@xmanrui/dsh-im 0.9.0 → 0.10.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xmanrui/dsh-im",
-  "version": "0.9.0",
+  "version": "0.10.0",
   "description": "通过扫码、App Manifest或机器人凭据把IM机器人接入DeepSeek Harness（支持飞书、微信、钉钉、企业微信、QQ、Slack、Telegram、Discord和WhatsApp）。",
   "license": "MIT",
   "type": "module",

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

@@ -1,5 +1,7 @@
 import { randomUUID } from 'node:crypto';
 
+import { fetchImageBuffer, ImagePromptError } from '../shared/image-prompt.mjs';
+
 export const DINGTALK_REGISTRATION_BASE_URL = 'https://oapi.dingtalk.com/';
 export const DINGTALK_API_BASE_URL = 'https://api.dingtalk.com/';
 export const DINGTALK_REGISTRATION_SOURCE = 'DING_DWS_CLAW';
@@ -14,6 +16,7 @@ export class DingtalkApiError extends Error {
     this.name = 'DingtalkApiError';
     this.code = code;
     this.status = options.status;
+    this.providerCode = options.providerCode;
   }
 }
 
@@ -21,6 +24,25 @@ function nonEmptyString(value) {
   return typeof value === 'string' && value.trim() ? value.trim() : null;
 }
 
+function safeProviderCode(value) {
+  const code = nonEmptyString(value);
+  return code && /^[A-Za-z0-9_.:-]{1,160}$/.test(code) ? code : undefined;
+}
+
+function secureDingtalkDownloadUrl(value) {
+  let url;
+  try {
+    url = new URL(value);
+  } catch (error) {
+    throw new DingtalkApiError('invalid-image-download', '钉钉服务返回了无效的图片下载地址。', { cause: error });
+  }
+  if (url.protocol === 'http:' && (!url.port || url.port === '80')) {
+    url.protocol = 'https:';
+    url.port = '';
+  }
+  return url;
+}
+
 function isDingtalkHost(hostname) {
   const normalized = hostname.toLowerCase().replace(/\.$/, '');
   return normalized === 'dingtalk.com' || normalized.endsWith('.dingtalk.com');
@@ -123,10 +145,17 @@ async function requestJson(fetchImpl, url, {
       signal: controller.signal,
     });
     if (!response.ok) {
+      let providerCode;
+      try {
+        const errorBody = await response.json();
+        providerCode = safeProviderCode(errorBody?.code ?? errorBody?.errcode);
+      } catch {
+        // DingTalk occasionally returns an empty or non-JSON error body.
+      }
       throw new DingtalkApiError(
         'http-error',
         `钉钉服务请求失败（HTTP ${response.status}）。`,
-        { status: response.status },
+        { status: response.status, providerCode },
       );
     }
     try {
@@ -403,6 +432,58 @@ export function createDingtalkApi({
 
     accessToken,
 
+    async downloadImage({
+      clientId,
+      clientSecret,
+      robotCode,
+      downloadCode,
+      signal,
+      maxBytes,
+    }) {
+      const botCode = nonEmptyString(robotCode);
+      const fileCode = nonEmptyString(downloadCode);
+      if (!botCode || !fileCode) throw new TypeError('robotCode and downloadCode are required');
+      const token = await accessToken({ clientId, clientSecret, signal });
+      let response;
+      try {
+        response = await requestJson(
+          fetchImpl,
+          endpoint(apiBase, 'v1.0/robot/messageFiles/download'),
+          {
+            body: { downloadCode: fileCode, robotCode: botCode },
+            headers: { 'x-acs-dingtalk-access-token': token },
+            signal,
+            action: '图片下载地址',
+          },
+        );
+      } catch (error) {
+        if (signal?.aborted) throw error;
+        throw new DingtalkApiError(
+          'image-download-address-failed',
+          '钉钉图片下载地址获取失败。',
+          { cause: error, status: error?.status, providerCode: error?.providerCode },
+        );
+      }
+      const downloadUrl = nonEmptyString(response?.downloadUrl ?? response?.download_url);
+      if (!downloadUrl) {
+        throw new DingtalkApiError('invalid-image-download', '钉钉服务没有返回图片下载地址。');
+      }
+      try {
+        return await fetchImageBuffer(secureDingtalkDownloadUrl(downloadUrl), {
+          fetchImpl,
+          signal,
+          maxBytes,
+        });
+      } catch (error) {
+        if (signal?.aborted || error instanceof ImagePromptError) throw error;
+        throw new DingtalkApiError(
+          'image-content-download-failed',
+          '钉钉图片内容下载失败。',
+          { cause: error },
+        );
+      }
+    },
+
     async createAiCard({ clientId, clientSecret, target, initialText, signal }) {
       const appKey = nonEmptyString(clientId);
       const appSecret = nonEmptyString(clientSecret);

