win-wechat-summary

KO | EN

gitlite — search

#hacktoberfest #llm #go #android #deep-learning #linux #docker #macos #kubernetes #awesome-list #vue #pytorch #ios #open-source

win-wechat-summary

(★ 21)

🤖 Windows 微信群聊 AI 总结工具 | 本地安全读取聊天记录，支持 Claude/DeepSeek/ChatGPT 多模型

win-wechat-summary 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# 微信群聊 AI 总结 / WeChat Group AI Summarizer

> 🔒 **零封号风险** — 只读本地数据库，绝不与微信服务器通信  
> 🔒 **Zero ban risk** — read-only local database, never communicates with WeChat servers
>
> 🤖 **多模型支持** — Claude / ChatGPT / DeepSeek / 通义千问 / Ollama  
> 🤖 **Multi-model support** — Claude / ChatGPT / DeepSeek / Qwen / Ollama
>
> 🖥️ **本地运行** — 所有数据留在你的电脑上，API Key 加密存储  
> 🖥️ **Runs locally** — all data stays on your machine, API keys encrypted

---

## 简介 / Introduction

**中文**：一个 Windows 桌面工具，通过 [PyWxDump](https://github.com/xaoyaoo/PyWxDump) 读取本地微信数据库，调用大语言模型对群聊消息进行智能总结。适合每天群消息太多、没时间全部看完的朋友。

**English**: A Windows desktop tool that reads your local WeChat database via [PyWxDump](https://github.com/xaoyaoo/PyWxDump) and uses large language models to generate smart summaries of group chat messages. Perfect for staying on top of busy group chats without reading every message.

---

## 功能特色 / Features

| 功能 / Feature | 说明 / Description |
|------|------|
| 📋 **群聊总结 / Group Summary** | 按时间段或最近 N 条消息生成 AI 总结 / AI summary by time range or latest N messages |
| 🔖 **阅读书签 / Bookmark** | 自动记录上次总结位置，下次一键续读 / Auto-saves last summary position for easy catch-up |
| 🔍 **关键词搜索 / Search** | 跨群全文搜索 + AI 归纳搜索结果 / Cross-group full-text search with AI-powered digest |
| 📜 **历史记录 / History** | 所有总结自动存档，支持导出为 Markdown / All summaries archived, exportable as Markdown |
| ⏰ **定时任务 / Scheduler** | 每天自动总结指定群聊，静默写入历史 / Daily auto-summary for selected groups |
| 🧩 **Map-Reduce 分块 / Chunked Summary** | 超大群聊自动分块总结再合并，减少上下文超限失败 / Splits huge chats into chunks and reduces them into one summary |
| 📊 **进度可视化 / Progress UI** | 后台总结任务实时显示读取、清洗、分块和保存进度 / Real-time progress for reading, cleaning, chunking, and saving |
| 🛡️ **SQLite 隔离锁 / SQLite Isolation** | 同步与读取互斥隔离，降低本地库替换时的并发风险 / Isolates sync and read paths to reduce concurrent replacement risk |
| ⚡ **WAL 近实时同步 / WAL Sync** | 强制 WAL Checkpoint + 加密 WAL patch，尽量补齐微信运行中的最新消息 / WAL checkpoint and encrypted WAL patch for fresher live messages |
| 🤖 **多模型支持 / Multi-model** | 自由切换 Claude / ChatGPT / DeepSeek / 通义千问 / Ollama / Switch between Claude, ChatGPT, DeepSeek, Qwen, Ollama |
| 🔐 **安全存储 / Secure Storage** | API Key 使用 Windows DPAPI 加密，只有当前用户可解密 / API keys encrypted with Windows DPAPI |
| 🗂️ **系统托盘 / System Tray** | 关闭浏览器后程序继续在托盘运行 / Keeps running in the tray after closing the browser |

---

## 截图 / Screenshots

---

## 支持的 AI 模型 / Supported AI Models

| 提供商 / Provider | 模型示例 / Example Models | 说明 / Notes |
|--------|----------|------|
| **Anthropic Claude** | claude-3-5-sonnet-20241022 | 需要 Claude API Key / Requires Claude API Key |
| **OpenAI ChatGPT** | gpt-4o, gpt-4o-mini | 需要 OpenAI API Key / Requires OpenAI API Key |
| **DeepSeek** | deepseek-chat | 需要 DeepSeek API Key，性价比高 / Requires DeepSeek API Key, great value |
| **通义千问 / Qwen** | qwen-plus, qwen-turbo | 需要阿里云 API Key / Requires Alibaba Cloud API Key |
| **Ollama** | llama3, qwen2.5 等 / etc. | 本地运行，无需 API Key / Runs locally, no API key needed |

