wechat-to-anything 0.6.3 → 0.6.5

package/README.en.md

@@ -0,0 +1,165 @@
+<p align="center">
+  <img src="docs/banner.png" alt="wechat-to-anything" />
+</p>
+
+<h1 align="center">wechat-to-anything</h1>
+
+<p align="center">
+  <a href="https://www.npmjs.com/package/wechat-to-anything"><img src="https://img.shields.io/npm/v/wechat-to-anything?style=flat-square&color=cb3837" alt="npm" /></a>
+  <a href="https://github.com/kellyvv/wechat-to-anything"><img src="https://img.shields.io/github/stars/kellyvv/wechat-to-anything?style=flat-square&color=yellow" alt="stars" /></a>
+  <a href="LICENSE"><img src="https://img.shields.io/github/license/kellyvv/wechat-to-anything?style=flat-square" alt="license" /></a>
+  <a href="https://github.com/kellyvv/wechat-to-anything"><img src="https://img.shields.io/badge/node-%3E%3D22-brightgreen?style=flat-square" alt="node" /></a>
+</p>
+
+<p align="center">
+  <a href="#quick-start">Quick Start</a> · <a href="#full-multimodal-matrix">Multimodal</a> · <a href="#media-protocol">Media Protocol</a> · <a href="#multi-agent-mode">Multi-Agent</a> · <a href="#bring-your-own-agent">Custom Agent</a>
+</p>
+
+<p align="center">
+  <a href="README.md">中文</a> | English
+</p>
+
+> ⭐ If this project helps you, please give it a Star!
+
+**The first open-source project** to support full multimodal bidirectional communication between WeChat and AI Agents — text, images, voice, video, and files, both sending and receiving.
+
+<p align="center">
+  <img src="docs/wechat-image-send.png" width="250" alt="Agent sends files, images, voice" />
+  <img src="docs/wechat-image-receive.png" width="250" alt="Agent sends images, video, voice" />
+  <a href="https://github.com/kellyvv/wechat-to-anything/raw/main/docs/wechat-voice-demo.mp4">
+    <img src="docs/wechat-voice-demo.gif" width="250" alt="Voice demo (click for audio)" />
+  </a>
+</p>
+
+## Features
+
+- 🔌 **Zero-config setup** — One `npx` command, no cloning, no configuration
+- 🧠 **Agent-agnostic** — Works with any OpenAI-compatible API (Codex / Gemini / Claude / custom)
+- 📡 **Full multimodal** — Text, images, voice, video, files — bidirectional
+- 🤖 **Multi-Agent** — Connect multiple Agents simultaneously, route with `@` prefix
+- ⌨️ **Typing indicator** — Shows "typing..." while Agent is thinking
+
+### Full Multimodal Matrix
+
+| Modality | WeChat → Agent | Agent → WeChat |
+|------|:---:|:---:|
+| 📝 Text | ✅ | ✅ |
+| 📷 Image | ✅ Auto-detect | ✅ HD original |
+| 🎤 Voice | ✅ Speech-to-text | ✅ Voice bubble |
+| 🎬 Video | ✅ Auto-receive | ✅ With thumbnail |
+| 📄 File | ✅ Content extraction | ✅ Downloadable |
+
+## Quick Start
+
+```bash
+# Pick your favorite Agent:
+npx wechat-to-anything --codex     # OpenAI Codex
+npx wechat-to-anything --gemini    # Google Gemini
+npx wechat-to-anything --claude    # Claude Code
+npx wechat-to-anything --openclaw  # OpenClaw
+
+# Or pass a URL directly:
+npx wechat-to-anything http://your-agent:8000/v1
+```
+
+> First time: A QR code pops up in terminal → Scan with WeChat → Done. Login is cached automatically.
+
+### Dependencies
+
+```bash
+# 1. Node.js >= 22
+curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.3/install.sh | bash
+nvm install 22
+
+# 2. Python 3 + pip
+brew install python3       # macOS
+apt install python3 python3-pip  # Linux
+
+# 3. ffmpeg
+brew install ffmpeg        # macOS
+apt install ffmpeg         # Linux
+
+# 4. pilk
+pip install pilk
+```
+
+## How It Works
+
+```
+WeChat User ←→ Tencent ilinkai API ←→ wechat-to-anything ←→ Your Agent (HTTP)
+```
+
+Directly calls Tencent's ilinkai API to send/receive WeChat messages. No middleware, no reverse engineering, no web client. Your Agent just needs an OpenAI-compatible HTTP endpoint.
+
+## Bring Your Own Agent
+
+Any language — just expose `POST /v1/chat/completions`:
+
+```python
+@app.post("/v1/chat/completions")
+def chat(request):
+    message = request.json["messages"][-1]["content"]
+    reply = your_agent(message)
+    return {"choices": [{"message": {"role": "assistant", "content": reply}}]}
+```
+
+Then: `npx wechat-to-anything http://your-agent:8000/v1`
+
+## Media Protocol
+
+Include specific formats in Agent responses to automatically send media:
+
+| Type | Agent Response Format | Notes |
+|------|----------------------|-------|
+| Image | `![desc](URL or path)` | URL, local path, or data URI |
+| Voice | `[audio:path or URL]` | MP3/WAV/OGG, requires `ffmpeg` + `pilk` |
+| Video | `[video:path or URL]` | Requires `ffmpeg` |
+| File | `[file:path or URL]` | Any file type |
+
+**Image receiving** (WeChat → Agent) follows the [OpenAI Vision API](https://platform.openai.com/docs/guides/vision):
+
+```json
+{
+  "messages": [{
+    "role": "user",
+    "content": [
+      { "type": "text", "text": "What is this?" },
+      { "type": "image_url", "image_url": { "url": "data:image/jpeg;base64,..." } }
+    ]
+  }]
+}
+```
+
+> Examples: [image-test.mjs](examples/image-test.mjs) · [voice-test.mjs](examples/voice-test.mjs) · [video-test-local.mjs](examples/video-test-local.mjs) · [file-test.mjs](examples/file-test.mjs)
+
+## Multi-Agent Mode
+
+Connect multiple Agents simultaneously, route with `@` prefix. Supports OpenAI format and [ACP](https://agentcommunicationprotocol.dev/):
+
+```bash
+npx wechat-to-anything \
+  --agent codex=http://localhost:3001/v1 \
+  --agent gemini=http://localhost:3002/v1 \
+  --agent bee=acp://localhost:8000/chat \
+  --default codex
+```
+
+| WeChat Message | Effect |
+|---|---|
+| `Hello` | Sent to default Agent |
+| `@codex write a sort` | Routes to Codex |
+| `@gemini review code` | Routes to Gemini |
+| `@list` | List all Agents |
+| `@switch gemini` | Switch default |
+
+## Credentials
+
+Login credentials are saved in `~/.wechat-to-anything/credentials.json`. Delete to re-login.
+
+## Star History
+
+If this project helped you, please give it a ⭐ Star — it's the best support!
+
+## License
+
+[MIT](LICENSE)

