - TASK_MAX_SIZE_PER_USER=2 # Maximum parallel document generation tasks per user for AI
      - CHAT_MODEL=DeepSeek-V3 # Model must support function calling
      - ANALYSIS_MODEL= # Analysis model for generating repository directory structure
      - CHAT_API_KEY= # Your API Key
      - LANGUAGE= # Default generation language, e.g., "Chinese"
      - ENDPOINT=https://api.token-ai.cn/v1
      - DB_TYPE=sqlite
      - MODEL_PROVIDER=OpenAI # Model provider, supports OpenAI, AzureOpenAI
      - DB_CONNECTION_STRING=Data Source=/data/KoalaWiki.db
      - EnableSmartFilter=true # Whether to enable smart filtering, affects AI's ability to get repository file directories
      - UPDATE_INTERVAL=5 # Repository incremental update interval in days
      - MAX_FILE_LIMIT=100 # Maximum upload file limit in MB
      - DEEP_RESEARCH_MODEL= # Deep research model, if empty uses CHAT_MODEL
      - ENABLE_INCREMENTAL_UPDATE=true # Whether to enable incremental updates
      - ENABLE_CODED_DEPENDENCY_ANALYSIS=false # Whether to enable code dependency analysis, may affect code quality
      - ENABLE_WAREHOUSE_COMMIT=true # Whether to enable warehouse commit
      - ENABLE_FILE_COMMIT=true # Whether to enable file commit
      - REFINE_AND_ENHANCE_QUALITY=false # Whether to refine and enhance quality
      - CATALOGUE_FORMAT=compact # Directory structure format (compact, json, pathlist, unix)
      - CUSTOM_BODY_PARAMS= # Custom request body parameters, format: key1=value1,key2=value2 (e.g., stop=<|im_end|>,max_tokens=4096)
      - READ_MAX_TOKENS=100000 # The maximum token limit for reading files in AI is set to prevent unlimited file reading. It is recommended to fill in 70% of the model's maximum token.
      - MCP_STREAMABLE= # MCP service streamable configuration, format: serviceName=streamableUrl (e.g., claude=http://localhost:8080/api/mcp,windsurf=http://localhost:8080/api/mcp)
      # Auto Context Compression configuration (optional)
      - AUTO_CONTEXT_COMPRESS_ENABLED=false # Whether to enable AI-powered intelligent context compression
      - AUTO_CONTEXT_COMPRESS_TOKEN_LIMIT=100000 # Token limit to trigger compression (required when enabled)
      - AUTO_CONTEXT_COMPRESS_MAX_TOKEN_LIMIT=200000 # Maximum allowed token limit (default: 200000)
      # Feishu Bot configuration (optional)
      - FeishuAppId=
      - FeishuAppSecret=
      - FeishuBotName=KoalaWiki

• AzureOpenAI and Anthropic configurations are similar, only need to adjust ENDPOINT and MODEL_PROVIDER.

Database Configuration

SQLite (Default)

- DB_TYPE=sqlite
- DB_CONNECTION_STRING=Data Source=/data/KoalaWiki.db

PostgreSQL

- DB_TYPE=postgres
- DB_CONNECTION_STRING=Host=localhost;Database=KoalaWiki;Username=postgres;Password=password

SQL Server

- DB_TYPE=sqlserver
- DB_CONNECTION_STRING=Server=localhost;Database=KoalaWiki;Trusted_Connection=true;

MySQL

- DB_TYPE=mysql
- DB_CONNECTION_STRING=Server=localhost;Database=KoalaWiki;Uid=root;Pwd=password;

3. Start services

Using Makefile commands:

# Build all Docker images
make build

# Start all services in background
make up

# Start in development mode (with visible logs)
make dev

Visit http://localhost:8090 to access the knowledge base page.

For Windows users without make environment, use Docker Compose directly:

docker-compose build
docker-compose up -d
docker-compose up
docker-compose down
docker-compose logs -f

---

Deployment Recommendations

• Build for specific architecture:

docker-compose build --build-arg ARCH=arm64
docker-compose build --build-arg ARCH=amd64

• Build only backend or frontend:

docker-compose build koalawiki
docker-compose build koalawiki-web

• One-click deployment to Sealos (supports public network access):

For detailed steps, please refer to: One-click Sealos Deployment of OpenDeepWiki

---

🔍 How It Works

OpenDeepWiki leverages AI to achieve:

