@xmanrui/dsh-im 1.2.0 → 1.3.0

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

@@ -26,6 +26,10 @@ import {
   imagePromptUserMessage,
   promptContentForMessage,
 } from '../shared/image-prompt.mjs';
+import {
+  hasInboundFiles,
+  inboundFileUserMessage,
+} from '../shared/inbound-file.mjs';
 import { rememberConnectionTestTarget } from '../shared/connection-test.mjs';
 import {
   materializeOutboundArtifact,
@@ -43,7 +47,7 @@ const DEFAULT_FILE_UPLOAD_TIMEOUT_MS = 120_000;
 const HELP_TEXT = [
   '企业微信机器人已连接 DeepSeek Harness。',
   '',
-  '直接发送文字或图片即可继续当前会话。',
+  '直接发送文字、图片或文件即可继续当前会话。',
   '/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',
@@ -110,6 +114,13 @@ function imageContents(frame) {
     .map((item) => item.image);
 }
 
+function fileContents(frame) {
+  const body = bodyOf(frame);
+  return body.msgtype === 'file' && body.file && typeof body.file === 'object'
+    ? [body.file]
+    : [];
+}
+
 function imageSource(client, image) {
   const url = nonEmptyString(image?.url);
   if (!url) return null;
@@ -139,10 +150,56 @@ function imageSource(client, image) {
   };
 }
 
+function fileSource(client, file) {
+  const url = nonEmptyString(file?.url);
+  if (!url) return null;
+  const aeskey = nonEmptyString(file?.aeskey) ?? undefined;
+  return {
+    name: nonEmptyString(file?.filename ?? file?.file_name ?? file?.name) ?? 'file',
+    async load({ signal } = {}) {
+      signal?.throwIfAborted();
+      if (typeof client?.downloadFile !== 'function') {
+        throw new Error('Enterprise WeChat file download is unavailable');
+      }
+      const result = await client.downloadFile(url, aeskey);
+      signal?.throwIfAborted();
+      const raw = result?.buffer ?? result?.data;
+      if (!Buffer.isBuffer(raw) && !(raw instanceof Uint8Array)) {
+        throw new Error('Enterprise WeChat file download returned no data');
+      }
+      return {
+        data: Buffer.from(raw),
+        ...(nonEmptyString(result?.filename) ? { name: result.filename.trim() } : {}),
+      };
+    },
+  };
+}
+
 export function wecomInboundMessage(frame, client) {
   return {
     content: messageText(frame),
     images: imageContents(frame).map((image) => imageSource(client, image)).filter(Boolean),
+    files: fileContents(frame).map((file) => fileSource(client, file)).filter(Boolean),
+  };
+}
+
+function prefetchInboundFiles(message, signal) {
+  if (!Array.isArray(message?.files) || message.files.length === 0) return message;
+  return {
+    ...message,
+    files: message.files.map((source) => {
+      const download = source.load({ signal });
+      download.catch(() => undefined);
+      return {
+        ...source,
+        async load({ signal: loadSignal } = {}) {
+          loadSignal?.throwIfAborted();
+          const result = await download;
+          loadSignal?.throwIfAborted();
+          return result;
+        },
+      };
+    }),
   };
 }
 