package/README.md

@@ -1,79 +1,99 @@
-# wechat-to-anything
+<p align="center">
+  <img src="docs/banner.png" alt="wechat-to-anything" />
+</p>
 
-> 把微信变成任何 AI Agent 的前端。零依赖，一条命令。
->
-> ⭐ 如果这个项目对你有帮助，请给个 Star！本项目仅用于技术学习和交流，开源不易。
+<h1 align="center">wechat-to-anything</h1>
 
-微信双向支持 Agent 多种模态消息发送和接收，支持文本、图片、语音、视频、文件。
+<p align="center">
+  <a href="https://www.npmjs.com/package/wechat-to-anything"><img src="https://img.shields.io/npm/v/wechat-to-anything?style=flat-square&color=cb3837" alt="npm" /></a>
+  <a href="https://github.com/kellyvv/wechat-to-anything"><img src="https://img.shields.io/github/stars/kellyvv/wechat-to-anything?style=flat-square&color=yellow" alt="stars" /></a>
+  <a href="LICENSE"><img src="https://img.shields.io/github/license/kellyvv/wechat-to-anything?style=flat-square" alt="license" /></a>
+  <a href="https://github.com/kellyvv/wechat-to-anything"><img src="https://img.shields.io/badge/node-%3E%3D22-brightgreen?style=flat-square" alt="node" /></a>
+</p>
 
