@marshulll/openclaw-wecom 0.1.9 → 0.1.10

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@marshulll/openclaw-wecom",
-  "version": "0.1.9",
+  "version": "0.1.10",
   "type": "module",
   "description": "OpenClaw WeCom channel plugin (intelligent bot + internal app)",
   "author": "OpenClaw",

package/wecom/src/wecom-bot.ts

@@ -31,6 +31,17 @@ type StreamState = {
   content: string;
 };
 
+type InboundMedia = {
+  path: string;
+  type: string;
+  url?: string;
+};
+
+type InboundBody = {
+  text: string;
+  media?: InboundMedia;
+};
+
 const streams = new Map<string, StreamState>();
 const msgidToStreamId = new Map<string, string>();
 const recentEncrypts = new Map<string, { ts: number; streamId?: string }>();
@@ -320,7 +331,8 @@ async function startAgentForStream(params: {
   const userid = msg.from?.userid?.trim() || "unknown";
   const chatType = msg.chattype === "group" ? "group" : "direct";
   const chatId = msg.chattype === "group" ? (msg.chatid?.trim() || "unknown") : userid;
-  const rawBody = await buildInboundBody({ target, msg });
+  const inbound = await buildInboundBody({ target, msg });
+  const rawBody = inbound.text;
 
   const route = core.channel.routing.resolveAgentRoute({
     cfg: config,
@@ -367,6 +379,14 @@ async function startAgentForStream(params: {
     OriginatingTo: `wecom:${chatId}`,
   });
 
+  if (inbound.media) {
+    ctxPayload.MediaPath = inbound.media.path;
+    ctxPayload.MediaType = inbound.media.type;
+    if (inbound.media.url) {
+      ctxPayload.MediaUrl = inbound.media.url;
+    }
+  }
+
   await core.channel.session.recordInboundSession({
     storePath,
     sessionKey: ctxPayload.SessionKey ?? route.sessionKey,
@@ -550,7 +570,7 @@ async function buildBotMediaMessage(params: {
   url?: string;
   base64?: string;
   filename?: string;
-}): Promise<string> {
+}): Promise<InboundBody> {
   const { target, msgtype, url, base64, filename } = params;
 
   const fallbackLabel = msgtype === "image"
@@ -561,28 +581,31 @@ async function buildBotMediaMessage(params: {
         ? "[video]"
         : "[file]";
 
-  if (!url && !base64) return fallbackLabel;
+  if (!url && !base64) return { text: fallbackLabel };
 
   try {
     let buffer: Buffer | null = null;
     let contentType = "";
     if (base64) {
       buffer = Buffer.from(base64, "base64");
-      contentType = msgtype === "image" ? "image/jpeg" : "application/octet-stream";
+      if (msgtype === "image") contentType = "image/jpeg";
+      else if (msgtype === "voice") contentType = "audio/amr";
+      else if (msgtype === "video") contentType = "video/mp4";
+      else contentType = "application/octet-stream";
     } else if (url) {
       const media = await fetchMediaFromUrl(url, target.account);
       buffer = media.buffer;
       contentType = media.contentType;
     }
 
-    if (!buffer) return fallbackLabel;
+    if (!buffer) return { text: fallbackLabel };
 
     const maxBytes = resolveMediaMaxBytes(target);
     if (maxBytes && buffer.length > maxBytes) {
-      if (msgtype === "image") return "[图片过大，未处理]\n\n请发送更小的图片。";
-      if (msgtype === "voice") return "[语音消息过大，未处理]\n\n请发送更短的语音消息。";
-      if (msgtype === "video") return "[视频过大，未处理]\n\n请发送更小的视频。";
-      return "[文件过大，未处理]\n\n请发送更小的文件。";
+      if (msgtype === "image") return { text: "[图片过大，未处理]\n\n请发送更小的图片。" };
+      if (msgtype === "voice") return { text: "[语音消息过大，未处理]\n\n请发送更短的语音消息。" };
+      if (msgtype === "video") return { text: "[视频过大，未处理]\n\n请发送更小的视频。" };
+      return { text: "[文件过大，未处理]\n\n请发送更小的文件。" };
     }
 
     const tempDir = resolveMediaTempDir(target);
@@ -606,7 +629,10 @@ async function buildBotMediaMessage(params: {
       const safeName = sanitizeFilename(filename || "", `file-${Date.now()}.${ext}`);
       const tempFilePath = join(tempDir, safeName);
       await writeFile(tempFilePath, buffer);
-      return `[用户发送了一个文件: ${safeName}，已保存到: ${tempFilePath}]\n\n请根据文件内容回复用户。`;
+      return {
+        text: `[用户发送了一个文件: ${safeName}，已保存到: ${tempFilePath}]\n\n请根据文件内容回复用户。`,
+        media: { path: tempFilePath, type: contentType || "application/octet-stream", url },
+      };
     }
 
     const tempPath = join(
@@ -616,34 +642,43 @@ async function buildBotMediaMessage(params: {
     await writeFile(tempPath, buffer);
 