@@ -400,7 +457,7 @@ export class WecomHarnessBridge {
     const pending = this.#pendingInteractions.get(key);
     const commandMessage = wecomInboundMessage(frame, this.#client);
     const commandText = nonEmptyString(commandMessage.content) ?? '';
-    const commandRunner = isControlCommand(commandText)
+    const commandRunner = hasInboundFiles(commandMessage) ? null : isControlCommand(commandText)
       ? runControlCommand
       : (isModelCommand(commandText)
           ? runModelCommand
@@ -510,6 +567,7 @@ export class WecomHarnessBridge {
         preparedMessage = imageQueueFullMessage(inboundMessage);
       }
     }
+    preparedMessage = prefetchInboundFiles(preparedMessage, this.#signal);
     const previous = this.#queues.get(key) ?? Promise.resolve();
     const current = previous
       .catch(() => undefined)
@@ -543,6 +601,7 @@ export class WecomHarnessBridge {
     const result = await runner(message.content, this.#harness, this.#state, key, {
       signal: this.#signal,
       hasImages: hasInboundImages(message),
+      hasFiles: hasInboundFiles(message),
       pendingInteraction: this.#pendingInteractions.has(key)
         || this.#approvals.hasPending(key),
       control: { owner: this, key },
@@ -701,34 +760,35 @@ export class WecomHarnessBridge {
     const message = preparedMessage ?? wecomInboundMessage(frame, this.#client);
     const text = message.content;
     const hasImages = hasInboundImages(message);
+    const hasFiles = hasInboundFiles(message);
     const key = conversationKey(frame);
     let streamId = null;
     let streamStarted = false;
     try {
-      if (!text && !hasImages) {
-        await this.#sendImmediate(frame, chatId, '目前支持文字、图片和语音转写消息。');
+      if (!text && !hasImages && !hasFiles) {
+        await this.#sendImmediate(frame, chatId, '目前支持文字、图片、文件和语音转写消息。');
         await this.#state.markSeen(messageId);
         return;
       }
       const command = text.toLowerCase();
-      if (!hasImages && command === '/help') {
+      if (!hasImages && !hasFiles && command === '/help') {
         await this.#sendImmediate(frame, chatId, HELP_TEXT);
         await this.#state.markSeen(messageId);
         return;
       }
-      if (!hasImages && command === '/status') {
+      if (!hasImages && !hasFiles && command === '/status') {
         await this.#harness.ensureRunning({ signal: this.#signal });
         await this.#sendImmediate(frame, chatId, '企业微信机器人与 DeepSeek Harness 连接正常。');
         await this.#state.markSeen(messageId);
         return;
       }
-      if (!hasImages && command === '/new') {
+      if (!hasImages && !hasFiles && command === '/new') {
         await this.#state.clearSession(key);
         await this.#sendImmediate(frame, chatId, '已开启新会话。请发送你的问题。');
         await this.#state.markSeen(messageId);
         return;
       }
-      const workspaceCommand = hasImages
+      const workspaceCommand = hasImages || hasFiles
         ? null
         : await runWorkspaceCommand(text, this.#harness, key);
       if (workspaceCommand) {
@@ -738,7 +798,7 @@ export class WecomHarnessBridge {
         await this.#state.markSeen(messageId);
         return;
       }
-      const compactCommand = hasImages
+      const compactCommand = hasImages || hasFiles
         ? null
         : await runCompactCommand(
             text,
@@ -789,6 +849,7 @@ export class WecomHarnessBridge {
             requiresMention: body.chattype === 'group',
           }),
           onInteractionResolved: (resolution) => this.#handleInteractionResolved(resolution),
+          files: message.files,
         },
       });
 
@@ -862,7 +923,9 @@ export class WecomHarnessBridge {
       if (this.#signal?.aborted) return;
       this.#status.lastError = error?.message ?? String(error);
       this.#logger.error?.('[dsh-im:wecom] failed to process an inbound message');
-      const errorText = imagePromptUserMessage(error) ?? '消息处理失败，请稍后重试。';
+      const errorText = inboundFileUserMessage(error)
+        ?? imagePromptUserMessage(error)
+        ?? '消息处理失败，请稍后重试。';
       try {
         if (streamStarted && streamId) {
           await this.#client.replyStream(frame, streamId, errorText, true);

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

@@ -196,6 +196,47 @@ export function extractWeixinImages(message, { fetchImpl = fetch } = {}) {
   return images;
 }
 
