Content
# Tool List
[English Version](README_EN.md) |
## 📖 Project Introduction
XiaoyaoSearch is a cross-platform local desktop application (Windows/MacOS/Linux) designed for knowledge workers, content creators, and technical developers. Through integrated AI models, it supports multiple input methods such as voice input (within 30 seconds), text input, and image input, converting user queries into semantic intelligence for local file deep retrieval.
## ⭐️ Important Notes
- This project is completely free for non-commercial use, allowing modification and distribution (with copyright declaration and agreement retained); commercial use requires authorization, see [XiaoyaoSearch Software Authorization Agreement](LICENSE) for details.
- This project is fully implemented through Vibe Coding, providing all source code and development documents (context) for everyone to learn and exchange.
## Author Introduction
<p align="center">
<img src="docs/产品文档/产品截图/author-avatar.jpg" alt="dtsola" width="120" height="120" style="border-radius: 50%;">
</p>
<p align="center">
<b>dtsola</b> — IT Architect | Solo Company Practitioner
</p>
<p align="center">
🌐 <a href="https://www.dtsola.com">Personal Station</a> |
📺 <a href="https://space.bilibili.com/736015">Bilibili</a> |
💬 WeChat: dtsola (Technical Exchange | Business Cooperation)
</p>
<p align="center">
<img src="docs/产品文档/产品截图/个人二维码.png" alt="WeChat QR Code" width="120">
<img src="docs/运营文档/开发者交流群图.png" alt="Developer Exchange Group" width="120">
<img src="docs/运营文档/用户交流群图.png" alt="User Exchange Group" width="120">
</p>
<p align="center">
<small>WeChat Contact Developer Exchange Group User Exchange Group</small>
</p>
### ✨ Core Features
- **🎤 Multimodal Input**: Supports voice recording, text input, and image upload.
- **🔍 Deep Search**: Supports content and filename search for videos (mp4, avi), audio (mp3, wav), and documents (txt, markdown, office, pdf).
- **🧠 AI Enhanced**: Integrates advanced AI models such as BGE-M3, FasterWhisper, CN-CLIP, and OLLAMA.
- **☁️ Cloud Large Model**: Supports OpenAI/DeepSeek/Aliyun compatible API, freely switching with local models (v1.3.0).
- **☁️ Cloud Embedding Model**: Supports OpenAI/DeepSeek/Aliyun embedding API to improve search quality (v1.6.0).
- **📚 Professional Terminology Library**: Supports terminology extension and synonym matching, improving professional field search recall rate by 60% (v1.9.0).
- **⚡ High Performance**: Based on Faiss vector search and Whoosh full-text search hybrid retrieval architecture.
- **🔒 Privacy Controllable**: Local operation defaults to no data upload, supporting cloud API options, and performance and privacy trade-offs are up to you.
- **🎨 Notion Warm and Bright Design**: Adopts Notion warm and bright design style, establishing a complete design system specification, improving visual comfort by 67% (v2.0.0).
- **System Font Stack**: Zero network dependency, optimal performance, font loading < 10ms.
- **Lucide Icons**: 1000+ icons, unified visual language.
- **Exquisite Multi-layer Shadows**: 4-5 layer shadow stacking, enhancing visual hierarchy.
- **🤖 AI Ecosystem Integration**:
- **MCP Server Support**: Supports Model Context Protocol, can be connected by Claude Desktop for local file intelligent search (v1.4.0).
- **Agent Skills Support**: Provides standardized tool calling capabilities for AI assistants like Claude Code, VS Code, and Cursor (v1.5.0).