---

## 快速开始 / Quick Start

### 前提条件 / Prerequisites

1. **Windows 10/11 64位 / 64-bit**
2. **微信 PC 版处于登录状态 / WeChat PC client, currently logged in**

### 方法一：直接运行 exe（推荐）/ Method 1: Run the exe (Recommended)

**中文**：
1. 前往 [Releases](https://github.com/yanyan1115/win-wechat-summary/releases/tag/v1.1.0) 下载最新的 `WeChat-Summary.exe`
2. 保持微信 PC 版处于**登录状态**
3. 双击 `WeChat-Summary.exe` 启动，**首次会弹出 UAC 提权对话框，点「是」即可**（读取微信内存需要管理员权限）；浏览器会自动打开 `http://127.0.0.1:5000`
4. **首次启动**会弹出设置向导，点击「自动检测微信账号」，选择账号后确认即可完成初始化（无需任何命令行操作）
5. 在「AI 设置」中配置 API Key，点击「同步」，即可开始使用

**English**:
1. Download the latest `WeChat-Summary.exe` from [Releases](https://github.com/yanyan1115/win-wechat-summary/releases/tag/v1.1.0)
2. Make sure WeChat PC client is **logged in**
3. Double-click `WeChat-Summary.exe` — a **UAC prompt will appear on first launch, click "Yes"** (admin rights are required to read WeChat's memory); your browser will automatically open `http://127.0.0.1:5000`
4. On **first launch**, a setup wizard will appear — click "Auto-detect WeChat account", select your account, and confirm (no command line needed)
5. Configure your API Key in "AI Settings", click "Sync", and you're ready to go

### 方法二：从源码运行 / Method 2: Run from Source

```powershell
# 1. 克隆仓库 / Clone the repo
git clone https://github.com/yanyan1115/win-wechat-summary.git
cd win-wechat-summary

# 2. 创建虚拟环境并安装依赖 / Create venv and install dependencies
python -m venv .venv
.\.venv\Scripts\activate
pip install -r requirements.txt

# 3. 启动应用（首次启动时浏览器会弹出设置向导）
# Start the app (setup wizard will appear on first launch)
python app.py
```

### 方法三：自行打包 / Method 3: Build from Source

```powershell
pip install pyinstaller
pyinstaller build.spec
# 生成文件在 dist/WeChat-Summary.exe
# Output: dist/WeChat-Summary.exe
```

---

## wxdump_work 目录结构 / wxdump_work Directory

**exe 版本 / exe version**：首次启动时设置向导会自动完成初始化，无需手动操作。  
The setup wizard handles initialization automatically on first launch — no manual steps needed.

**源码运行 / Source run**：需在项目根目录创建 `wxdump_work/conf_auto.json`（可由 `pywxdump showkey` 生成，或直接启动 app.py 后通过设置向导自动生成）。  
You need `wxdump_work/conf_auto.json` in the project root (generated by `pywxdump showkey`, or automatically by the setup wizard when you first run app.py).

```
wxdump_work/
└── conf_auto.json          # 微信账号配置（含数据库密钥，自动生成）
                            # WeChat account config (includes DB key, auto-generated)
└── wxid_xxxxxxxx/
    └── merge_all.db        # 同步后生成的合并数据库 / Merged database generated after sync
```

---

## 配置说明 / Configuration

**中文**：应用配置保存在 `~/.wechat-summary/config.json`（自动创建），其中 API Key 使用 Windows DPAPI 加密，**只有当前 Windows 登录用户才能读取**。

**English**: App config is saved to `~/.wechat-summary/config.json` (auto-created). API keys are encrypted with Windows DPAPI and can **only be decrypted by the current Windows user**.