+async function fetchWeixinFileCiphertext(url, { fetchImpl, signal }) {
+  const response = await fetchImpl(new URL(url), {
+    method: 'GET',
+    signal,
+    redirect: 'manual',
+  });
+  if (!response?.ok) {
+    await response?.body?.cancel?.().catch?.(() => undefined);
+    throw new WeixinApiError(
+      'file-download-failed',
+      `微信文件下载失败（HTTP ${response?.status ?? 'unknown'}）。`,
+      { status: response?.status },
+    );
+  }
+  return Buffer.from(await response.arrayBuffer());
+}
+
+/** Convert native iLink file items into lazily downloaded, decrypted file references. */
+export function extractWeixinFiles(message, { fetchImpl = fetch } = {}) {
+  if (typeof fetchImpl !== 'function') throw new TypeError('fetchImpl must be a function');
+  const files = [];
+  for (const item of message?.item_list ?? []) {
+    const fileItem = item?.file_item;
+    if (!fileItem || typeof fileItem !== 'object') continue;
+    const declaredSize = Number(fileItem.len);
+    files.push({
+      name: nonEmptyString(fileItem.file_name) ?? (files.length === 0 ? 'file' : `file-${files.length + 1}`),
+      ...(Number.isFinite(declaredSize) && declaredSize >= 0 ? { size: declaredSize } : {}),
+      load: async ({ signal } = {}) => {
+        signal?.throwIfAborted();
+        const key = parseWeixinImageAesKey(fileItem);
+        const url = weixinImageDownloadUrl(fileItem.media);
+        const ciphertext = await fetchWeixinFileCiphertext(url, { fetchImpl, signal });
+        signal?.throwIfAborted();
+        return decryptWeixinImage(ciphertext, key);
+      },
+    });
+  }
+  return files;
+}
+
 function isWeixinHost(hostname) {
   const normalized = hostname.toLowerCase().replace(/\.$/, '');
   return normalized === 'weixin.qq.com' || normalized.endsWith('.weixin.qq.com');
@@ -421,6 +462,10 @@ export function createWeixinApi({ fetchImpl = fetch } = {}) {
       return extractWeixinImages(message, { fetchImpl });
     },
 
+    inboundFiles(message) {
+      return extractWeixinFiles(message, { fetchImpl });
+    },
+
     async beginLogin({ localTokens = [], botType = DEFAULT_BOT_TYPE, signal } = {}) {
       const tokens = [...new Set(localTokens.map(nonEmptyString).filter(Boolean))].slice(-10);
       const response = await requestJson(fetchImpl, {

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

@@ -1,4 +1,5 @@
 import {
+  extractWeixinFiles,
   extractWeixinImages,
   extractWeixinText,
   splitWeixinText,
@@ -31,6 +32,11 @@ import {
   imagePromptUserMessage,
   promptContentForMessage,
 } from '../shared/image-prompt.mjs';
+import {
+  hasInboundFiles,
+  inboundFileUserMessage,
+  prefetchInboundFiles,
+} from '../shared/inbound-file.mjs';
 import { rememberConnectionTestTarget } from '../shared/connection-test.mjs';
 import {
   materializeOutboundArtifact,
@@ -49,7 +55,7 @@ const GENERIC_PROCESSING_ERROR = '消息处理失败，请稍后重试。';
 const HELP_TEXT = [
   '微信已连接 DeepSeek Harness。',
   '',
-  '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。',
+  '直接发送文字、图片、文件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',
@@ -77,15 +83,33 @@ function nonEmptyString(value) {
   return typeof value === 'string' && value.trim() ? value.trim() : null;
 }
 
+export function weixinInboundMessage(message, api) {
+  return {
+    content: extractWeixinText(message) ?? '',
+    images: typeof api?.inboundImages === 'function'
+      ? api.inboundImages(message)
+      : extractWeixinImages(message),
+    files: typeof api?.inboundFiles === 'function'
+      ? api.inboundFiles(message)
+      : extractWeixinFiles(message),
+  };
+}
+
 function hasWeixinImageItems(message) {
   return Array.isArray(message?.item_list)
     && message.item_list.some((item) => item?.image_item && typeof item.image_item === 'object');
 }
 
+function hasWeixinFileItems(message) {
+  return Array.isArray(message?.item_list)
+    && message.item_list.some((item) => item?.file_item && typeof item.file_item === 'object');
+}
+
 function canClaimInteractionReply(message, pending) {
   return pending.questions[pending.index]
     && nonEmptyString(message?.from_user_id) === pending.actor
     && !hasWeixinImageItems(message)
+    && !hasWeixinFileItems(message)
     && nonEmptyString(extractWeixinText(message));
 }
 