 <p align="center">
-  <img src="docs/wechat-image-send.png" width="250" alt="发送图片给 Agent 识别" />
-  <img src="docs/wechat-image-receive.png" width="250" alt="Agent 发送图片到微信" />
-  <a href="https://github.com/kellyvv/wechat-to-anything/raw/main/docs/wechat-voice-demo.mp4">
-    <img src="docs/wechat-voice-demo.gif" width="250" alt="语音发送演示（点击播放有声版）" />
-  </a>
+  <a href="#快速开始">快速开始</a> · <a href="#全模态支持矩阵">全模态</a> · <a href="#多媒体协议">多媒体协议</a> · <a href="#多-agent-模式">多 Agent</a> · <a href="#接入自己的-agent">自定义 Agent</a>
 </p>
 
-## 原理
+<p align="center">
+  中文 | <a href="README.en.md">English</a>
+</p>
 
-```
-微信 ←→ ilinkai API (腾讯) ←→ wechat-to-anything ←→ 你的 Agent (HTTP)
-```
+> ⭐ 如果这个项目对你有帮助，请给个 Star！
+
+**全网首个**支持微信与任何 AI Agent 全模态双向通信的开源项目 —— 文本、图片、语音、视频、文件，发送和接收全覆盖。
 
-直接调用腾讯 ilinkai 接口收发微信消息，无中间层。你的 Agent 只需暴露一个 OpenAI 兼容的 HTTP 接口（`POST /v1/chat/completions`），任何语言都行。
+<p align="center">
+  <img src="docs/wechat-image-send.png" width="250" alt="Agent 发送文件、图片、语音" />
+  <img src="docs/wechat-image-receive.png" width="250" alt="Agent 发送图片、视频、语音" />
+  <a href="https://github.com/kellyvv/wechat-to-anything/raw/main/docs/wechat-voice-demo.mp4">
+    <img src="docs/wechat-voice-demo.gif" width="250" alt="语音演示（点击播放有声版）" />
+  </a>
+</p>
 
-### 多种模态支持
+## 特性
 
-| 方向 | 图片 | 语音 | 视频 | 文件 |
-|---|---|---|---|---|
-| **微信 → Agent** | ✅ 自动识别 | ✅ 语音转文字 | ✅ 自动接收 | ✅ 提取文本 |
-| **Agent → 微信** | ✅ 自动发图 | ✅ 语音消息 | ✅ 视频消息 | 文本回复 |
+- 🔌 **零依赖接入** — `npx` 一条命令，无需 clone、无需配置
+- 🧠 **Agent 无关** — 支持任何 OpenAI 兼容 API（Codex / Gemini / Claude / 自建）
+- 📡 **全模态** — 文本、图片、语音、视频、文件，双向全覆盖
+- 🤖 **多 Agent** — 同时接入多个 Agent，`@` 路由切换
+- ⌨️ **打字指示器** — Agent 思考时显示"对方正在输入"
 
-## 前置条件
+### 全模态支持矩阵
 
-- Node.js >= 22（`nvm install 22`）
+| 模态 | 微信 → Agent | Agent → 微信 |
+|------|:---:|:---:|
+| 📝 文本 | ✅ | ✅ |
+| 📷 图片 | ✅ 自动识别 | ✅ HD 原图 |
+| 🎤 语音 | ✅ 语音转文字 | ✅ 语音气泡 |
+| 🎬 视频 | ✅ 自动接收 | ✅ 带缩略图 |
+| 📄 文件 | ✅ 提取内容 | ✅ 可下载 |
 
 ## 快速开始
 
 ```bash
-# 一条命令，选你喜欢的 Agent：
+# 选你喜欢的 Agent：
 npx wechat-to-anything --codex     # OpenAI Codex
 npx wechat-to-anything --gemini    # Google Gemini
 npx wechat-to-anything --claude    # Claude Code
 npx wechat-to-anything --openclaw  # OpenClaw
 
-# 多 Agent 同时用：
-npx wechat-to-anything --codex --gemini
+# 或直接传 URL：
+npx wechat-to-anything http://your-agent:8000/v1
 ```
 
-> 需要先安装对应 CLI：`npm i -g @openai/codex` / `@google/gemini-cli` / `@anthropic-ai/claude-code` / `openclaw`
->
-> 也支持直接传 URL：`npx wechat-to-anything http://your-agent:8000/v1`
+> 首次使用：终端弹出二维码 → 微信扫码 → 完成。之后自动复用登录。
 
