@xmanrui/dsh-im 0.9.0 → 0.10.0

package/src/channels/weixin/weixin-bridge.mjs

@@ -1,4 +1,5 @@
 import {
+  extractWeixinImages,
   extractWeixinText,
   splitWeixinText,
   weixinMessageId,
@@ -12,13 +13,18 @@ import { HarnessApprovalQueue } from '../shared/harness-approval.mjs';
 import { runCompactCommand } from '../shared/compact-command.mjs';
 import { runWorkspaceCommand } from '../shared/workspace-command.mjs';
 import { askInWorkspaceSession } from '../shared/workspace-session.mjs';
+import {
+  hasInboundImages,
+  imagePromptUserMessage,
+  promptContentForMessage,
+} from '../shared/image-prompt.mjs';
 
 const INTERACTION_RESOLVED_TEXT = '这个问题已在其他客户端处理，无需再次回答。';
 
 const HELP_TEXT = [
   '微信已连接 DeepSeek Harness。',
   '',
-  '直接发送文字或带文字识别结果的语音即可继续当前会话。',
+  '直接发送文字、图片或带文字识别结果的语音即可继续当前会话。',
   '/new  开启一个全新会话',
   '/compact  压缩当前会话的较早上下文',
   '/workspace 工作区绝对路径  切换工作区',
@@ -37,9 +43,15 @@ function nonEmptyString(value) {
   return typeof value === 'string' && value.trim() ? value.trim() : null;
 }
 
+function hasWeixinImageItems(message) {
+  return Array.isArray(message?.item_list)
+    && message.item_list.some((item) => item?.image_item && typeof item.image_item === 'object');
+}
+
 function canClaimInteractionReply(message, pending) {
   return pending.questions[pending.index]
     && nonEmptyString(message?.from_user_id) === pending.actor
+    && !hasWeixinImageItems(message)
     && nonEmptyString(extractWeixinText(message));
 }
 
@@ -124,7 +136,7 @@ export class WeixinHarnessBridge {
       key,
       actor: sender,
       messageId,
-      text: extractWeixinText(message),
+      text: hasWeixinImageItems(message) ? '' : extractWeixinText(message),
       addressed: true,
       hasPendingQuestion: Boolean(pending),
       questionCompletion: pending?.submitting || pending?.claimedReplyMessageId
@@ -216,33 +228,40 @@ export class WeixinHarnessBridge {
 
     const contextToken = typeof message.context_token === 'string' ? message.context_token : undefined;
     const runId = typeof message.run_id === 'string' ? message.run_id : undefined;
-    const text = extractWeixinText(message);
+    const text = extractWeixinText(message) ?? '';
     try {
-      if (!text) {
-        await this.#send(sender, '目前仅支持文字消息，以及微信已转成文字的语音消息。', contextToken, runId);
+      const images = typeof this.#api.inboundImages === 'function'
+        ? this.#api.inboundImages(message)
+        : extractWeixinImages(message);
+      const promptMessage = { content: text, images };
+      const hasImages = hasInboundImages(promptMessage);
+      if (!text && !hasImages) {
+        await this.#send(sender, '目前支持文字、图片，以及微信已转成文字的语音消息。', contextToken, runId);
         await this.#state.markSeen(messageId);
         return;
       }
 
       const command = text.trim().toLowerCase();
-      if (command === '/help') {
+      if (!hasImages && command === '/help') {
         await this.#send(sender, HELP_TEXT, contextToken, runId);
         await this.#state.markSeen(messageId);
         return;
       }
-      if (command === '/status') {
+      if (!hasImages && command === '/status') {
         await this.#harness.ensureRunning({ signal: this.#signal });
         await this.#send(sender, '微信与 DeepSeek Harness 连接正常。', contextToken, runId);
         await this.#state.markSeen(messageId);
         return;
       }
-      if (command === '/new') {
+      if (!hasImages && command === '/new') {
         await this.#state.clearSession(key);
         await this.#send(sender, '已开启新会话。请发送你的问题。', contextToken, runId);
         await this.#state.markSeen(messageId);
         return;
       }
-      const workspaceCommand = await runWorkspaceCommand(text, this.#harness, key);
+      const workspaceCommand = hasImages
+        ? null
+        : await runWorkspaceCommand(text, this.#harness, key);
       if (workspaceCommand) {
         for (const reply of workspaceCommand.messages ?? [workspaceCommand.message]) {
           await this.#send(sender, reply, contextToken, runId);
@@ -250,26 +269,31 @@ export class WeixinHarnessBridge {
         await this.#state.markSeen(messageId);
         return;
       }
-      const compactCommand = await runCompactCommand(
-        text,
-        this.#harness,
-        this.#state,
-        key,
-        { signal: this.#signal },
-      );
+      const compactCommand = hasImages
+        ? null
+        : await runCompactCommand(
+            text,
+            this.#harness,
+            this.#state,
+            key,
+            { signal: this.#signal },
+          );
       if (compactCommand) {
         await this.#send(sender, compactCommand.message, contextToken, runId);
         await this.#state.markSeen(messageId);
         return;
       }
 
