@sunnoy/wecom 1.0.0 → 1.1.0

package/README.md

@@ -9,6 +9,7 @@
 - 🌊 **Streaming Output**: Built on WeCom's latest AI bot streaming mechanism for smooth typewriter-style responses.
 - 🤖 **Dynamic Agent Management**: Automatically creates isolated agents per direct message user or group chat, with independent workspaces and conversation contexts.
 - 👥 **Deep Group Chat Integration**: Supports group message parsing with @mention triggering.
+- 🎤 **Voice Message Support**: Automatically processes voice messages transcribed by WeCom into text for AI interaction (direct messages only).
 - 🖼️ **Image Support**: Automatic base64 encoding and sending of local images (screenshots, generated images) without requiring additional configuration.
 - 🛠️ **Command Enhancement**: Built-in commands (e.g., `/new` for new sessions, `/status` for status) with allowlist configuration.
 - 🔒 **Security & Authentication**: Full support for WeCom message encryption/decryption, URL verification, and sender validation.
@@ -216,6 +217,10 @@ AI: [Takes screenshot] → Image displays properly in WeCom ✅
 
 If an image fails to process (size limit, invalid format), the text response will still be delivered and an error will be logged.
 
+### Q: Does the bot support voice messages?
+
+**A:** Yes! Voice messages in direct chats are automatically transcribed by WeCom and processed as text. No additional configuration needed.
+
 ### Q: How to configure auth token for public-facing OpenClaw with WeCom callbacks?
 
 **A:** WeCom bot **does not need** OpenClaw's Gateway Auth Token.

package/README_ZH.md

@@ -9,6 +9,7 @@
 - 🌊 **流式输出 (Streaming)**: 基于企业微信最新的 AI 机器人流式分片机制，实现流畅的打字机式回复体验。
 - 🤖 **动态 Agent 管理**: 默认按"每个私聊用户 / 每个群聊"自动创建独立 Agent。每个 Agent 拥有独立的工作区与对话上下文，实现更强的数据隔离。
 - 👥 **群聊深度集成**: 支持群聊消息解析，可通过 @提及（At-mention）精准触发机器人响应。
+- 🎤 **语音消息支持**: 自动处理企业微信转录后的语音消息，转换为文本进行 AI 交互（仅限私聊）。
 - 🖼️ **图片支持**: 自动将本地图片（截图、生成的图像）进行 base64 编码并发送，无需额外配置。
 - 🛠️ **指令增强**: 内置常用指令支持（如 `/new` 开启新会话、`/status` 查看状态等），并提供指令白名单配置功能。
 - 🔒 **安全与认证**: 完整支持企业微信消息加解密、URL 验证及发送者身份校验。
@@ -216,6 +217,10 @@ AI：[执行截图] → 图片在企业微信中正常显示 ✅
 
 如果图片处理失败（超出大小限制、格式不支持等），文本回复仍会正常发送，错误信息会记录在日志中。
 
+### Q: 机器人支持语音消息吗？
+
+**A:** 支持！私聊中的语音消息会被企业微信自动转录为文字并作为文本处理，无需额外配置。
+
 ### Q: OpenClaw 开放公网需要 auth token，企业微信回调如何配置？
 
 **A:** 企业微信机器人**不需要**配置 OpenClaw 的 Gateway Auth Token。

package/index.js

@@ -805,9 +805,45 @@ async function deliverWecomReply({ payload, account, responseUrl, senderId, stre
     senderId,
   });
 
+  // 处理绝对路径的 MEDIA: 行（OpenClaw 会拒绝它们，所以我们需要手动处理）
+  const mediaRegex = /^MEDIA:\s*(.+)$/gm;
+  const mediaMatches = [];
+  let match;
+  while ((match = mediaRegex.exec(text)) !== null) {
+    const mediaPath = match[1].trim();
+    // 检查是否是绝对路径（以 / 开头）
+    if (mediaPath.startsWith("/")) {
+      mediaMatches.push({
+        fullMatch: match[0],
+        path: mediaPath
+      });
+      logger.debug("Detected absolute path MEDIA line", {
+        streamId,
+        mediaPath,
+        line: match[0]
+      });
+    }
+  }
+
+  // 如果检测到绝对路径的 MEDIA 行，将图片加入队列并从文本中移除
+  let processedText = text;
+  if (mediaMatches.length > 0 && streamId) {
+    for (const media of mediaMatches) {
+      const queued = streamManager.queueImage(streamId, media.path);
+      if (queued) {
+        // 从文本中移除这行
+        processedText = processedText.replace(media.fullMatch, "").trim();
+        logger.info("Queued absolute path image for stream", {
+          streamId,
+          imagePath: media.path
+        });
+      }
+    }
+  }
+
   // 所有消息都通过流式发送
-  if (!text.trim()) {
-    logger.debug("WeCom: empty block, skipping stream update");
+  if (!processedText.trim()) {
+    logger.debug("WeCom: empty block after processing, skipping stream update");
     return;
   }
 
@@ -834,10 +870,10 @@ async function deliverWecomReply({ payload, account, responseUrl, senderId, stre
     // 尝试从 activeStreams 获取
     const activeStreamId = activeStreams.get(senderId);
     if (activeStreamId && streamManager.hasStream(activeStreamId)) {
-      appendToStream(activeStreamId, text);
+      appendToStream(activeStreamId, processedText);
       logger.debug("WeCom stream appended (via activeStreams)", {
         streamId: activeStreamId,
-        contentLength: text.length,
+        contentLength: processedText.length,
       });
       return;
     }
@@ -850,10 +886,10 @@ async function deliverWecomReply({ payload, account, responseUrl, senderId, stre
     return;
   }
 
-  appendToStream(streamId, text);
+  appendToStream(streamId, processedText);
   logger.debug("WeCom stream appended", {
     streamId,
-    contentLength: text.length,
+    contentLength: processedText.length,
     to: senderId
   });
 }

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@sunnoy/wecom",
-    "version": "1.0.0",
+    "version": "1.1.0",
     "description": "Enterprise WeChat AI Bot channel plugin for OpenClaw",
     "type": "module",
     "main": "index.js",

package/webhook.js

@@ -209,6 +209,50 @@ export class WecomWebhook {
                 query: { timestamp, nonce },
             };
         }
+        else if (msgtype === "voice") {
+            // Voice message (single chat only) - WeCom automatically transcribes to text
+            const content = data.voice?.content || "";
+            const msgId = data.msgid || `msg_${Date.now()}`;
+            const fromUser = data.from?.userid || "";
+            const responseUrl = data.response_url || "";
+            const chatType = data.chattype || "single";
+            const chatId = data.chatid || "";
+
+            // Check for duplicates
+            if (this.deduplicator.isDuplicate(msgId)) {
+                logger.debug("Duplicate voice message ignored", { msgId });
+                return null;
+            }
+
+            // Validate content
+            if (!content.trim()) {
+                logger.warn("Empty voice message received", { msgId, fromUser });
+                return null;
+            }
+
+            logger.info("Received voice message (auto-transcribed by WeCom)", {
+                fromUser,
+                chatType,
+                chatId: chatId || "(private)",
+                originalType: "voice",
+                transcribedLength: content.length,
+                preview: content.substring(0, 50)
+            });
+
+            // Treat voice as text since WeCom already transcribed it
+            return {
+                message: {
+                    msgId,
+                    msgType: "text",
+                    content,
+                    fromUser,
+                    chatType,
+                    chatId,
+                    responseUrl,
+                },
+                query: { timestamp, nonce },
+            };
+        }
         else if (msgtype === "event") {
             logger.info("Received event", { event: data.event });
             return {