-### 接入 OpenClaw
+### 环境依赖
 
 ```bash
-# 1. 安装并配置 OpenClaw
-npm i -g openclaw
-openclaw configure         # 设置模型（如 Gemini / OpenAI）
+# 1. Node.js >= 22
+curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.3/install.sh | bash
+nvm install 22
+
+# 2. Python 3 + pip
+brew install python3       # macOS
+apt install python3 python3-pip  # Linux
 
-# 2. 启动 Gateway
-openclaw gateway
+# 3. ffmpeg
+brew install ffmpeg        # macOS
+apt install ffmpeg         # Linux
 
-# 3. 启动桥（另一个终端）
-npx wechat-to-anything --openclaw
+# 4. pilk
+pip install pilk
 ```
 
-> OpenClaw 的 Gateway 需要先配好模型 provider（运行 `openclaw configure`）。
-> 如果 OpenClaw 已有 `openclaw-weixin` 插件，需先禁用以避免消息冲突。
+## 原理
 
-### 首次使用
+```
+微信用户 ←→ 腾讯 ilinkai API ←→ wechat-to-anything ←→ 你的 Agent (HTTP)
+```
 
-终端会弹出二维码 → 微信扫码 → 完成。之后自动复用登录凭证。
+直接调用腾讯 ilinkai 接口收发微信消息，无中间层、无逆向、无网页版。Agent 只需暴露一个 OpenAI 兼容的 HTTP 接口。
 
-## 接入你自己的 Agent
+## 接入自己的 Agent
 
-暴露 `POST /v1/chat/completions` 即可，任何语言：
+任何语言，暴露 `POST /v1/chat/completions` 即可：
 
 ```python
 @app.post("/v1/chat/completions")
@@ -83,38 +103,20 @@ def chat(request):
     return {"choices": [{"message": {"role": "assistant", "content": reply}}]}
 ```
 
-然后 `npx wechat-to-anything http://your-agent:8000/v1`。
-
-## 多 Agent 模式
-
-同时接入多个 Agent，通过 `@` 前缀路由消息。支持 OpenAI 兼容格式和 [ACP (Agent Communication Protocol)](https://agentcommunicationprotocol.dev/) 两种协议：
-
-```bash
-npx wechat-to-anything \
-  --agent codex=http://localhost:3001/v1 \
-  --agent gemini=http://localhost:3002/v1 \
-  --agent bee=acp://localhost:8000/chat \
-  --default codex
-```
-
-> `http://` → OpenAI 格式，`acp://` → ACP 协议，自动识别。
+然后：`npx wechat-to-anything http://your-agent:8000/v1`
 
-微信里使用：
-
-| 消息 | 效果 |
-|---|---|
-| `你好` | 发给默认 Agent |
-| `@codex 写个排序` | 路由到 Codex |
-| `@gemini 审查代码` | 路由到 Gemini |
-| `@bee 分析数据` | 路由到 ACP Agent |
-| `@list` | 查看已注册的 Agent |
-| `@切换 gemini` | 切换默认 Agent |
+## 多媒体协议
 
-多 Agent 模式下回复自动带 `[agentName]` 前缀标识来源。每个用户独立维护默认 Agent。
+Agent 回复中包含特定格式即可自动发送多媒体：
 
-## 多媒体格式
+| 类型 | Agent 回复格式 | 说明 |
+|------|--------------|------|
+| 图片 | `![描述](URL或路径)` | 支持 URL、本地路径、data URI |
+| 语音 | `[audio:路径或URL]` | MP3/WAV/OGG，需 `ffmpeg` + `pilk` |
+| 视频 | `[video:路径或URL]` | 需 `ffmpeg` |
+| 文件 | `[file:路径或URL]` | 任意文件类型 |
 
-**图片（微信 → Agent）**：遵循 [OpenAI Vision API](https://platform.openai.com/docs/guides/vision)，`content` 为数组：
+**图片接收**（微信 → Agent）遵循 [OpenAI Vision API](https://platform.openai.com/docs/guides/vision)：
 
 ```json
 {
@@ -128,35 +130,32 @@ npx wechat-to-anything \
 }
 ```
 
