wechat-ai-radar

WeChat AI Radar Skill Documentation

Skill Overview

Project Path: ~/wechat_ai_radar/

WeChat Moments AI Radar is a complete automated content collection + analysis system. Core workflow:

微信朋友圈滚动截图 → AI视觉提取内容 → 存储数据库 → AI分类聚合分析 → 生成结构化简报

---

Core Architecture

Module Description

Module	Path	Function
config.py	~/wechat_ai_radar/config.py	Global configuration (coordinates, API, logs, etc.)
app.py	~/wechat_ai_radar/app.py	Main program entry point, supports multiple run modes
run_300.py	~/wechat_ai_radar/run_300.py	Complete workflow script for collecting 300 posts
automation/	automation/scroll_moments.py	PyAutoGUI simulated scrolling screenshots
extractor/	extractor/moments_extractor.py	AI vision extracts Moments content from screenshots
db/	db/database.py	SQLite database read/write
ai/	ai/classify.py, ai/summarize.py	AI classification and aggregation analysis
reports/	reports/daily_report.py	Daily report generator

---

Key Configuration Parameters

Moments window coordinates (calibrated)

WECHAT_MOMENTS_FALLBACK_X = 0   # 朋友圈窗口x坐标（屏幕左上角）
WECHAT_MOMENTS_FALLBACK_Y = 0   # 朋友圈窗口y坐标
WECHAT_CONTENT_WIDTH = 550      # 内容区宽度
WECHAT_CONTENT_HEIGHT = 969     # 内容区高度
滚动鼠标目标 = (275, 484)       # 内容区中心

Scrolling parameters

SCROLL_COUNT = 50              # 默认滚动次数
SCROLL_INTERVAL = 1.0          # 滚动间隔（秒）
SCROLL_AMOUNT = -300           # 每次滚动像素量（负值=向上）
SCROLL_MOUSE_CLICKS = 3        # 每次滚动执行几次scroll调用

AI API configuration

AI_API_URL = "https://api.qingyuntop.top/v1"  # 青禾聚合API
AI_API_KEY = config.AI_API_KEY  # 从config读取
AI_MODEL = "gpt-4o-mini"
AI_MAX_TOKENS = 4000
AI_TEMPERATURE = 0.7
AI_REQUEST_DELAY = 1.0         # API请求间隔（秒）
AI_BATCH_SIZE = 10             # 每批处理数量

---

Usage

Method 1: Complete collection workflow (collection + analysis + briefing)

cd ~/wechat_ai_radar && uv run python run_300.py

Workflow：

1. Automatically open/activate WeChat
2. Execute 150 scrolling screenshots (about 300 posts)
3. AI vision extracts content → store in moments.db
4. AI aggregation analysis
5. Generate briefing (Markdown format)

Target: about 300 posts / takes about 15-20 minutes

---

Method 2: Collect screenshots only (no analysis)

cd ~/wechat_ai_radar && uv run python -c "
from automation.open_wechat import WeChatOpener
from automation.scroll_moments import MomentsScroller
import config

opener = WeChatOpener()
opener.activate_wechat()

scroller = MomentsScroller(scroll_count=50, scroll_interval=1.5)
screenshots = scroller.scroll_full(capture=True)
print(f'截图完成: {len(screenshots)} 张')
"

---

Method 3: Generate briefing based on existing data

When there is already a extracted_*.json data file, directly generate the briefing:

cd ~/wechat_ai_radar && uv run python reports/generate_daily_report.py

Daily report structure(user-specified framework):

# 今日朋友圈热点
## 🔥 热门关键词
## 📈 热门分类（AI/活动/创业/SaaS/出海/促销）
## 💼 商机发现（AI合作/合作需求/创业动态/投资动态）
## 👥 高频人物
## 🧠 AI总结

---

Method 4: app.py multi-mode operation

# 仅滚动截图
uv run python app.py scroll

# OCR文字提取（需要paddleocr）
uv run python app.py ocr

# AI分析已有数据
uv run python app.py analyze

# AI总结
uv run python app.py summarize

# 生成日报
uv run python app.py report

# 启动定时任务（每天21:00自动运行）
uv run python app.py schedule --hour 21 --minute 0

---

Data file paths

Purpose	Path
Database	~/wechat_ai_radar/data/moments.db
Screenshot directory	~/wechat_ai_radar/screenshots/
Report directory	~/wechat_ai_radar/reports/
Log directory	~/wechat_ai_radar/logs/
Latest extracted data	~/wechat_ai_radar/data/extracted_YYYYMMDD_HHMMSS.json

---

Database Structure

-- moments 表
id, content, author, image_path, created_at, hash,
likes, comments, is_analyzed, category, keywords,
people, companies, business_opportunity, sentiment, summary

# 读取数据示例
from db.database import Database
from db.storage import Storage

db = Database()
storage = Storage(db)
moments = storage.get_recent(days=7, limit=100)
for m in moments:
    print(f"[{m.author}] {m.content[:50]}")

---

Packaged Convenience Scripts

1. reports/generate_daily_report.py

• Input:extracted_*.json file
• Output: Markdown format daily report + Word document
• Framework: trending keywords/categories/business opportunities/figures/AI summary

2. reports/daily_report_to_docx.py

• Convert daily report data to Word document (A4 landscape, blue theme)

3. extract_and_analyze.py

• Batch process screenshots → AI extraction → store in database

---

Prerequisites

# 安装依赖
cd ~/wechat_ai_radar && uv sync

# 核心依赖（pyproject.toml）
pyautogui, Pillow, python-dateutil, openai, httpx, APScheduler

---

Known Limitations

1. Window must be visible: the Moments window must be visible on the screen (cannot be minimized)
2. Coordinates depend on screen layout: calibrated to (x=0, y=0) top-left screen layout
3. Mac-exclusive: uses PyAutoGUI + macOS-specific configuration
4. OCR optional: when paddleocr is not installed, skip OCR and use AI vision extraction instead

---

Quick Command Summary

# 完整采集+简报
cd ~/wechat_ai_radar && uv run python run_300.py

# 快速生成日报（基于已有JSON）
cd ~/wechat_ai_radar && uv run python reports/generate_daily_report.py

# 生成Word版日报
cd ~/wechat_ai_radar && uv run python reports/daily_report_to_docx.py

# 采集50次滚动截图
cd ~/wechat_ai_radar && uv run python -c "
from automation.open_wechat import WeChatOpener
from automation.scroll_moments import MomentsScroller
opener = WeChatOpener()
opener.activate_wechat()
scroller = MomentsScroller(scroll_count=50, scroll_interval=1.5)
screenshots = scroller.scroll_full(capture=True)
print(f'完成: {len(screenshots)} 张截图')
"