• Clone code repository locally
• Read .gitignore configuration to ignore irrelevant files
• Recursively scan directories to get all files and directories
• Determine if file count exceeds threshold; if so, call AI model for intelligent directory filtering
• Parse AI-returned directory JSON data
• Generate or update README.md
• Call AI model to generate repository classification information and project overview
• Clean project analysis tag content and save project overview to database
• Call AI to generate thinking directory (task list)
• Recursively process directory tasks to generate document directory structure
• Save directory structure to database
• Process incomplete document tasks
• If Git repository, clean old commit records, call AI to generate update log and save

---

OpenDeepWiki Repository Parsing to Documentation Detailed Flow Chart

graph TD
    A[Clone code repository] --> B[Read .gitignore configuration to ignore files]
    B --> C[Recursively scan directories to get all files and directories]
    C --> D{Does file count exceed threshold?}
    D -- No --> E[Directly return directory structure]
    D -- Yes --> F[Call AI model for intelligent directory structure filtering]
    F --> G[Parse AI-returned directory JSON data]
    E --> G
    G --> H[Generate or update README.md]
    H --> I[Call AI model to generate repository classification information]
    I --> J[Call AI model to generate project overview information]
    J --> K[Clean project analysis tag content]
    K --> L[Save project overview to database]
    L --> M[Call AI to generate thinking directory task list]
    M --> N[Recursively process directory tasks to generate DocumentCatalog]
    N --> O[Save directory structure to database]
    O --> P[Process incomplete document tasks]
    P --> Q{Is repository type Git?}
    Q -- Yes --> R[Clean old commit records]
    R --> S[Call AI to generate update log]
    S --> T[Save update log to database]
    Q -- No --> T

Loading

---

Advanced Configuration

Environment Variables

• KOALAWIKI_REPOSITORIES: Repository storage path
• TASK_MAX_SIZE_PER_USER: Maximum parallel document generation tasks per user for AI
• CHAT_MODEL: Chat model (must support function calling)
• ENDPOINT: API endpoint
• ANALYSIS_MODEL: Analysis model for generating repository directory structure
• CHAT_API_KEY: API key
• LANGUAGE: Document generation language
• DB_TYPE: Database type, supports sqlite, postgres, sqlserver, mysql (default: sqlite)
• MODEL_PROVIDER: Model provider, default OpenAI, supports AzureOpenAI, Anthropic
• DB_CONNECTION_STRING: Database connection string
• EnableSmartFilter: Whether to enable smart filtering, affects AI's ability to get repository directories
• UPDATE_INTERVAL: Repository incremental update interval (days)
• MAX_FILE_LIMIT: Maximum upload file limit (MB)
• DEEP_RESEARCH_MODEL: Deep research model, if empty uses CHAT_MODEL
• ENABLE_INCREMENTAL_UPDATE: Whether to enable incremental updates
• ENABLE_CODED_DEPENDENCY_ANALYSIS: Whether to enable code dependency analysis, may affect code quality
• ENABLE_WAREHOUSE_COMMIT: Whether to enable warehouse commit
• ENABLE_FILE_COMMIT: Whether to enable file commit
• REFINE_AND_ENHANCE_QUALITY: Whether to refine and enhance quality
• ENABLE_WAREHOUSE_FUNCTION_PROMPT_TASK: Whether to enable warehouse function prompt task
• ENABLE_WAREHOUSE_DESCRIPTION_TASK: Whether to enable warehouse description task
• CATALOGUE_FORMAT: Directory structure format (compact, json, pathlist, unix)
• ENABLE_CODE_COMPRESSION: Whether to enable code compression
• CUSTOM_BODY_PARAMS: Custom request body parameters, format: key1=value1,key2=value2 (e.g., stop=<|im_end|>,max_tokens=4096). These parameters will be added to all AI model API requests
• READ_MAX_TOKENS: Maximum token limit for reading files in AI, prevents unlimited file reading. It is recommended to fill in 70% of the model's maximum token (default: 100000)
• MAX_FILE_READ_COUNT: Maximum file read count limit for AI, prevents unlimited file reading and improves processing efficiency (default: 10, 0 = no limit)
• AUTO_CONTEXT_COMPRESS_ENABLED: Whether to enable AI-powered intelligent context compression for long conversations (default: false)
• AUTO_CONTEXT_COMPRESS_TOKEN_LIMIT: Token threshold to trigger context compression. Required when compression is enabled (default: 100000)
• AUTO_CONTEXT_COMPRESS_MAX_TOKEN_LIMIT: Maximum allowed token limit, ensures the token limit doesn't exceed model capabilities (default: 200000)

Intelligent Context Compression Features: Uses Prompt Encoding Compression - an ultra-dense, structured format that achieves 90%+ compression while preserving ALL critical information.