-**图片（Agent → 微信）**：回复中包含 `![desc](https://...)` 自动发图（HD 原图质量，自动生成缩略图）。支持 URL、本地路径、data URI。
-
-**语音（Agent → 微信）**：回复中包含 `[audio:path 或 URL]` 自动发语音气泡。支持 MP3、WAV、OGG 等。需要 `ffmpeg` 和 `pip install pilk`。
+> 示例：[image-test.mjs](examples/image-test.mjs) · [voice-test.mjs](examples/voice-test.mjs) · [video-test-local.mjs](examples/video-test-local.mjs) · [file-test.mjs](examples/file-test.mjs)
 
-```python
-@app.post("/v1/chat/completions")
-def chat(request):
-    audio_path = your_tts(message)  # → /tmp/reply.mp3
-    reply = f"[audio:{audio_path}]\n这是文字版内容"
-    return {"choices": [{"message": {"role": "assistant", "content": reply}}]}
-```
+## 多 Agent 模式
 
-**视频（Agent → 微信）**：回复中包含 `[video:path 或 URL]` 自动发视频消息（含缩略图）。需要 `ffmpeg`。
+同时接入多个 Agent，`@` 前缀路由。支持 OpenAI 格式和 [ACP 协议](https://agentcommunicationprotocol.dev/)：
 
