Content
# Xiaohongshu Content Auto-Generation and Publishing System
中文 | [English](README_EN.md)
An AI-based intelligent Xiaohongshu content generation and automatic publishing Web application that integrates multiple tool services through MCP (Model Context Protocol) to achieve a fully automated process from topic input to content publishing.
Simply put: Enter a topic and let AI handle everything for you, directly publishing it to your Xiaohongshu account.
System main interface:
## ✨ Features
- 🎨 **Modern Web Interface** - Clean and beautiful left-right layout, responsive design
- ⚙️ **Configuration Management** - All configurations are entered through the Web interface and automatically saved to the configuration file
- 🔌 **MCP Service Integration** - Integrates Jina search, Tavily search, and Xiaohongshu publishing three major MCP services
- 🤖 **Intelligent Content Generation** - Automatically retrieves information, analyzes, organizes, and writes articles based on user topics
- 📝 **Automatic Formatting** - Automatically adapts to Xiaohongshu publishing format (title, body, tags, images)
- 🚀 **One-Click Publishing** - Automatically publishes to the Xiaohongshu platform after generation
- 📊 **Real-Time Progress** - Displays execution progress and status information
- ✅ **Result Display** - Displays the final published title, content, tags, images, and other information
- 🔥 **Hot Topic Recommendation** - Automatically obtains today's hot news and intelligently extracts topics for selection
- 🎯 **Batch Generation and Publishing** - Supports batch selection of multiple topics, one-click batch generation and publishing
- 🌐 **URL Content Extraction** - Supports entering web links, automatically crawls and extracts topic content
- 🏷️ **Quick Domain Filtering** - Provides quick filter buttons for domains such as AI, financing, papers, and robots
**Advertisement**
This is my own account, all articles are written by this project
[Personal Homepage](https://www.xiaohongshu.com/user/profile/60857eed0000000001008826) Welcome to view, click to follow, and take you to get first-hand AI resources
## 📋 Prerequisites
### 1. Start Xiaohongshu MCP Service
**You must first start the [xiaohongshu-mcp](https://github.com/xpzouying/xiaohongshu-mcp) service**
```bash
# 1. Clone the Xiaohongshu MCP project
git clone https://github.com/xpzouying/xiaohongshu-mcp.git
cd xiaohongshu-mcp
# 2. Start the service according to the instructions in the project README
# Default service address: http://localhost:18060/mcp
```
⚠️Note: You must first log in to your Xiaohongshu account (log in according to the method in this mcp service)
### 2. System Environment
- Python 3.8+
- Node.js 16+ (for MCP tools)
- npm/npx (for running MCP tools)
## 🚀 Quick Start
### 1. Install Dependencies
```bash
cd xhs_web_app
pip install -r requirements.txt
```
### 2. Start Application
```bash
python app.py
```
The application starts by default at `http://localhost:8080`.
### 3. Configure System
Visit `http://localhost:8080` and configure the following information in the left panel:
#### Required Configuration
| Configuration Item | Description | Example |
|--------|------|------|
| **LLM API Key** | OpenAI compatible API key | `sk-xxx...` |
| **OpenAI Base URL** | API base address | `https://api.openai.com/v1` or `https://usw.sealos.io/v1` |
| **Default Model** | LLM model used | `claude-sonnet-4-20250514` (recommended) |
| **Xiaohongshu MCP Service Address** | MCP service URL | `http://localhost:18060/mcp` |
#### Optional Configuration (must choose one)
| Configuration Item | Description | How to Obtain |
|--------|------|----------|
| **Jina API Key** | Jina search service key | [Jina Official Website](https://jina.ai/) |
| **Tavily API Key** | Tavily search service key | [Tavily Official Website](https://tavily.com/) |
It is recommended to use Tavily as a search tool, which can be used for 1000 free search requests per month.
> 💡 **Tip**: Click the "Get Key" or "View Documentation" link next to the configuration item to directly jump to the corresponding service website.