## 📖 Core Interface
### Search Interface
#### Main Interface
#### Text Search
#### Voice Search
#### Image Search
### Index Management Interface
### Settings Interface
### Terminology Library Management Interface (v2.0.0)
## 🏗️ Technical Architecture
### System Architecture Diagram
### Technology Stack
**Frontend Technology**
- **Framework**: Electron + Vue 3 + TypeScript
- **UI Library**: Ant Design Vue
- **State Management**: Pinia
- **Build Tool**: Vite
**Backend Technology**
- **Framework**: Python 3.10 + FastAPI + Uvicorn
- **AI Model**: BGE-M3 + FasterWhisper + CN-CLIP + Ollama
- **Search Engine**: Faiss (vector search) + Whoosh (full-text search)
- **Database**: SQLite + Index Files
### Project Structure
```
xiaoyaosearch/
├── backend/ # Backend Service (Python FastAPI)
│ ├── app/ # Application Core Code
│ │ ├── api/ # API Route Layer
│ │ ├── core/ # Core Configuration
│ │ ├── models/ # Data Models
│ │ ├── services/ # Business Services
│ │ ├── schemas/ # Data Schemas
│ │ └── utils/ # Utility Functions
│ ├── requirements.txt # Python Dependencies
│ ├── main.py # Application Entry
│ └── .env # Environment Variables
├── frontend/ # Frontend Application (Electron + Vue3)
│ ├── src/ # Source Code
│ │ ├── main/ # Electron Main Process
│ │ ├── preload/ # Preload Script
│ │ └── renderer/ # Vue Render Process
│ ├── out/ # Build Output
│ ├── dist-electron/ # Packaging Output
│ ├── resources/ # Application Resources
│ ├── package.json # Node.js Dependencies
│ └── electron-builder.yml # Packaging Configuration
├── docs/ # Project Documents
│ ├── 00-mrd.md # Market Research
│ ├── 01-prd.md # Product Requirements
│ ├── 02-prototype.md # Product Prototype
│ ├── 03-technical-solution.md # Technical Solution
│ ├── 04-development-task-list.md # Development Task List
│ ├── 05-development-schedule.md # Development Schedule
│ ├── development-progress.md # Progress Tracking
│ ├── api-documentation.md # API Documentation
│ ├── database-design-document.md # Database Design
│ └── high-fidelity-prototype/ # UI Prototype
├── data/ # Data Directory
│ ├── database/ # SQLite Database
│ ├── indexes/ # Search Indexes
│ │ ├── faiss/ # Vector Index
│ │ └── whoosh/ # Full-text Index
│ ├── models/ # Model Files
│ └── logs/ # Log Files
├── .claude/ # Claude Assistant Configuration
├── LICENSE # Software License Agreement (Chinese Version)
├── LICENSE_EN # Software License Agreement (English Version)
├── README.md # Project Description (Chinese Version)
└── README_EN.md # Project Description (English Version)
```
## 🚀 Quick Start
### Method 1: Integrated Package Deployment (Recommended for Ordinary Users)
> **Applicable Users**: Non-developers, users who want to experience XiaoyaoSearch quickly.
> **Supported Platforms**: Only supports Windows.
> **Deployment Difficulty**: ⭐ Simple (one-click installation).
#### Download Integrated Package
Download the latest Windows integrated package from Baidu Netdisk:
- Link: https://pan.baidu.com/s/1lDaWjMCRXIT-Sqx9UFjerg?pwd=37ed
- Extraction Code: 37ed
Please download the latest version (e.g., `XiaoyaoSearch-Windows-v1.1.1.zip`).
#### Installation Steps
**1. Unzip Integrated Package**
Unzip the downloaded compressed package to any directory (recommended to avoid Chinese paths).
**2. Run Environment Preparation Script**
Double-click to run `scripts/setup.bat`, the script will automatically complete the following operations:
- Unzip Python embedded runtime.
- Install backend Python dependencies.
- Install frontend Node dependencies.
- Generate configuration files.
- Create data directories.
> **RTX 50 Series Graphics Card Users**: If you use RTX 50 series graphics cards, please run `scripts/setup_rtx50显卡.bat`, this script will install PyTorch version supporting CUDA 12.8 for optimal performance.
**3. Install Ollama**
Double-click to run `runtime\ollama\OllamaSetup.exe`, follow the prompts to complete the installation.
After installation, open the command line and run:
```bash
ollama serve
ollama pull qwen2.5:1.5b
```
**4. Download AI Models**
Download default models from Baidu Netdisk:
- Link: https://pan.baidu.com/s/1jRcTztvjf8aiExUh6oayVg
- Extraction Code: ycr5
Unzip the models to the corresponding directories:
- `data/models/embedding/BAAI/bge-m3/`
- `data/models/cn-clip/`
- `data/models/faster-whisper/`
**5. Start Application**
Double-click to run `scripts/startup.bat`, the script will:
- Start backend service.
- Start frontend service.
**Detailed Documentation**: [Integrated Package Deployment Guide](docs/部署文档/整合包部署指南.md)
---
### Method 2: Developer Deployment
> **Applicable Users**: Developers, users who want to participate in project contributions.
> **Supported Platforms**: Windows / macOS / Linux.