+      const content = hasImages
+        ? await promptContentForMessage(promptMessage, { signal: this.#signal })
+        : undefined;
       let answer;
       try {
         ({ answer } = await askInWorkspaceSession({
           harness: this.#harness,
           state: this.#state,
           key,
-          text,
+          ...(hasImages ? { content } : { text }),
           createOptions: { signal: this.#signal },
           existsOptions: { signal: this.#signal },
           askOptions: {
@@ -300,7 +324,12 @@ export class WeixinHarnessBridge {
       this.#status.lastError = error?.message ?? String(error);
       this.#logger.error?.('[dsh-weixin] failed to process an inbound message:', error);
       try {
-        await this.#send(sender, '消息处理失败，请稍后重试。', contextToken, runId);
+        await this.#send(
+          sender,
+          imagePromptUserMessage(error) ?? '消息处理失败，请稍后重试。',
+          contextToken,
+          runId,
+        );
         await this.#state.markSeen(messageId);
       } catch (sendError) {
         this.#logger.error?.('[dsh-weixin] failed to send the safe error reply:', sendError);
@@ -326,7 +355,7 @@ export class WeixinHarnessBridge {
     const text = nonEmptyString(extractWeixinText(message));
     const contextToken = nonEmptyString(message?.context_token) ?? undefined;
     const runId = nonEmptyString(message?.run_id) ?? undefined;
-    if (!text) {
+    if (!text || hasWeixinImageItems(message)) {
       await this.#send(
         expected.actor,
         '请用文字回答当前问题。',

package/src/channels/whatsapp/whatsapp-runtime.mjs

@@ -1,12 +1,45 @@
 import {
   areJidsSameUser,
+  downloadMediaMessage,
   normalizeMessageContent,
 } from '@whiskeysockets/baileys';
 
 import { splitMessageText } from '../shared/editable-message-stream.mjs';
+import { ImagePromptError } from '../shared/image-prompt.mjs';
 import { createWhatsappBridgeStatus, WhatsappHarnessBridge } from './whatsapp-bridge.mjs';
 import { createWhatsappWebSession } from './whatsapp-web-session.mjs';
 
+const IMAGE_MEDIA_TYPES = new Set(['image/jpeg', 'image/png', 'image/webp', 'image/gif']);
+const DEFAULT_MAX_IMAGE_BYTES = 5 * 1024 * 1024;
+const IMAGE_DOWNLOAD_TIMEOUT_MS = 15_000;
+const MESSAGE_WRAPPER_KEYS = [
+  'ephemeralMessage',
+  'viewOnceMessage',
+  'documentWithCaptionMessage',
+  'viewOnceMessageV2',
+  'viewOnceMessageV2Extension',
+  'editedMessage',
+  'associatedChildMessage',
+  'groupStatusMessage',
+  'groupStatusMessageV2',
+];
+const VIEW_ONCE_WRAPPER_KEYS = new Set([
+  'viewOnceMessage',
+  'viewOnceMessageV2',
+  'viewOnceMessageV2Extension',
+]);
+
+function hasViewOnceWrapper(content) {
+  let current = content;
+  for (let depth = 0; depth < 5 && current && typeof current === 'object'; depth += 1) {
+    const wrapperKey = MESSAGE_WRAPPER_KEYS.find((key) => current[key]?.message);
+    if (!wrapperKey) return false;
+    if (VIEW_ONCE_WRAPPER_KEYS.has(wrapperKey)) return true;
+    current = current[wrapperKey].message;
+  }
+  return false;
+}
+
 function messageContext(content) {
   return content?.extendedTextMessage?.contextInfo
     ?? content?.imageMessage?.contextInfo
@@ -24,7 +57,114 @@ function messageText(content) {
     ?? '';
 }
 