### 4. Generate Content
The system adopts a brand-new immersive dual-view design:
#### Method 1: Creation Center (Home View)
This is the system's default view, focusing on in-depth creation of a single topic.
1. **Enter Topic**: In the input box below "What will you create today?" in the center of the screen, enter the topic you want to create.
2. **Start Generation**: Click the "✨ Generate" button on the right.
3. **Real-Time Status**:
- The "Current Task Status Card" will automatically pop up below the input box.
- You can see the steps that AI is performing (Retrieve Information -> Write Article -> Format Adaptation -> Automatic Publishing).
- The progress bar will be updated in real time, showing the current execution percentage.
**Topic Example:**
- "Analysis of the Generative AI Priority Paradox"
#### Method 2: Hotspot Discovery (Trending View)
Click the "🔥 Hotspot" icon in the left sidebar to enter this view, which is suitable for finding inspiration and mass production.
##### 1. Get Hotspots by Domain
At the top of the page, click the glass-morphic domain tag to quickly get popular topics in that domain:
- 🤖 **AI** - Artificial intelligence, cutting-edge dynamics of large models
- 💰 **Financing** - Latest financing news in the venture capital circle
- 📄 **Papers** - Latest research results in the academic world
- 🦾 **Robots** - Embodied intelligence and automation technology
##### 2. Extract from Web Links
If you have a specific article link:
1. Paste the URL in the "Or paste a web link to extract the topic..." input box.
2. Click the "Extract" button, and the system will intelligently analyze the web content and extract the core topic.
##### 3. Batch Generation and Publishing
After getting the topic card:
1. **Select Topic**: Click on the topic card you are interested in (the card will be highlighted and a ✓ mark will appear).
2. **Select All Function**: Click the "✓ Select All" button above the list to select all current topics with one click.
3. **Batch Execution**:
- The "Batch Operation Bar" will automatically float at the bottom, showing the number of selected items.
- Click the "🚀 Batch Generate Selected Topics" button.
- The system will automatically create a task queue and process the generation and publishing of each topic in sequence.
#### Task History and Management
Click the "📝 History" icon in the left sidebar to open the history record modal box:
- **View Details**: Click the "View" button of any task to preview the complete Xiaohongshu copy (title, body, tags) and images generated.
- **Status Filtering**: Use the segmented controller at the top (All/Success/Failure) to quickly filter task statuses.
## 📂 Project Structure
```
xhs_web_app/
├── app.py # FastAPI main program
├── requirements.txt # Python dependencies
├── README.md # This document
│
├── config/ # Configuration module
│ ├── __init__.py
│ ├── config_manager.py # Configuration manager
│ ├── app_config.json # Application configuration (automatically generated)
│ ├── servers_config.json # MCP service configuration (automatically generated)
│ └── .env # Environment variables (automatically generated)
│
├── core/ # Core function module
│ ├── __init__.py
│ ├── content_generator.py # Content generator (based on MCP tool execution)
│ └── xhs_llm_client.py # MCP client and LLM interaction
│
├── cache/ # Cache module
│ ├── __init__.py
│ ├── cache_manager.py # Task history cache management
│ └── task_history.json # Task history record (automatically generated)
│
├── static/ # Static resources
│ ├── css/
│ │ └── style.css # Style file
│ └── js/
│ └── app.js # Front-end interaction logic
│
└── templates/ # HTML template
└── index.html # Main page
```
## 🔧 Configuration File Description
### 1. app_config.json
Stores the main configuration information of the application:
```json
{
"llm_api_key": "sk-xxx...",
"openai_base_url": "https://api.openai.com/v1",
"default_model": "claude-sonnet-4-20250514",
"jina_api_key": "jina_xxx...",
"tavily_api_key": "tvly_xxx...",
"xhs_mcp_url": "http://localhost:18060/mcp"
}
```
### 2. servers_config.json
MCP server configuration (automatically generated according to app_config.json):
```json
{
"mcpServers": {
"jina-mcp-tools": {
"command": "npx",
"args": ["jina-mcp-tools"],
"env": {
"JINA_API_KEY": "jina_xxx..."