Compression Strategy:

100 messages (50k tokens) → 1 encoded snapshot (3k tokens)
Compression ratio: 94% ✨

What Gets Preserved (100%):

• System Messages: All system-level instructions
• Function Calls & Results: Complete tool invocation history (preserves core behavior)
• Recent Conversation: Most recent 30% of messages in original form

What Gets Encoded (Older Messages): Instead of selecting/deleting messages, older messages are compressed into an ultra-dense structured snapshot:

## CONTEXT_SNAPSHOT
### FILES
✓ src/File.cs:modified(L:25-48) → README.md:pending

### TASKS
✓ Implement feature X ✓ Fix bug Y → Add tests (pending)

### TECH_STACK
IChatClient, Semantic Kernel, AutoContextCompress, TokenHelper

### DECISIONS
[D1] Use message filtering: preserve structure
[D2] Keep 30% recent: based on Google Gemini best practice

### CODE_PATTERNS
```cs
if (message.Contents.Any(c => c is FunctionCallContent)) { ... }

USER_INTENT

Enable configurable compression via env vars. Must preserve core behavior.

NEXT_ACTIONS

1. Update documentation 2. Add unit tests

**Encoding Format Features:**
- ✓ Ultra-dense: Uses symbols (✓=done, →=pending, ✗=blocked)
- ✓ Structured: 8 semantic sections (FILES, TASKS, TECH_STACK, etc.)
- ✓ Precise: Preserves file paths, line numbers, function names, decisions
- ✓ Actionable: Clear next steps for AI to continue work
- ✓ Lossless: All critical information encoded, zero loss

**Key Benefits:**
- ✅ 90-95% compression ratio (vs 30-40% with message filtering)
- ✅ Zero loss of function calls and results
- ✅ Maintains temporal context (recent messages untouched)
- ✅ AI can reconstruct full understanding from snapshot
- ✅ One snapshot replaces hundreds of messages
- `FeishuAppId`: Feishu App ID (required if enabling Feishu Bot)
- `FeishuAppSecret`: Feishu App Secret (required if enabling Feishu Bot)
- `FeishuBotName`: Feishu bot display name (optional)

---

# Feishu Bot Integration

- Purpose: Connect the current repository to Feishu group/DM as a knowledge bot for Q&A and content delivery.
- Callback route: `/api/feishu-bot/{owner}/{name}` (copy the full URL from the "Feishu Bot" button on the repository page).
- Requirement: Service must be publicly accessible; message encryption (Encrypt Key) is not supported yet.

## 1) Create an App in Feishu Open Platform

- Type: Internal App (for your organization).
- Capability: Enable "Bot". Publish to the organization and install it.
- Permissions (at minimum, per platform guidance):
  - Message send/read related scopes (e.g., `im:message`, `im:message:send_as_bot`).
  - Event subscription related scopes (for receiving message events).

## 2) Configure Event Subscriptions (Important)

- Open "Event Subscriptions" and disable "Encrypt Key".
- Subscribe to event: `im.message.receive_v1`.
- Request URL: `https://your-domain/api/feishu-bot/{owner}/{name}`.
  - `{owner}` is the repository organization or owner, e.g., `AIDotNet`.
  - `{name}` is the repository name, e.g., `OpenDeepWiki`.
- Save to complete the URL verification (backend already handles the challenge response).

Tip: You can also click the "Feishu Bot" button on the repository page to copy the dedicated callback URL.

## 3) Configure Server Environment Variables

Set the following in your backend service (see docker-compose example below):

- `FeishuAppId`: Feishu App ID
- `FeishuAppSecret`: Feishu App Secret
- `FeishuBotName`: Bot display name (optional)

## 4) Add the Bot to a Group and Use It

- After installing the app, add the bot to the target group.
- Group: @bot + your question (answers using the current repository's knowledge).
- Direct Message: send your question directly.
- Supports text and image replies (e.g., mind map images).

## Feishu FAQ

- No response/callback failed: ensure the Request URL is publicly reachable and that Nginx proxies `/api/` to the backend.
- "Encryption enabled" message: disable Encrypt Key (current version doesn't support encrypted messages).
- 403/insufficient permissions: make sure the app is installed in the org and required scopes/events are granted.

## Build for Different Architectures

Makefile commands:

```bash
make build-arm    # ARM architecture
make build-amd    # AMD architecture
make build-backend-arm   # Backend only ARM
make build-frontend-amd  # Frontend only AMD

---

Community

• Discord: join us
• Feishu QR Code:

---

📄 License

This project is licensed under the MIT License. See LICENSE for details.

---

⭐ Star History