@@ -204,7 +228,7 @@ export class WeixinHarnessBridge {
     const runId = nonEmptyString(message?.run_id) ?? undefined;
     const pending = this.#pendingInteractions.get(key);
     const commandText = nonEmptyString(extractWeixinText(message)) ?? '';
-    const commandRunner = isControlCommand(commandText)
+    const commandRunner = hasWeixinFileItems(message) ? null : isControlCommand(commandText)
       ? runControlCommand
       : (isModelCommand(commandText)
           ? runModelCommand
@@ -238,7 +262,9 @@ export class WeixinHarnessBridge {
       key,
       actor: sender,
       messageId,
-      text: hasWeixinImageItems(message) ? '' : extractWeixinText(message),
+      text: hasWeixinImageItems(message) || hasWeixinFileItems(message)
+        ? ''
+        : extractWeixinText(message),
       addressed: true,
       hasPendingQuestion: Boolean(pending),
       questionCompletion: pending?.submitting || pending?.claimedReplyMessageId
@@ -290,10 +316,16 @@ export class WeixinHarnessBridge {
     releaseMessageId = true,
     alreadyRecorded = false,
   } = {}) {
+    const preparedMessage = message.from_user_id === this.#ownerUserId
+      ? prefetchInboundFiles(
+          weixinInboundMessage(message, this.#api),
+          { signal: this.#signal },
+        )
+      : undefined;
     const previous = this.#queues.get(key) ?? Promise.resolve();
     const current = previous
       .catch(() => undefined)
-      .then(() => this.#process(message, key, { alreadyRecorded }))
+      .then(() => this.#process(message, key, { alreadyRecorded, preparedMessage }))
       .finally(() => {
         if (releaseMessageId) this.#acceptedMessageIds.delete(messageId);
         if (this.#queues.get(key) === current) this.#queues.delete(key);
@@ -331,6 +363,7 @@ export class WeixinHarnessBridge {
     const result = await runner(text, this.#harness, this.#state, key, {
       signal: this.#signal,
       hasImages: hasWeixinImageItems(message),
+      hasFiles: hasWeixinFileItems(message),
       pendingInteraction: this.#pendingInteractions.has(key)
         || this.#approvals.hasPending(key),
       control: { owner: this, key },
@@ -348,7 +381,7 @@ export class WeixinHarnessBridge {
     this.#status.lastMessageError = null;
   }
 
-  async #process(message, key, { alreadyRecorded = false } = {}) {
+  async #process(message, key, { alreadyRecorded = false, preparedMessage } = {}) {
     this.#signal?.throwIfAborted();
     const messageId = weixinMessageId(message);
     const sender = nonEmptyString(message?.from_user_id);
@@ -366,38 +399,36 @@ 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) ?? '';
     try {
-      const images = typeof this.#api.inboundImages === 'function'
-        ? this.#api.inboundImages(message)
-        : extractWeixinImages(message);
-      const promptMessage = { content: text, images };
+      const promptMessage = preparedMessage ?? weixinInboundMessage(message, this.#api);
+      const text = promptMessage.content;
       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);
+      const hasFiles = hasInboundFiles(promptMessage);
+      if (!text && !hasImages && !hasFiles) {
+        await this.#send(sender, '目前支持文字、图片、文件，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 (!hasImages && command === '/help') {
+      if (!hasImages && !hasFiles && command === '/help') {
         await this.#send(sender, HELP_TEXT, contextToken, runId);
         await this.#state.markSeen(messageId);
         return;
       }
-      if (!hasImages && command === '/status') {
+      if (!hasImages && !hasFiles && 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 (!hasImages && command === '/new') {
+      if (!hasImages && !hasFiles && 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 = hasImages
+      const workspaceCommand = hasImages || hasFiles
         ? null
         : await runWorkspaceCommand(text, this.#harness, key);
       if (workspaceCommand) {
@@ -407,7 +438,7 @@ export class WeixinHarnessBridge {
         await this.#state.markSeen(messageId);
         return;
       }
-      const compactCommand = hasImages
+      const compactCommand = hasImages || hasFiles
         ? null
         : await runCompactCommand(
             text,
@@ -446,6 +477,7 @@ export class WeixinHarnessBridge {
               runId,
             }),
             onInteractionResolved: (resolution) => this.#handleInteractionResolved(resolution),
+            files: promptMessage.files,
           },
         }));
       } finally {
@@ -490,7 +522,9 @@ export class WeixinHarnessBridge {
       }
       if (this.#signal?.aborted) return;
       this.#status.lastError = error?.message ?? String(error);
-      const userMessage = imagePromptUserMessage(error) ?? GENERIC_PROCESSING_ERROR;
+      const userMessage = inboundFileUserMessage(error)
+        ?? imagePromptUserMessage(error)
+        ?? GENERIC_PROCESSING_ERROR;
       this.#status.lastMessageError = safeMessageError(error, userMessage);
       this.#logger.error?.('[dsh-weixin] failed to process an inbound message:', error);
       try {