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

@@ -12,6 +12,11 @@ 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 CARD_INITIAL_TEXT = '已连接 DeepSeek Harness，正在思考…';
 const CARD_ERROR_TEXT = '消息处理失败，请稍后重试。';
@@ -20,7 +25,7 @@ const INTERACTION_RESOLVED_TEXT = '这个问题已在其他客户端处理，无
 const HELP_TEXT = [
   '钉钉机器人已连接 DeepSeek Harness。',
   '',
-  '直接发送文字即可继续当前会话。',
+  '直接发送文字或图片即可继续当前会话。',
   '/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',
@@ -35,10 +40,123 @@ function nonEmptyString(value) {
   return typeof value === 'string' && value.trim() ? value.trim() : null;
 }
 
+function safeErrorDiagnostic(error) {
+  const chain = [];
+  const seen = new Set();
+  let current = error;
+  while (current && typeof current === 'object' && chain.length < 3 && !seen.has(current)) {
+    seen.add(current);
+    const name = nonEmptyString(current.name)?.slice(0, 80);
+    const code = nonEmptyString(current.code)?.slice(0, 80);
+    const providerCode = nonEmptyString(current.providerCode)?.slice(0, 160);
+    const status = Number.isInteger(current.status) ? current.status : undefined;
+    chain.push({
+      ...(name ? { name } : {}),
+      ...(code ? { code } : {}),
+      ...(providerCode ? { providerCode } : {}),
+      ...(status ? { status } : {}),
+    });
+    current = current.cause;
+  }
+  return chain;
+}
+
+function dingtalkImageErrorUserMessage(error) {
+  let current = error;
+  const seen = new Set();
+  while (current && typeof current === 'object' && !seen.has(current)) {
+    seen.add(current);
+    if (current.code === 'image-download-address-failed') {
+      return '钉钉未能换取图片下载地址，请重新发送；若持续失败，请检查机器人的“企业内机器人发送消息权限”。';
+    }
+    if (current.code === 'invalid-image-download') {
+      return '钉钉没有返回图片下载地址，请重新发送。';
+    }
+    if (current.code === 'image-content-download-failed') {
+      return '钉钉返回的图片临时地址无法读取，请重新发送。';
+    }
+    current = current.cause;
+  }
+  return imagePromptUserMessage(error);
+}
+
 function senderStaffId(message) {
   return nonEmptyString(message?.senderStaffId) ?? nonEmptyString(message?.senderId);
 }
 