> **Deployment Difficulty**: ⭐⭐⭐ Requires development environment.
#### Environment Requirements
- **Operating System**: Windows / macOS / Linux.
- **Python**: 3.10.11+ (https://www.python.org/downloads/).
- **Node.js**: 21.x+ (https://nodejs.org/en/download).
- **Memory**: Recommended 16GB or more.
- **Graphics Card**: Recommended RTX3060 6GB or more.
#### Installation Steps
**1. Clone Project**
```bash
git clone https://github.com/dtsola/xiaoyaosearch.git
cd xiaoyaosearch
```
**2. Backend Deployment**
```shell
# Enter backend directory
cd backend
# Install dependency packages (default CPU version inference engine)
pip install -r requirements.txt
# Install faster-whisper
pip install faster-whisper
# Enable CUDA (optional, note: cuda version needs to be determined according to the environment)
pip uninstall torch torchaudio torchvision
# RTX 40 series and earlier graphics cards (CUDA 12.1)
pip install torch==2.1.0+cu121 torchaudio==2.1.0+cu121 torchvision==0.16.0+cu121 --index-url https://download.pytorch.org/whl/cu121
# RTX 50 series graphics cards (CUDA 12.8)
pip install torch==2.10.0+cu128 torchaudio==2.10.0+cu128 torchvision==0.25.0+cu128 --index-url https://download.pytorch.org/whl/cu128
```
**Install ffmpeg**:
https://ffmpeg.org/download.html
**Install ollama**:
https://ollama.com/
**Configure `.env` file**:
```env
# Data configuration
FAISS_INDEX_PATH=../data/indexes/faiss
WHOOSH_INDEX_PATH=../data/indexes/whoosh
DATABASE_PATH=../data/database/xiaoyao_search.db
# API configuration
API_HOST=127.0.0.1
API_PORT=8000
API_RELOAD=true
# Log configuration
LOG_LEVEL=info
LOG_FILE=../data/logs/app.log
```
**Prepare Models**:
Default model description:
- ollama: qwen2.5:1.5b
- Embedding model: BAAI/bge-m3
- Speech recognition model: Systran/faster-whisper-base
- Visual model: OFA-Sys/chinese-clip-vit-base-patch16
Note: It is recommended to prepare default models first, successfully start the application, and then replace models.
ollama model:
ollama pull qwen2.5:1.5b (select according to the situation)
All model download addresses: (Baidu disk)
Link: https://pan.baidu.com/s/1jRcTztvjf8aiExUh6oayVg?pwd=ycr5 Extraction Code: ycr5
Embedding model:
- Model root directory: data/models/embedding
- Unzip and place the downloaded model directly into the root directory, the following is the corresponding relationship
- data/models/embedding/BAAI/bge-m3
- data/models/embedding/BAAI/bge-small-zh
- data/models/embedding/BAAI/bge-large-zh
Speech recognition model:
- Model root directory: data/models/faster-whisper
- Unzip and place the downloaded model directly into the root directory, the following is the corresponding relationship
- data/models/faster-whisper/Systran/faster-whisper-base
- data/models/faster-whisper/Systran/faster-whisper-small
- data/models/faster-whisper/Systran/faster-whisper-medium
- data/models/faster-whisper/Systran/faster-whisper-large-v3
Visual model:
- Model root directory: data/models/cn-clip
- Unzip and place the downloaded model directly into the root directory, the following is the corresponding relationship
- data/models/cn-clip/OFA-Sys/chinese-clip-vit-base-patch16
- data/models/cn-clip/OFA-Sys/chinese-clip-vit-large-patch14
**Start Backend Service**:
```shell
# Start with built-in configuration
python main.py
# Or start with uvicorn
uvicorn main:app --host 127.0.0.1 --port 8000 --reload
```
#### 3. Frontend Deployment
```shell
# Enter frontend directory
cd frontend
# Install dependencies
npm install
# Start development server
npm run dev
```
---
## 🔄 Version Upgrade Guide
When upgrading to a new version, please refer to [Version Upgrade Guide](docs/技术文档/版本升级指南.md) to easily retain your index data and configuration.
---
## 🤝 How to Share Code
Thank you for your attention to XiaoyaoSearch! We welcome any form of contribution, whether it's code, documentation, bug fixing, or new feature suggestions.