-export function normalizeWhatsappMessage(message, accountJid) {
+function mediaSize(value) {
+  if (Number.isSafeInteger(value) && value >= 0) return value;
+  let converted;
+  try {
+    converted = Number(value?.toString?.());
+  } catch {
+    return undefined;
+  }
+  return Number.isSafeInteger(converted) && converted >= 0 ? converted : undefined;
+}
+
+async function downloadWhatsappImage(message, download, {
+  signal,
+  maxBytes = DEFAULT_MAX_IMAGE_BYTES,
+} = {}) {
+  signal?.throwIfAborted();
+  const timeout = AbortSignal.timeout(IMAGE_DOWNLOAD_TIMEOUT_MS);
+  const downloadSignal = signal ? AbortSignal.any([signal, timeout]) : timeout;
+  const pendingStream = Promise.resolve().then(() => (
+    download(message, 'stream', { options: { signal: downloadSignal } })
+  ));
+  const stream = await new Promise((resolve, reject) => {
+    let settled = false;
+    const finish = (callback, value) => {
+      if (settled) return false;
+      settled = true;
+      downloadSignal.removeEventListener('abort', onAbort);
+      callback(value);
+      return true;
+    };
+    const onAbort = () => finish(reject, downloadSignal.reason);
+    downloadSignal.addEventListener('abort', onAbort, { once: true });
+    pendingStream.then((value) => {
+      if (!finish(resolve, value)) value?.destroy?.();
+    }, (error) => finish(reject, error));
+    if (downloadSignal.aborted) onAbort();
+  }).catch((error) => {
+    if (signal?.aborted) throw signal.reason ?? error;
+    if (timeout.aborted) {
+      throw new ImagePromptError(
+        'image-download-failed',
+        `WhatsApp image download timed out after ${IMAGE_DOWNLOAD_TIMEOUT_MS} ms`,
+        '图片下载失败，请重新发送后再试。',
+      );
+    }
+    throw error;
+  });
+  const chunks = [];
+  let size = 0;
+  const abortStream = () => stream?.destroy?.(downloadSignal.reason);
+  downloadSignal.addEventListener('abort', abortStream, { once: true });
+  try {
+    for await (const chunk of stream) {
+      downloadSignal.throwIfAborted();
+      const data = Buffer.from(chunk);
+      size += data.length;
+      if (size > maxBytes) {
+        stream.destroy?.();
+        throw new ImagePromptError(
+          'image-too-large',
+          `WhatsApp image exceeded ${maxBytes} bytes`,
+          '图片超过 5 MB，请压缩后重试。',
+        );
+      }
+      chunks.push(data);
+    }
+  } catch (error) {
+    if (signal?.aborted) throw signal.reason ?? error;
+    if (timeout.aborted) {
+      throw new ImagePromptError(
+        'image-download-failed',
+        `WhatsApp image stream timed out after ${IMAGE_DOWNLOAD_TIMEOUT_MS} ms`,
+        '图片下载失败，请重新发送后再试。',
+      );
+    }
+    throw error;
+  } finally {
+    downloadSignal.removeEventListener('abort', abortStream);
+  }
+  return Buffer.concat(chunks, size);
+}
+
+function whatsappImageSource(message, content, download, { viewOnce = false } = {}) {
+  let media;
+  let name;
+  if (!viewOnce && content?.imageMessage && content.imageMessage.viewOnce !== true) {
+    media = content.imageMessage;
+  } else if (content?.documentMessage) {
+    const type = typeof content.documentMessage.mimetype === 'string'
+      ? content.documentMessage.mimetype.toLowerCase() : '';
+    if (!IMAGE_MEDIA_TYPES.has(type)) return null;
+    media = content.documentMessage;
+    name = typeof media.fileName === 'string' ? media.fileName : undefined;
+  }
+  if (!media) return null;
+  const mediaType = typeof media.mimetype === 'string' ? media.mimetype.toLowerCase() : '';
+  if (!IMAGE_MEDIA_TYPES.has(mediaType)) return null;
+  return {
+    name,
+    mediaType,
+    size: mediaSize(media.fileLength),
+    load: (options) => downloadWhatsappImage(message, download, options),
+  };
+}
+
+export function normalizeWhatsappMessage(message, accountJid, {
+  download = downloadMediaMessage,
+} = {}) {
   const remoteJid = typeof message?.key?.remoteJid === 'string' ? message.key.remoteJid : '';
   const alternateRemoteJid = typeof message?.key?.remoteJidAlt === 'string'
     ? message.key.remoteJidAlt : '';
@@ -38,12 +178,14 @@ export function normalizeWhatsappMessage(message, accountJid) {
   if (fromMe && !selfChat) return null;
   const senderJid = selfChat ? accountJid : group ? message.key.participant : remoteJid;
   if (typeof senderJid !== 'string' || !senderJid) return null;
+  const viewOnce = hasViewOnceWrapper(message.message);
   const content = normalizeMessageContent(message.message);
   const context = messageContext(content);
   const mentioned = Array.isArray(context?.mentionedJid)
     && context.mentionedJid.some((jid) => areJidsSameUser(jid, accountJid));
   const replyToSelf = typeof context?.participant === 'string'
     && areJidsSameUser(context.participant, accountJid);
+  const image = whatsappImageSource(message, content, download, { viewOnce });
   return {
     messageId: `${remoteJid}:${messageId}`,
     providerMessageId: messageId,
@@ -52,6 +194,7 @@ export function normalizeWhatsappMessage(message, accountJid) {
     kind: group ? 'group' : 'direct',
     conversationId: remoteJid,
     content: messageText(content),
+    images: image ? [image] : [],
     addressed: !group || mentioned || replyToSelf,
     selfChat,
     replyTarget: { jid: remoteJid, quoted: message, selfChat },