+function parsedMessageContent(message) {
+  if (message?.content && typeof message.content === 'object') return message.content;
+  if (typeof message?.content !== 'string') return null;
+  try {
+    const parsed = JSON.parse(message.content);
+    return parsed && typeof parsed === 'object' ? parsed : null;
+  } catch {
+    return null;
+  }
+}
+
+function richTextEntries(content) {
+  const entries = content?.richText ?? content?.rich_text;
+  return Array.isArray(entries) ? entries : [];
+}
+
+function richTextEntryText(entry) {
+  if (typeof entry?.text === 'string') return nonEmptyString(entry.text);
+  if (typeof entry?.text?.content === 'string') return nonEmptyString(entry.text.content);
+  if (String(entry?.type).toLowerCase() === 'text' && typeof entry?.content === 'string') {
+    return nonEmptyString(entry.content);
+  }
+  return null;
+}
+
+function downloadCodeFor(value) {
+  return nonEmptyString(value?.downloadCode) ?? nonEmptyString(value?.pictureDownloadCode);
+}
+
+/** Normalize DingTalk picture and richText callbacks into lazy image references. */
+export function dingtalkInboundMessage(message, {
+  api,
+  clientId,
+  clientSecret,
+} = {}) {
+  const msgtype = String(message?.msgtype ?? '').toLowerCase();
+  const content = parsedMessageContent(message);
+  const richEntries = msgtype === 'richtext' ? richTextEntries(content) : [];
+  const text = msgtype === 'text'
+    ? nonEmptyString(message?.text?.content) ?? ''
+    : richEntries.map(richTextEntryText).filter(Boolean).join('\n');
+  const imageCodes = [];
+  if (msgtype === 'picture') {
+    const code = downloadCodeFor(content);
+    if (code) imageCodes.push(code);
+  } else if (msgtype === 'richtext') {
+    for (const entry of richEntries) {
+      if (String(entry?.type ?? '').toLowerCase() !== 'picture') continue;
+      const code = downloadCodeFor(entry);
+      if (code) imageCodes.push(code);
+    }
+  }
+  return {
+    content: text,
+    images: imageCodes.map((downloadCode, index) => ({
+      name: index === 0 ? 'image' : `image-${index + 1}`,
+      load: ({ signal, maxBytes }) => {
+        if (typeof api?.downloadImage !== 'function') {
+          throw new Error('DingTalk API does not support image downloads');
+        }
+        return api.downloadImage({
+          clientId,
+          clientSecret,
+          robotCode: message?.robotCode,
+          downloadCode,
+          signal,
+          maxBytes,
+        });
+      },
+    })),
+  };
+}
+
 function conversationKey(message, sender) {
   if (String(message?.conversationType) === '2') {
     const conversationId = nonEmptyString(message?.conversationId);
@@ -317,49 +435,63 @@ export class DingtalkHarnessBridge {
       return;
     }
 
-    const text = message?.msgtype === 'text' ? nonEmptyString(message?.text?.content) : null;
+    const promptMessage = dingtalkInboundMessage(message, {
+      api: this.#api,
+      clientId: this.#clientId,
+      clientSecret: this.#clientSecret,
+    });
+    const text = promptMessage.content;
+    const hasImages = hasInboundImages(promptMessage);
+    const isPlainText = String(message?.msgtype).toLowerCase() === 'text';
     let cardStream = null;
     let cardStarted = false;
     try {
-      if (!text) {
-        await this.#send(sessionWebhook, '目前仅支持文字消息。');
+      if (!text && !hasImages) {
+        await this.#send(sessionWebhook, '目前支持文字和图片消息。');
         return;
       }
 