### Contribution Methods
#### Method 1: Submit Pull Request (Recommended)
**Step 1: Fork Project**
1. Visit [xiaoyaosearch](https://github.com/dtsola/xiaoyaosearch) repository.
2. Click the "Fork" button in the upper right corner to fork the project to your GitHub account.
**Step 2: Clone to Local**
```bash
git clone https://github.com/<your-username>/xiaoyaosearch.git
cd xiaoyaosearch
```
**Step 3: Create Feature Branch**
```bash
git checkout -b feature/your-feature-name
```
# Tool List
## Or
git checkout -b fix/issue-description
**Step 4: Development**
- Follow the project's code specifications for development
- Ensure that the code has appropriate comments
- Run tests to ensure the functionality is normal
**Step 5: Submit Code**
```bash
git add .
git commit -m "feat(scope): Brief description of your changes"
```
Submission format specifications:
- `feat`: New feature
- `fix`: Bug fix
- `docs`: Document update
- `style`: Code format adjustment
- `refactor`: Code refactoring
- `perf`: Performance optimization
- `test`: Test-related
- `chore`: Build/toolchain related
**Step 6: Push to GitHub**
```bash
git push origin feature/your-feature-name
```
**Step 7: Create Pull Request**
1. Visit your Fork repository page
2. Click the "Compare & pull request" button
3. Fill in the PR description:
- Title: Briefly describe the changes
- Description: Detailed description of the changes, implementation method, and test results
4. Wait for the maintainer to review
#### Method 2: Submit Issue
If you find a Bug or have a feature suggestion:
1. Visit the [Issues](https://github.com/dtsola/xiaoyaosearch/issues) page
2. Click "New Issue"
3. Choose the appropriate Issue template
4. Describe the problem or suggestion in detail
### Code Specifications
#### Frontend Specifications
- Component naming: PascalCase (e.g., `SearchPanel.vue`)
- Variables/functions: camelCase (e.g., `searchResults`)
- Constants: UPPER_SNAKE_CASE (e.g., `MAX_FILE_SIZE`)
- Code comments: Use English
#### Backend Specifications
- File naming: snake_case (e.g., `search_service.py`)
- Class names: PascalCase (e.g., `SearchService`)
- Functions/variables: snake_case (e.g., `search_files`)
- Constants: UPPER_SNAKE_CASE (e.g., `MAX_RESULTS`)
- Code comments: Use English
### Contribution Guidelines
- ✅ Follow the project's code specifications
- ✅ Keep the code concise and avoid over-design
- ✅ Add appropriate error handling
- ✅ Ensure the code has appropriate testing
- ✅ Update relevant documentation
### Get Help
- 💬 WeChat: dtsola (please remark "Xiaoyao Search Contribution")
- 📧 Email: Contact through the official website https://www.dtsola.com
- 📺 Bilibili: https://space.bilibili.com/736015
### Contributor Benefits
- 📝 Your name will appear in the project contributor list
- 🌟 Your changes will help thousands of users
- 🤝 Join the independent developer community for exchange and learning
- 🎁 Excellent contributors can receive project周边礼品
---
**Let's build a better local search experience together!** 🚀
## Product Roadmap
[Product Roadmap](ROADMAP.md)
## Data Source Plugin List
Xiaoyao Search supports a **plugin-based architecture**, which can extend multiple data sources through plugins:
### Supported Data Source Types
| Type | Description | Status |
|------|------|------|
| 📁 Local Files | System built-in, no configuration required | ✅ Implemented |
| ☁️ Yuque | Alibaba Yuque knowledge base | ✅ Implemented |
| ☁️ Feishu | Feishu documents (metadata block parsing) | ✅ Implemented |
| ☁️ DingTalk | DingTalk documents (.xyddjson metadata file) | ✅ Implemented |
| ☁️ Notion | Notion notes | 📋 Planned |
| 🔗 GitHub | Code repository and Wiki | 📋 Planned |
| 🔗 GitLab | GitLab code repository | 📋 Planned |
### Complete List
View the complete list of data source plugins (13 types):
**📖 [Data Source Plugin List](docs/技术文档/数据源插件列表.md)** | [English Version](docs/技术文档/数据源插件列表_EN.md)
### Develop Plugins
Want to develop new data source plugins?
**📖 [Plugin Development Document](docs/技术文档/插件开发文档.md)**
---
## 🔥 MCP Server Support
Xiaoyao Search now supports **Model Context Protocol (MCP)**, which can be connected to AI applications like Claude Desktop for local file intelligent search.
