wechatdownload

A tool designed for batch downloading content from WeChat Official Accounts. It supports Windows and MacOS and does not require manual certificate installation.

Key Features:

• Download historical messages from Official Accounts.
• Supported formats: html, mhtml, md, pdf, docx, and csv tables.
• Supports downloading comments, collections (columns), images, videos, and audio files.
• Supports MCP (Model Context Protocol) and Skill calls.
• Can export article data including view counts, likes, shares, and comment counts.

Follow these steps to download content:

1. Get the Official Account ID:

   • Open any article from the target Official Account.
   • Copy the article link.
   • Paste the link into the software and click 获取公众号id (Get Official Account ID).

2. Obtain the Secret Key (Automatic):

   • Send the copied link to your 'File Transfer Assistant' (文件传输助手) in WeChat.
   • Open the link within the WeChat built-in browser.
   • Wait for a moment; the software will automatically capture the required secret key.

3. Start Downloading:

   • Once the ID and key are acquired, you can proceed to download the articles using the software interface.

Important Note: You MUST use the WeChat built-in browser to open the link for the key capture to work.

Before performing advanced tasks like batch downloading or exporting data, verify that the local MCP server is running at http://127.0.0.1:4545/mcp.

If the service is unavailable, you must:

1. Open the WeChat article download tool.
2. Check the '启动MCP' checkbox to start the MCP service.
3. Wait for the confirmation message indicating the service is running on port 4545.

import requests

LOCAL_MCP_ENDPOINT = "http://127.0.0.1:4545/mcp"

def check_local_mcp():
    response = requests.post(
        LOCAL_MCP_ENDPOINT,
        json={"jsonrpc": "2.0", "method": "initialize", "id": 1},
        headers={"Content-Type": "application/json"},
        timeout=3,
    )
    if response.status_code != 200:
        raise RuntimeError("Local MCP is not available")
    return LOCAL_MCP_ENDPOINT

Once credentials have been successfully obtained, use the batch_download_articles tool via the local MCP server to download all articles from a specific public account. The tool will download articles based on the date range and filters configured in the tool's interface and organize them by public account name.

def batch_download_articles():
    payload = {
        "jsonrpc": "2.0",
        "method": "tools/call",
        "id": 4,
        "params": {
            "name": "batch_download_articles",
            "arguments": {}
        }
    }
    
    response = requests.post('http://127.0.0.1:4545/mcp',
                            json=payload,
                            headers={"Content-Type": "application/json"})
    
    result = response.json()
    return result

result = batch_download_articles()
print(result)

To export article metadata (such as titles, URLs, publish dates, read counts, and like counts) to a CSV file, use the export_article_data tool via the local MCP server. The resulting CSV file will be saved in the tool's download directory.

def export_article_data():
    payload = {
        "jsonrpc": "2.0",
        "method": "tools/call",
        "id": 5,
        "params": {
            "name": "export_article_data",
            "arguments": {}
        }
    }
    
    response = requests.post('http://127.0.0.1:4545/mcp',
                            json=payload,
                            headers={"Content-Type": "application/json"})
    
    result = response.json()
    return result

result = export_article_data()
print(result)

To download a collection of articles, use the remote fallback MCP (https://changfengbox.top/api/mcp) with the tool name wechat_collection. This requires a collection URL passed in the arguments.

def download_collection(collection_url):
    payload = {
        "jsonrpc": "2.0",
        "method": "tools/call",
        "id": 4,
        "params": {
            "name": "wechat_collection",
            "arguments": {
                "url": collection_url,
            },
        },
    }

    response = requests.post(
        "https://changfengbox.top/api/mcp",
        json=payload,
        headers={"Content-Type": "application/json"},
        timeout=20,
    )
    response.raise_for_status()
    return response.json()

There are two primary open-source approaches for web-based OCR (Optical Character Recognition) using large vision models:

1. Ollama OCR for web: An open-source tool that includes a built-in Web frontend.
2. Gemini-based OCR: An alternative implementation leveraging Google's Gemini models for recognition.