       const command = text.toLowerCase();
-      if (command === '/help') {
+      if (isPlainText && !hasImages && command === '/help') {
         await this.#send(sessionWebhook, HELP_TEXT);
         return;
       }
-      if (command === '/status') {
+      if (isPlainText && !hasImages && command === '/status') {
         await this.#harness.ensureRunning({ signal: this.#signal });
         await this.#send(sessionWebhook, '钉钉机器人与 DeepSeek Harness 连接正常。');
         return;
       }
-      if (command === '/new') {
+      if (isPlainText && !hasImages && command === '/new') {
         await this.#state.clearSession(key);
         await this.#send(sessionWebhook, '已开启新会话。请发送你的问题。');
         return;
       }
-      const workspaceCommand = await runWorkspaceCommand(text, this.#harness, key);
+      const workspaceCommand = isPlainText && !hasImages
+        ? await runWorkspaceCommand(text, this.#harness, key)
+        : null;
       if (workspaceCommand) {
         for (const reply of workspaceCommand.messages ?? [workspaceCommand.message]) {
           await this.#send(sessionWebhook, reply);
         }
         return;
       }
-      const compactCommand = await runCompactCommand(
-        text,
-        this.#harness,
-        this.#state,
-        key,
-        { signal: this.#signal },
-      );
+      const compactCommand = isPlainText && !hasImages
+        ? await runCompactCommand(
+            text,
+            this.#harness,
+            this.#state,
+            key,
+            { signal: this.#signal },
+          )
+        : null;
       if (compactCommand) {
         await this.#send(sessionWebhook, compactCommand.message);
         return;
       }
 
+      const content = hasImages
+        ? await promptContentForMessage(promptMessage, { signal: this.#signal })
+        : undefined;
       if (typeof this.#api.createAiCard === 'function'
         && typeof this.#api.updateAiCard === 'function'
         && typeof this.#api.finishAiCard === 'function') {
@@ -377,7 +509,7 @@ export class DingtalkHarnessBridge {
         harness: this.#harness,
         state: this.#state,
         key,
-        text,
+        ...(hasImages ? { content } : { text }),
         createOptions: { signal: this.#signal },
         existsOptions: { signal: this.#signal },
         askOptions: {
@@ -400,13 +532,17 @@ export class DingtalkHarnessBridge {
       increment(this.#status, 'messagesReplied');
       this.#status.lastReplyAt = new Date().toISOString();
       this.#status.lastError = null;
-    } catch {
+    } catch (error) {
       if (this.#signal?.aborted) return;
       this.#status.lastError = '钉钉消息处理失败。';
-      this.#logger.error?.('[dsh-dingtalk] failed to process an inbound message');
+      this.#logger.error?.(
+        '[dsh-dingtalk] failed to process an inbound message',
+        safeErrorDiagnostic(error),
+      );
       try {
-        const streamed = cardStarted && await cardStream.finish(CARD_ERROR_TEXT);
-        if (!streamed) await this.#send(sessionWebhook, CARD_ERROR_TEXT);
+        const errorText = dingtalkImageErrorUserMessage(error) ?? CARD_ERROR_TEXT;
+        const streamed = cardStarted && await cardStream.finish(errorText);
+        if (!streamed) await this.#send(sessionWebhook, errorText);
       } catch {
         this.#logger.error?.('[dsh-dingtalk] failed to send the safe error reply');
       }

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

@@ -108,7 +108,7 @@ export class DiscordApi {
         headers: {
           authorization: `Bot ${this.#token}`,
           'content-type': 'application/json',
-          'user-agent': 'DeepSeek-Harness-dsh-im (https://github.com/xmanrui/dsh-im, 0.9.0)',
+          'user-agent': 'DeepSeek-Harness-dsh-im (https://github.com/xmanrui/dsh-im, 0.10.0)',
         },
         ...(body === undefined ? {} : { body: JSON.stringify(body) }),
         signal: requestSignal(signal, timeoutMs),

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

@@ -1,9 +1,19 @@
 import { createEditableMessageStream, splitMessageText } from '../shared/editable-message-stream.mjs';
+import { fetchImageBuffer } from '../shared/image-prompt.mjs';
 import { DiscordApi } from './discord-api.mjs';
 import { createDiscordBridgeStatus, DiscordHarnessBridge } from './discord-bridge.mjs';
 