### What is MCP?
MCP (Model Context Protocol) is an open protocol launched by Anthropic, allowing AI applications (like Claude Desktop) to connect to local data sources. Through MCP, Claude can directly search and access your local files, providing smarter Q&A and assistance.
### Agent Skills Support
Xiaoyao Search now supports **Agent Skills**, providing standardized MCP tool calling capabilities for Claude Code, VS Code, Cursor, and other AI assistants.
**Install Skill**:
```bash
# Project level
cp -r skills/ .claude/skills/
# Or global level
cp -r skills/ ~/.claude/skills/
```
After installation, AI assistants can automatically discover the MCP tools of Xiaoyao Search and provide correct usage guidance.
### MCP Client Configuration
The Xiaoyao Search MCP server uses the **HTTP transport protocol**, and any client that supports HTTP MCP can connect.
#### Claude Code CLI Configuration
Official command-line tool for quick configuration:
```bash
# Add HTTP MCP server
claude mcp add --transport http xiaoyao-search http://127.0.0.1:8000/mcp
# Check if MCP is added successfully (assuming MCP is already started, run the following command)
claude mcp list
```
#### Other MCP-supported Clients
Any client that supports the MCP protocol can connect to: `http://127.0.0.1:8000/mcp`
**Basic Configuration Template**:
```json
{
"name": "xiaoyao-search",
"url": "http://127.0.0.1:8000/mcp",
"type": "sse"
}
```
**Common Client Configuration Examples**:
- **Cline (VSCode plugin)**: In the VSCode settings, search for `cline.mcpServers` and add the above configuration
- **Cursor**: In the Cursor settings' MCP server configuration, add the above configuration
- **Other MCP clients**: Refer to the client documentation and use the SSE transmission method to connect
### Supported Search Tools
| Tool Name | Description | AI Model |
|---------|------|---------|
| semantic_search | Semantic search, supports natural language query understanding | BGE-M3 |
| fulltext_search | Full-text search, supports precise keyword matching and Chinese word segmentation | Whoosh |
| voice_search | Voice search, supports voice input conversion to text and then search | FasterWhisper |
| image_search | Image search, supports image upload to find similar content | CN-CLIP |
| hybrid_search | Hybrid search, combining the advantages of semantic and full-text search | BGE-M3 + Whoosh |
### Usage Examples
After configuration, you can perform the following operations in Claude Desktop:
**Semantic Search**:
```
User: Help me find the document about asynchronous programming
Claude: [Call semantic_search tool] Found 5 related documents...
```
**Full-text Search**:
```
User: Search for files containing "async def"
Claude: [Call fulltext_search tool] Found 3 code files...
```
**Image Search**:
```
User: [Upload image] Find similar charts
Claude: [Call image_search tool] Found 2 similar charts...
```
### Verify MCP Connection
Access the health check endpoint to verify the MCP service status:
```bash
curl http://127.0.0.1:8000/mcp/health
```
Return example:
```json
{
"status": "enabled",
"server": "fastmcp",
"tools_count": 5,
"tools": ["semantic_search", "fulltext_search", "voice_search", "image_search", "hybrid_search"]
}
```
### Configuration Options
Configure the MCP service in `backend/.env`:
```bash
# MCP server configuration
MCP_SSE_ENABLED=true # Whether to enable MCP SSE service
MCP_SERVER_NAME=xiaoyao-search # Server name
MCP_DEFAULT_LIMIT=20 # Default result count
MCP_DEFAULT_THRESHOLD=0.5 # Default similarity threshold
MCP_VOICE_ENABLED=true # Whether to enable voice search
```
### Technical Implementation
- **Protocol implementation**: Using the [fastmcp](https://github.com/PrefectHQ/fastmcp) framework
- **Transmission method**: HTTP SSE (Server-Sent Events)
- **Architecture pattern**: FastAPI integration, sharing AI models and search services
- **Memory optimization**: Single process, models loaded only once, saving 4-6GB of memory
### Detailed Documentation
- [MCP PRD](docs/特性开发/mcp/mcp-01-prd.md) - Product requirements document