}
},
"tavily-remote": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://mcp.tavily.com/mcp/?tavilyApiKey=tvly_xxx..."]
},
"xhs": {
"type": "streamable_http",
"url": "http://localhost:18060/mcp"
}
}
}
```
## 🎯 Content Generation Process
The system uses a 4-step process to automatically generate and publish content:
### Step 1: Information Retrieval
- Use Jina or Tavily search tools
- Retrieve relevant information from the latest 7-30 days
- Collect 5-8 high-quality materials
- Get 3-4 related pictures
### Step 2: Write Article
- Write professional articles based on the collected information
- The title is controlled within 20 characters
- The body is 800-1200 words
- Use a youthful and lively language style
- Add appropriate emoji to increase fun
### Step 3: Format Adaptation
- Adjust to Xiaohongshu special format
- Remove # tags and replace them with natural language
- Extract 5 accurate topic tags
- Verify the validity of image links
- Generate standard JSON format
### Step 4: Automatic Publishing
- Call Xiaohongshu MCP service
- Publish content to Xiaohongshu platform
- Return publishing results
- **Automatically stop iteration after detecting successful publishing**
## 🌐 API Endpoints
| Method | Path | Description |
|------|------|------|
| GET | `/` | Main page |
| GET | `/api/config` | Get configuration (hide sensitive information) |
| POST | `/api/config` | Save configuration |
| POST | `/api/validate-model` | Verify whether the LLM model is available |
| POST | `/api/test-login` | Test Xiaohongshu MCP connection |
| POST | `/api/generate-and-publish` | Generate and publish a single topic |
| POST | `/api/fetch-trending-topics` | Get today's hot topics (support domain filtering) |
| POST | `/api/fetch-topics-from-url` | Crawl and extract topics from URL |
| POST | `/api/batch-generate-and-publish` | Batch generate and publish multiple topics |
| GET | `/api/history` | Get task history records |
| DELETE | `/api/history/{task_id}` | Delete the specified task record |
| GET | `/api/history/statistics` | Get task statistics |
## 🛠️ Technology Stack
### Backend
- **FastAPI** - Modern Web framework
- **MCP (Model Context Protocol)** - Tool service integration
- **OpenAI SDK** - LLM API call
- **aiohttp** - Asynchronous HTTP client
### Frontend
- **Native HTML/CSS/JavaScript** - Lightweight, no build required
- **Element UI Style** - Simple and beautiful UI design
- **Responsive Layout** - Support desktop and mobile
### MCP Tool Service
- **Jina MCP Tools** - Web search and content crawling
- **Tavily Remote** - Deep web search
- **XHS MCP** - Xiaohongshu content publishing
## 📝 License
This project is for learning and research purposes only. When using this project, please abide by:
- OpenAI API Terms of Use
- Xiaohongshu Platform Rules
- Usage agreements of each MCP service provider
## 🔗 Related Links
- [Xiaohongshu MCP Service](https://github.com/xpzouying/xiaohongshu-mcp) - Required dependency service
- [Sealos API Platform](https://usw.sealos.io/) - Get LLM API key
- [Jina AI](https://jina.ai/) - Get Jina API key
- [Tavily](https://tavily.com/) - Get Tavily API key
## 📧 Support and Feedback
If you have any questions or suggestions, please:
- Submit Issue
- Initiate Pull Request
- Contact the project maintainer
## ⚠️ Write at the end
This project can be used as a learning case for agents, but it is not recommended to be directly used in a production environment.
The logic of the overall project is abstracting a more general agent architecture:
- Configurable mcp
- Different tools can be selected to complete complex tasks
- Agent's context retrieval and compression
- Agent's multi-turn dialogue and memory
- .....
---
** Welcome star Happy Creating! 🎉**