-```python
-@app.post("/v1/chat/completions")
-def chat(request):
-    reply = "[video:/tmp/demo.mp4]\n这是视频描述"
-    return {"choices": [{"message": {"role": "assistant", "content": reply}}]}
+```bash
+npx wechat-to-anything \
+  --agent codex=http://localhost:3001/v1 \
+  --agent gemini=http://localhost:3002/v1 \
+  --agent bee=acp://localhost:8000/chat \
+  --default codex
 ```
 
-> 示例：[examples/image-test.mjs](examples/image-test.mjs) · [examples/voice-test.mjs](examples/voice-test.mjs) · [examples/video-test-local.mjs](examples/video-test-local.mjs)
+| 微信消息 | 效果 |
+|---|---|
+| `你好` | 发给默认 Agent |
+| `@codex 写个排序` | 路由到 Codex |
+| `@gemini 审查代码` | 路由到 Gemini |
+| `@list` | 查看所有 Agent |
+| `@切换 gemini` | 切换默认 |
 
 ## 凭证
 
 登录凭证保存在 `~/.wechat-to-anything/credentials.json`，删除即可重新登录。
 
-
-
 ## Star History
 
 如果这个项目帮到了你，请给个 ⭐ Star，这是对我们最大的支持！

package/cli/agent-adapter.mjs

@@ -17,10 +17,10 @@ import { randomBytes } from "node:crypto";
 /**
  * 统一调用接口 — 根据 URL 自动选择适配器
  */
-export async function callAgentAuto(url, messages) {
-  if (url.startsWith("acp://")) return callACP(url, messages);
+export async function callAgentAuto(url, messages, userId) {
+  if (url.startsWith("acp://")) return callACP(url, messages, userId);
   if (url.startsWith("cli://")) return callCLI(url, messages);
-  return callOpenAI(url, messages);
+  return callOpenAI(url, messages, userId);
 }
 
 /**
@@ -50,11 +50,11 @@ export async function checkAgent(url) {
 
 // ========== OpenAI 适配器 ==========
 
-async function callOpenAI(agentUrl, messages) {
+async function callOpenAI(agentUrl, messages, userId) {
   const res = await fetch(`${agentUrl}/chat/completions`, {
     method: "POST",
     headers: { "Content-Type": "application/json" },
-    body: JSON.stringify({ messages }),
+    body: JSON.stringify({ messages, user: userId || undefined }),
     signal: AbortSignal.timeout(300_000),
   });
   if (!res.ok) {
@@ -74,7 +74,7 @@ function parseACPUrl(acpUrl) {
   return { httpUrl: `http://${withoutScheme.slice(0, slashIdx)}`, agentName: withoutScheme.slice(slashIdx + 1) };
 }
 
-async function callACP(acpUrl, messages) {
+async function callACP(acpUrl, messages, userId) {
   const { httpUrl, agentName } = parseACPUrl(acpUrl);
   const input = messages.map((msg) => {
     if (typeof msg.content === "string") {

package/cli/bridge.mjs

@@ -325,7 +325,7 @@ export async function start(agents, defaultAgent) {
           // 调用 Agent
           try {
             if (typing) await typing.onReplyStart();
-            const reply = await callAgentAuto(agentUrl, agentMessages);
+            const reply = await callAgentAuto(agentUrl, agentMessages, from);
             if (typing) typing.onIdle();
             const agentTag = multiMode ? `[${targetAgent}] ` : "";
 

package/cli/cdn.mjs

@@ -355,6 +355,7 @@ export async function uploadToCdn(filePath, toUserId, token, mediaType = 1) {
     aeskey: aeskey.toString("hex"),
     fileSize: rawsize,
     fileSizeCiphertext: filesize,
+    rawMd5: rawfilemd5,
     filekey,
   };
 }

package/cli/weixin.mjs

@@ -286,8 +286,11 @@ export async function sendVoiceMessage(token, to, contextToken, audioBase64, dur
 
 /**
  * 发送文件消息（通过 CDN 引用）
+ * 格式参考微信原始 file_item: 44-char aes_key, 无 encrypt_type, 有 md5
  */
 export async function sendFileMessage(token, to, contextToken, uploaded, fileName) {
+  // aes_key: hex string → UTF-8 bytes → base64 = 44 chars（微信格式）
+  const aesKeyB64 = Buffer.from(uploaded.aeskey).toString("base64");
   await apiPost(
     "ilink/bot/sendmessage",
     {
@@ -302,10 +305,10 @@ export async function sendFileMessage(token, to, contextToken, uploaded, fileNam
           file_item: {
             media: {
               encrypt_query_param: uploaded.downloadParam,
-              aes_key: Buffer.from(uploaded.aeskey, "hex").toString("base64"),
-              encrypt_type: 1,
+              aes_key: aesKeyB64,
             },
             file_name: fileName,
+            md5: uploaded.rawMd5 || "",
             len: String(uploaded.fileSize),
           },
         }],
@@ -459,6 +462,7 @@ export function extractMedia(msg) {
     }
     // 文件
     if (item.type === 4 && item.file_item?.media?.encrypt_query_param) {
+
       return {
         type: "file",
         encryptQueryParam: item.file_item.media.encrypt_query_param,

package/examples/file-test.mjs

@@ -0,0 +1,39 @@
+/**
+ * 文件发送测试脚本
+ *
+ * 用法: node examples/file-test.mjs [文件路径]
+ *
+ * 通过 CDN 上传文件，发送到自己的微信（用于测试）。
+ * 需要先扫码登录获取 credentials。
+ */
+import { readFileSync } from "fs";
+import { homedir, tmpdir } from "os";
+import { join, basename } from "path";
+
+const creds = JSON.parse(readFileSync(homedir() + "/.wechat-to-anything/credentials.json", "utf-8"));
+
+// 测试文件：命令行指定或自动生成
+let filePath = process.argv[2];
+let fileName;
+if (filePath) {
+  fileName = basename(filePath);
+} else {
+  const { writeFileSync } = await import("fs");
+  filePath = join(tmpdir(), "wxta-test-file.txt");
+  writeFileSync(filePath, `Hello from wechat-to-anything!\n测试文件 ${new Date().toISOString()}\n`);
+  fileName = "wxta-test-file.txt";
+  console.log("未指定文件，已生成测试文件:", filePath);
+}
+
+const { uploadToCdn } = await import("../cli/cdn.mjs");
+const { getUpdates, sendFileMessage } = await import("../cli/weixin.mjs");
+
+const msgs = await getUpdates(creds.token);
+const ct = msgs?.context_token || "";
+
+console.log("上传文件:", fileName);
+const uploaded = await uploadToCdn(filePath, creds.userId, creds.token, 3);
+console.log("✅ CDN 上传完成 | 大小:", uploaded.fileSize, "bytes | md5:", uploaded.rawMd5);
+
+await sendFileMessage(creds.token, creds.userId, ct, uploaded, fileName);
+console.log("✅ 发送成功！请检查微信");

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wechat-to-anything",
-  "version": "0.6.3",
+  "version": "0.6.5",
   "description": "一条命令，把微信变成任何 AI Agent 的入口",
   "type": "module",
   "bin": {