- [MCP Technical Solution](docs/特性开发/mcp/mcp-03-技术方案.md) - Technical implementation plan
- [MCP Official Documentation](https://modelcontextprotocol.io/) - MCP protocol specification
---
## Project Contributors
Thanks to the following individuals for their contributions to this project:
<p align="center">
<a href="https://github.com/luojiyin1987">
<img src="https://avatars.githubusercontent.com/u/6524977?v=4" width="80" height="80" style="border-radius: 50%;">
</a>
<a href="https://github.com/jidingliu">
<img src="https://avatars.githubusercontent.com/u/26623206?v=4" width="80" height="80" style="border-radius: 50%;">
</a>
</p>
<p align="center">
<a href="https://github.com/luojiyin1987">@luojiyin1987</a>
<a href="https://github.com/jidingliu">@jidingliu</a>
</p>
---
## 📝 Version Update Log
### v2.0.0 (2026-04-24) - UI Visual System Upgrade Edition 🎨
**Core Upgrades**:
- ✅ **Notion warm and bright design style**: Pure white background + brand blue + warm neutral gray, visual comfort improved by 67%
- ✅ **System font stack**: Zero network dependency, optimal performance, font loading < 10ms
- ✅ **Lucide Icons icon system**: 1000+ icons, unified visual language
- ✅ **Complete design system specification**: CSS Variables + Design Tokens, ensuring design consistency
- ✅ **Exquisite multi-layer shadows**: 4-5 layer shadow stacking, enhancing visual hierarchy
- ✅ **Responsive design**: Perfectly adapted to desktop, tablet, and mobile devices
- ✅ **All page UI upgrades**: 10 pages/components comprehensively upgraded
**Detailed Documentation**: [v2.0.0 Version Update Description](docs/产品文档/版本更新文档/v2.0.0-UI&UX-界面升级.md)
---
### v1.9.0 (2026-04-12) - Search Optimization Edition
**Core Optimizations**:
- ✅ **Term expansion timing optimization**: Term expansion moved to before LLM enhancement, based on original user input matching
- ✅ **Responsibility separation**: Term library (professional synonyms) vs. LLM (general query optimization) each responsible
- ✅ **Intelligent merging**: Expanded words automatically deduplicated, avoiding repeated searches
- ✅ **Concurrent control**: Intelligent concurrent control based on CPU core count (CPU core count × 2)
- ✅ **Recall rate improvement**: Professional term scene recall rate improved by 60%
**Detailed Documentation**: [v1.9.0 Version Update Description](docs/产品文档/版本更新文档/v1.9.0-搜索优化-术语库支持.md)
### v1.8.0 (2026-04-08) - DingTalk Document Support
**New Features**:
- ✅ DingTalk document data source support (cooperating with DingTalk export tool)
- ✅ Metadata file parsing (.xyddjson format)
- ✅ Original link jump function
**Detailed Documentation**: [v1.8.0 Version Update Description](docs/产品文档/版本更新文档/v1.8.0-钉钉文档+知识库支持.md)
### v1.7.0 (2026-03-31) - Feishu Document Support
**New Features**:
- ✅ Feishu document data source support
- ✅ Metadata block parsing
- ✅ Original link jump function
### v1.6.0 (2026-03-26) - Cloud Embedded Model
**New Features**:
- ✅ Cloud embedded model calling capability
- ✅ Local/cloud mutual exclusion switching
- ✅ Batch text embedding optimization
### v1.5.0 (2026-03-20) - Agent Skills
**New Features**:
- ✅ MCP Agent Skills support
- ✅ Tool calling capability for Claude Code/VS Code/Cursor
### v1.4.0 (2026-03-15) - MCP Server Support
**New Features**:
- ✅ MCP protocol support
- ✅ Can be connected by Claude Desktop
- ✅ SSE transmission method
### v1.3.0 (2026-03-10) - OpenAI Cloud Model
**New Features**:
- ✅ OpenAI compatible cloud large model
- ✅ Dynamic form configuration
- ✅ API key encryption storage
### v1.2.0 (2026-03-05) - Plugin Architecture
**New Features**:
- ✅ Plugin architecture framework
- ✅ Yuque knowledge base data source
### v1.1.0 (2026-02-28) - i18n Internationalization
**New Features**:
- ✅ Chinese-English bilingual support
- ✅ i18n framework integration
### v1.0.0 (2026-02-20) - MVP Version
**Core Features**:
- ✅ Multimodal search (text/voice/image)
- ✅ Local file deep retrieval
- ✅ BGE-M3 + FasterWhisper + CN-CLIP
- ✅ Faiss + Whoosh hybrid search