 const DISCORD_GATEWAY_INTENTS = (1 << 0) | (1 << 9) | (1 << 12);
 const RECONNECT_DELAYS_MS = Object.freeze([1_000, 3_000, 5_000, 10_000, 30_000]);
+const IMAGE_MEDIA_TYPES = new Set(['image/jpeg', 'image/png', 'image/webp', 'image/gif']);
+const IMAGE_FILE_TYPES = new Map([
+  ['.jpg', 'image/jpeg'],
+  ['.jpeg', 'image/jpeg'],
+  ['.png', 'image/png'],
+  ['.webp', 'image/webp'],
+  ['.gif', 'image/gif'],
+]);
+const DISCORD_IMAGE_HOSTS = Object.freeze(['cdn.discordapp.com']);
 
 function socketUrl(value) {
   const url = new URL(value);
@@ -48,7 +58,37 @@ function stripBotMention(text, botId) {
   return text.replace(new RegExp(`<@!?${botId}>`, 'g'), '').trim();
 }
 
-export function normalizeDiscordMessage(message, botId) {
+function attachmentMediaType(attachment) {
+  const value = typeof attachment?.content_type === 'string'
+    ? attachment.content_type.split(';', 1)[0].trim().toLowerCase() : '';
+  if (IMAGE_MEDIA_TYPES.has(value)) return value;
+  const filename = typeof attachment?.filename === 'string' ? attachment.filename.toLowerCase() : '';
+  for (const [extension, mediaType] of IMAGE_FILE_TYPES) {
+    if (filename.endsWith(extension)) return mediaType;
+  }
+  return null;
+}
+
+function attachmentSize(value) {
+  return Number.isSafeInteger(value) && value >= 0 ? value : undefined;
+}
+
+function discordImageSource(attachment, fetchImpl) {
+  const mediaType = attachmentMediaType(attachment);
+  if (!mediaType || typeof attachment?.url !== 'string') return null;
+  return {
+    name: typeof attachment.filename === 'string' ? attachment.filename : undefined,
+    mediaType,
+    size: attachmentSize(attachment.size),
+    load: (options) => fetchImageBuffer(attachment.url, {
+      ...options,
+      fetchImpl,
+      allowedHosts: DISCORD_IMAGE_HOSTS,
+    }),
+  };
+}
+
+export function normalizeDiscordMessage(message, botId, { fetchImpl = fetch } = {}) {
   if (!message?.id || !message?.channel_id || !message?.author?.id) return null;
   const direct = !message.guild_id;
   const addressed = direct
@@ -60,6 +100,9 @@ export function normalizeDiscordMessage(message, botId) {
     kind: direct ? 'direct' : 'group',
     conversationId: String(message.channel_id),
     content: stripBotMention(message.content ?? '', botId),
+    images: Array.isArray(message.attachments)
+      ? message.attachments.map((attachment) => discordImageSource(attachment, fetchImpl)).filter(Boolean)
+      : [],
     addressed,
     replyTarget: {
       channelId: String(message.channel_id),

package/src/channels/feishu/bridge.mjs

@@ -1,10 +1,16 @@
 import {
   conversationKey,
+  extractInboundMessage,
   extractText,
   isAllowedSender,
   isBotSender,
   splitText,
 } from './message-utils.mjs';
+import {
+  hasInboundImages,
+  imagePromptUserMessage,
+  promptContentForMessage,
+} from '../shared/image-prompt.mjs';
 import {
   harnessAnswerForQuestion,
   harnessQuestionText,
@@ -21,7 +27,7 @@ const RESOLVED_REPLY_TTL_MS = 30 * 60_000;
 const HELP_TEXT = [
   '北汇星河 AIOS 已连接 DeepSeek Harness。',
   '',
-  '直接发送问题即可继续当前会话。',
+  '直接发送文字或图片即可继续当前会话。',
   '/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',
@@ -273,7 +279,8 @@ export class FeishuHarnessBridge {
     await this.#finishReaction(messageId, processingReaction, 'ERROR');
     await this.#send(
       event.message.chat_id,
-      '处理失败，请稍后重试。如果问题持续，请在 DeepSeek Harness 的飞书插件页面检查连接状态。',
+      imagePromptUserMessage(error)
+        ?? '处理失败，请稍后重试。如果问题持续，请在 DeepSeek Harness 的飞书插件页面检查连接状态。',
     ).catch(() => undefined);
   }
 