     if (msgtype === "image") {
-      return `[用户发送了一张图片，已保存到: ${tempPath}]\n\n请使用 Read 工具查看这张图片并描述内容。`;
+      return {
+        text: `[用户发送了一张图片，已保存到: ${tempPath}]\n\n请使用 Read 工具查看这张图片并描述内容。`,
+        media: { path: tempPath, type: contentType || "image/jpeg", url },
+      };
     }
     if (msgtype === "voice") {
-      return `[用户发送了一条语音消息，已保存到: ${tempPath}]\n\n请根据语音内容回复用户。`;
+      return {
+        text: `[用户发送了一条语音消息，已保存到: ${tempPath}]\n\n请根据语音内容回复用户。`,
+        media: { path: tempPath, type: contentType || "audio/amr", url },
+      };
     }
     if (msgtype === "video") {
-      return `[用户发送了一个视频文件，已保存到: ${tempPath}]\n\n请根据视频内容回复用户。`;
+      return {
+        text: `[用户发送了一个视频文件，已保存到: ${tempPath}]\n\n请根据视频内容回复用户。`,
+        media: { path: tempPath, type: contentType || "video/mp4", url },
+      };
     }
-    return fallbackLabel;
+    return { text: fallbackLabel };
   } catch (err) {
     target.runtime.error?.(`wecom bot ${msgtype} download failed: ${String(err)}`);
-    if (msgtype === "image") return "[用户发送了一张图片，但下载失败]\n\n请告诉用户图片处理暂时不可用。";
-    if (msgtype === "voice") return "[用户发送了一条语音消息，但下载失败]\n\n请告诉用户语音处理暂时不可用。";
-    if (msgtype === "video") return "[用户发送了一个视频，但下载失败]\n\n请告诉用户视频处理暂时不可用。";
-    return "[用户发送了一个文件，但下载失败]\n\n请告诉用户文件处理暂时不可用。";
+    if (msgtype === "image") return { text: "[用户发送了一张图片，但下载失败]\n\n请告诉用户图片处理暂时不可用。" };
+    if (msgtype === "voice") return { text: "[用户发送了一条语音消息，但下载失败]\n\n请告诉用户语音处理暂时不可用。" };
+    if (msgtype === "video") return { text: "[用户发送了一个视频，但下载失败]\n\n请告诉用户视频处理暂时不可用。" };
+    return { text: "[用户发送了一个文件，但下载失败]\n\n请告诉用户文件处理暂时不可用。" };
   }
 }
 
-async function buildInboundBody(params: { target: WecomWebhookTarget; msg: WecomInboundMessage }): Promise<string> {
+async function buildInboundBody(params: { target: WecomWebhookTarget; msg: WecomInboundMessage }): Promise<InboundBody> {
   const { target, msg } = params;
   const msgtype = String(msg.msgtype ?? "").toLowerCase();
   if (msgtype === "text") {
     const content = (msg as any).text?.content;
-    return typeof content === "string" ? content : "";
+    return { text: typeof content === "string" ? content : "" };
   }
   if (msgtype === "voice") {
     const content = (msg as any).voice?.content;
-    if (typeof content === "string" && content.trim()) return content;
+    if (typeof content === "string" && content.trim()) return { text: content.trim() };
     const url = resolveBotMediaUrl(msg as any, "voice");
     const base64 = resolveBotMediaBase64(msg as any, "voice");
     return await buildBotMediaMessage({ target, msgtype: "voice", url, base64 });
@@ -651,7 +686,7 @@ async function buildInboundBody(params: { target: WecomWebhookTarget; msg: Wecom
   if (msgtype === "mixed") {
     const items = (msg as any).mixed?.msg_item;
     if (Array.isArray(items)) {
-      return items
+      const text = items
         .map((item: any) => {
           const t = String(item?.msgtype ?? "").toLowerCase();
           if (t === "text") return String(item?.text?.content ?? "");
@@ -660,8 +695,9 @@ async function buildInboundBody(params: { target: WecomWebhookTarget; msg: Wecom
         })
         .filter((part: string) => Boolean(part && part.trim()))
         .join("\n");
+      return { text };
     }
-    return "[mixed]";
+    return { text: "[mixed]" };
   }
   if (msgtype === "image") {
     const url = resolveBotMediaUrl(msg as any, "image");
@@ -681,13 +717,13 @@ async function buildInboundBody(params: { target: WecomWebhookTarget; msg: Wecom
   }
   if (msgtype === "event") {
     const eventtype = String((msg as any).event?.eventtype ?? "").trim();
-    return eventtype ? `[event] ${eventtype}` : "[event]";
+    return { text: eventtype ? `[event] ${eventtype}` : "[event]" };
   }
   if (msgtype === "stream") {
     const id = String((msg as any).stream?.id ?? "").trim();
-    return id ? `[stream_refresh] ${id}` : "[stream_refresh]";
+    return { text: id ? `[stream_refresh] ${id}` : "[stream_refresh]" };
   }
-  return msgtype ? `[${msgtype}]` : "";
+  return { text: msgtype ? `[${msgtype}]` : "" };
 }
 
 function normalizeMediaType(raw?: string): "image" | "voice" | "video" | "file" | null {