---

## API 调用示例 / API Examples

所有接口都只监听本机地址，默认是 `http://127.0.0.1:5000`。

创建异步总结任务：

```powershell
$body = @{
  room_id = "12345678@chatroom"
  mode = "count"
  count = 200
} | ConvertTo-Json

Invoke-RestMethod `
  -Uri "http://127.0.0.1:5000/api/summarize" `
  -Method Post `
  -ContentType "application/json" `
  -Body $body
```

返回示例：

```json
{
  "ok": true,
  "job_id": "8f9d...",
  "status": "queued",
  "message": "总结任务已创建"
}
```

轮询任务进度：

```powershell
Invoke-RestMethod "http://127.0.0.1:5000/api/jobs/8f9d..."
```

任务状态会依次经过 `queued`、`reading`、`preprocessing`、`calling_ai`、`saving`、`done`；失败时为 `failed`，并返回 `error`。长聊天记录会先清洗低信息量消息，并在超过模型上下文预算时自动使用 Map-Reduce 分块总结。

---

## 常见问题 / FAQ

**Q: 会不会导致微信封号？/ Will this get my WeChat account banned?**

A（中文）: 不会。本工具只读取本地 SQLite 数据库文件，不与微信服务器产生任何通信。

A (English): No. This tool only reads local SQLite database files and never communicates with WeChat's servers.

---

**Q: 找不到群聊 / 群列表为空？/ Group list is empty?**

A（中文）: 确认 `wxdump_work/conf_auto.json` 存在且路径正确。点击界面左上角「同步」按钮触发数据库解密合并。

A (English): Make sure `wxdump_work/conf_auto.json` exists and is correctly configured. Click the "Sync" button in the top-left to decrypt and merge the database.

---

**Q: 同步后消息不是最新的？/ Messages are not up to date after sync?**

A（中文）: 自 v1.1.0 起，程序已引入强制 WAL Checkpoint 机制，现在可以近乎实时地同步最新消息。若极极端情况下依然有延迟，才需要稍作等待。

A (English): Since v1.1.0, the app includes a forced WAL Checkpoint mechanism, so latest messages can now be synced almost in real time. Only in extremely rare cases should you need to wait briefly and sync again.

---

**Q: 总结质量不好？/ Summary quality is poor?**

A（中文）: 可以在「AI 设置」中切换 Prompt 模板（技术交流群 / 通用群），或选择更强的模型（如 Claude Sonnet / GPT-4o）。

A (English): Try switching the prompt template in "AI Settings" (Tech group / General group), or use a more capable model such as Claude Sonnet or GPT-4o.

---

**Q: 如何在没有网络的环境下使用？/ Can I use this offline?**

A（中文）: 前端已将 Tailwind / Alpine.js / Marked.js 打包为本地文件，无需网络。但 AI API 调用仍需联网（Ollama 除外）。

A (English): The frontend bundles Tailwind / Alpine.js / Marked.js as local files — no internet needed for the UI. AI API calls still require internet access (except Ollama, which runs locally).

---

**Q: 支持 Mac / Linux 吗？/ Is Mac / Linux supported?**

A（中文）: 暂不支持。微信 PC 版仅有 Windows 版本，且数据库解密依赖 Windows DPAPI。

A (English): Not currently. WeChat PC client is Windows-only, and database decryption relies on Windows DPAPI.

---

## 技术栈 / Tech Stack

| | |
|---|---|
| **后端 / Backend** | Python 3.11 + Flask |
| **前端 / Frontend** | 单文件 HTML SPA + Tailwind CSS + Alpine.js（无 Node.js 构建 / no Node.js build step） |
| **数据库 / Database** | SQLite（通过 PyWxDump 解密读取 / decrypted via PyWxDump） |
| **AI SDK** | anthropic / openai / requests (Ollama) |
| **定时任务 / Scheduler** | APScheduler |
| **打包 / Packaging** | PyInstaller（单文件 exe，约 38MB / single-file exe, ~38MB） |

---

## 贡献 / Contributing

**中文**：欢迎提 Issue 和 PR！特别欢迎：
- 新 AI Provider 适配
- Prompt 模板优化
- Bug 修复

**English**: Issues and PRs are welcome! Especially:
- New AI provider integrations
- Prompt template improvements
- Bug fixes

---

## 免责声明 / Disclaimer

**中文**：本工具仅供个人学习和使用，请遵守微信用户协议。作者不对任何滥用行为承担责任。

**English**: This tool is for personal use only. Please comply with WeChat's Terms of Service. The author is not responsible for any misuse.

---

## 致谢 / Credits

**中文**：本项目在以下 AI 的帮助下成形并实现：

**English**: This Project was shaped and implemented with help from:

- Anthropic Claude Code
- OpenAI ChatGPT Codex
- Google Gemini

---

## License

win-wechat-summary

win-wechat-summary 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

win-wechat-summary

win-wechat-summary 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README