@@ -297,40 +304,47 @@ export class FeishuHarnessBridge {
       this.#status.messagesReceived += 1;
     }
 
-    const text = extractText(event);
-    if (!text) {
-      await this.#send(event.message.chat_id, '目前仅支持文字消息。');
+    const message = extractInboundMessage(event, this.#client);
+    const text = message.content;
+    const hasImages = hasInboundImages(message);
+    const commandText = event.message.message_type === 'text' && !hasImages ? text : null;
+    if (!text && !hasImages) {
+      await this.#send(event.message.chat_id, '目前支持文字和图片消息。');
       return;
     }
 
-    if (text === '/help') {
+    if (commandText === '/help') {
       await this.#send(event.message.chat_id, HELP_TEXT);
       return;
     }
-    if (text === '/new') {
+    if (commandText === '/new') {
       await this.#state.clearSession(key);
       await this.#send(event.message.chat_id, '已开启全新 Harness 会话。');
       return;
     }
-    if (text === '/status') {
+    if (commandText === '/status') {
       await this.#harness.ensureRunning({ signal: this.#signal });
       await this.#send(event.message.chat_id, '飞书机器人与 DeepSeek Harness 连接正常。');
       return;
     }
-    const workspaceCommand = await runWorkspaceCommand(text, this.#harness, key);
+    const workspaceCommand = commandText === null
+      ? null
+      : await runWorkspaceCommand(text, this.#harness, key);
     if (workspaceCommand) {
       for (const reply of workspaceCommand.messages ?? [workspaceCommand.message]) {
         await this.#send(event.message.chat_id, reply);
       }
       return;
     }
-    const compactCommand = await runCompactCommand(
-      text,
-      this.#harness,
-      this.#state,
-      key,
-      { signal: this.#signal },
-    );
+    const compactCommand = commandText === null
+      ? null
+      : await runCompactCommand(
+          commandText,
+          this.#harness,
+          this.#state,
+          key,
+          { signal: this.#signal },
+        );
     if (compactCommand) {
       await this.#send(event.message.chat_id, compactCommand.message);
       return;
@@ -338,7 +352,7 @@ export class FeishuHarnessBridge {
 
     this.#logger.info?.(`[dsh-feishu] processing ${event.message.chat_type} message ${messageId}`);
     try {
-      await this.#answerWithStream(event, key, text);
+      await this.#answerWithStream(event, key, message);
       this.#status.messagesReplied += 1;
       this.#status.lastReplyAt = new Date().toISOString();
       this.#status.lastError = null;
@@ -362,15 +376,20 @@ export class FeishuHarnessBridge {
     };
   }
 
-  async #answerWithStream(event, key, text) {
+  async #answerWithStream(event, key, message) {
     const chatId = event.message.chat_id;
     const messageId = event.message.message_id;
+    const text = message.content;
+    const content = hasInboundImages(message)
+      ? await promptContentForMessage(message, { signal: this.#signal })
+      : undefined;
     if (!this.#channel?.stream) {
       const { answer } = await askInWorkspaceSession({
         harness: this.#harness,
         state: this.#state,
         key,
         text,
+        content,
         createOptions: { signal: this.#signal },
         existsOptions: { signal: this.#signal },
         askOptions: this.#interactionAskOptions(event, key),
@@ -398,6 +417,7 @@ export class FeishuHarnessBridge {
             state: this.#state,
             key,
             text,
+            content,
             createOptions: { signal: this.#signal },
             existsOptions: { signal: this.#signal },
             askOptions,
@@ -425,6 +445,7 @@ export class FeishuHarnessBridge {
         state: this.#state,
         key,
         text,
+        content,
         createOptions: { signal: this.#signal },
         existsOptions: { signal: this.#signal },
         askOptions: this.#interactionAskOptions(event, key),