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

@@ -170,6 +170,39 @@ function whatsappImageSource(message, content, download, { viewOnce = false } =
   };
 }
 
+function whatsappFileSource(message, content, download) {
+  const media = content?.documentMessage;
+  if (!media) return null;
+  const mediaType = typeof media.mimetype === 'string' && media.mimetype
+    ? media.mimetype.toLowerCase() : undefined;
+  if (IMAGE_MEDIA_TYPES.has(mediaType)) return null;
+  return {
+    name: typeof media.fileName === 'string' && media.fileName
+      ? media.fileName : 'whatsapp-file',
+    ...(mediaType ? { mediaType } : {}),
+    size: mediaSize(media.fileLength),
+    async load({ signal } = {}) {
+      signal?.throwIfAborted();
+      const stream = await download(message, 'stream', { options: { signal } });
+      signal?.throwIfAborted();
+      return { stream };
+    },
+  };
+}
+
+export function createWhatsappMediaDownloader({
+  socket,
+  logger = console,
+  download = downloadMediaMessage,
+} = {}) {
+  if (typeof download !== 'function') throw new TypeError('WhatsApp media downloader is required');
+  if (typeof socket?.updateMediaMessage !== 'function') return download;
+  return (message, type, options) => download(message, type, options, {
+    logger,
+    reuploadRequest: (candidate) => socket.updateMediaMessage(candidate),
+  });
+}
+
 export function normalizeWhatsappMessage(message, accountJid, {
   download = downloadMediaMessage,
 } = {}) {
@@ -195,6 +228,7 @@ export function normalizeWhatsappMessage(message, accountJid, {
   const replyToSelf = typeof context?.participant === 'string'
     && areJidsSameUser(context.participant, accountJid);
   const image = whatsappImageSource(message, content, download, { viewOnce });
+  const file = whatsappFileSource(message, content, download);
   return {
     messageId: `${remoteJid}:${messageId}`,
     providerMessageId: messageId,
@@ -205,6 +239,7 @@ export function normalizeWhatsappMessage(message, accountJid, {
     conversationId: remoteJid,
     content: messageText(content),
     images: image ? [image] : [],
+    files: file ? [file] : [],
     addressed: !group || mentioned || replyToSelf,
     selfChat,
     replyTarget: { jid: remoteJid, quoted: message, selfChat },
@@ -495,8 +530,13 @@ export class WhatsappRuntime {
           new Error('WhatsApp linked-device session must be scanned again'),
           { code: 'relink-required' },
         )),
-        onMessage: async (raw) => {
-          const message = normalizeWhatsappMessage(raw, this.#config.accountJid);
+        onMessage: async (raw, context) => {
+          const message = normalizeWhatsappMessage(raw, this.#config.accountJid, {
+            download: createWhatsappMediaDownloader({
+              socket: context?.socket,
+              logger: this.#logger,
+            }),
+          });
           if (!message || outboundIds.has(message.providerMessageId) || !this.#bridge) return;
           this.#status.lastCheckedAt = Date.now();
           if (!whatsappInboundAllowed(message, {

package/src/channels/whatsapp/whatsapp-web-session.mjs

@@ -190,7 +190,7 @@ export async function createWhatsappWebSession({
           const timestamp = messageTimestampMs(message?.messageTimestamp);
           if (timestamp === null || timestamp < sessionStartedAt - APPEND_RECENT_GRACE_MS) continue;
         }
-        Promise.resolve(onMessage(message)).catch(() => {
+        Promise.resolve(onMessage(message, { socket: nextSocket })).catch(() => {
           logger.error?.('[dsh-im:whatsapp] failed to process an inbound WhatsApp message');
         });
       }