merge: enable WeChat media for all users

Preserve Portal/Goose SSE cursor boundaries and recover Finish after reconnect.

server.mjs

@@ -820,6 +820,8 @@ async function bootstrapUserAuth() {
         const target = await tkmindProxy.resolveTarget(sessionId);
         return tkmindProxy.apiFetchTo(target, pathname, init);
       },
+      submitSessionReply: ({ userId, sessionId, requestId, userMessage }) =>
+        tkmindProxy.submitSessionReplyForUser(userId, sessionId, requestId, userMessage),
       scheduleService: process.env.H5_SCHEDULE_ENABLED === '1' ? scheduleService : null,
       wechatScheduleLlmConfigService,
       llmProviderService,

tkmind-proxy-vision.test.mjs

@@ -83,6 +83,42 @@ test('buildVisionPayload marks one billable image analysis when vision succeeds'
   );
 });
 
+test('buildVisionPayload sends all current-turn images to Qwen in order', async () => {
+  const imageUrls = [
+    'https://m.tkmind.cn/MindSpace/user-1/public/wechat-mp/first.jpg',
+    'https://m.tkmind.cn/MindSpace/user-1/public/wechat-mp/second.jpg',
+  ];
+  let analyzedImages = [];
+  const result = await buildVisionPayload({
+    userId: 'user-1',
+    publishLayout: { publicUrl: 'https://m.tkmind.cn/MindSpace/user-1' },
+    userMessage: {
+      content: [{ type: 'text', text: '请结合两张图片生成页面' }],
+      metadata: { imageUrls },
+    },
+    fetchImpl: async () =>
+      new Response(Buffer.from('fake-image'), {
+        status: 200,
+        headers: { 'Content-Type': 'image/jpeg' },
+      }),
+    llmProviderService: {
+      analyzeImagesWithVision: async (images) => {
+        analyzedImages = images;
+        return '第一张是宝塔，第二张是晚霞。';
+      },
+    },
+  });
+
+  assert.equal(analyzedImages.length, 2);
+  assert.deepEqual(analyzedImages.map((item) => item.rawUrl), imageUrls);
+  assert.deepEqual(result?.userMessage?.metadata?.imageUrls, [
+    '/MindSpace/user-1/public/wechat-mp/first.jpg',
+    '/MindSpace/user-1/public/wechat-mp/second.jpg',
+  ]);
+  assert.match(result?.userMessage?.content?.[0]?.text ?? '', /本轮用户仅上传 2 张图片/);
+  assert.match(result?.userMessage?.content?.[0]?.text ?? '', /图片2/);
+});
+
 test('buildVisionPayload does not mark billable usage when vision analysis fails', async () => {
   const result = await buildVisionPayload({
     userId: 'user-1',

tkmind-proxy.test.mjs

@@ -898,6 +898,67 @@ test('submitSessionReplyForUser adds goose metadata visibility flags before repl
   });
 });
 
+test('submitSessionReplyForUser applies the shared Qwen vision preprocessing path', async () => {
+  await withFakeGoosedSession(async ({ apiTarget, workingDir, replyBodies }) => {
+    const proxy = createTkmindProxy({
+      apiTarget,
+      apiSecret: 'test-secret',
+      userAuth: {
+        ...createMemoryTestUserAuth(workingDir),
+        async ownsSession() {
+          return true;
+        },
+        async canUseChat() {
+          return { ok: true };
+        },
+        async getUserById() {
+          return { id: 'user-1' };
+        },
+        async getUserPublishLayout() {
+          return { publicUrl: 'https://example.com/MindSpace/user-1' };
+        },
+        async resolveUserPolicies() {
+          return { unrestricted: true, policies: {} };
+        },
+      },
+      localFetchAsset: async () => ({
+        buffer: Buffer.from('fake-image'),
+        mimeType: 'image/jpeg',
+      }),
+      llmProviderService: {
+        async applyBestProviderForSession() {
+          return { ok: true };
+        },
+        async hasVisionKey() {
+          return true;
+        },
+        async analyzeImagesWithVision() {
+          return '一件蓝色产品，白色背景，竖版构图。';
+        },
+      },
+    });
+
+    await proxy.submitSessionReplyForUser(
+      'user-1',
+      'session-1',
+      'request-qwen-vision',
+      {
+        id: 'message-qwen-vision',
+        role: 'user',
+        content: [{ type: 'text', text: '请分析这张图片' }],
+        metadata: {
+          imageUrls: ['/api/mindspace/v1/assets/asset-1/download?inline=1'],
+          displayText: '请分析这张图片',
+        },
+      },
+    );
+
+    const forwardedText = replyBodies[0]?.user_message?.content?.[0]?.text ?? '';
+    assert.match(forwardedText, /Qwen VL 图片描述/);
+    assert.match(forwardedText, /一件蓝色产品/);
+  });
+});
+
 test('submitSessionReplyForUser passes current prompt to Memory V2 resolve before existing reply path', async () => {
   let resolveInput = null;
   await withFakeGoosedSession(async ({ apiTarget, workingDir }) => {

wechat-media.mjs

@@ -9,12 +9,19 @@ const __dirname = path.dirname(fileURLToPath(import.meta.url));
 
 const DEFAULT_WECHAT_MEDIA_URL = 'https://api.weixin.qq.com/cgi-bin/media/get';
 const DEFAULT_MAX_IMAGE_BYTES = 10 * 1024 * 1024;
+const DEFAULT_MAX_ATTACHMENT_BYTES = 30 * 1024 * 1024;
 const ALLOWED_IMAGE_MIME_TYPES = new Map([
   ['image/jpeg', 'jpg'],
   ['image/png', 'png'],
   ['image/webp', 'webp'],
   ['image/gif', 'gif'],
 ]);
+const ALLOWED_ATTACHMENT_EXTENSIONS = new Map([
+  ['.doc', 'application/msword'],
+  ['.docx', 'application/vnd.openxmlformats-officedocument.wordprocessingml.document'],
+  ['.xls', 'application/vnd.ms-excel'],
+  ['.xlsx', 'application/vnd.openxmlformats-officedocument.spreadsheetml.sheet'],
+]);
 
 function resolveImageExtension(contentType = '', fallbackUrl = '') {
   const normalized = String(contentType ?? '')
@@ -44,6 +51,32 @@ function ensureImageWithinLimit(buffer, maxBytes) {
   }
 }
 
+function sanitizeAttachmentFilename(filename = '') {
+  const basename = path.basename(String(filename ?? '').trim()).replace(/[\u0000-\u001f\u007f]/g, '');
+  if (!basename || basename === '.' || basename === '..') {
+    throw new Error('微信文件缺少有效文件名');
+  }
+  const extension = path.extname(basename).toLowerCase();
+  const mimeType = ALLOWED_ATTACHMENT_EXTENSIONS.get(extension);
+  if (!mimeType) {
+    throw new Error('当前服务号文件仅支持 Word（doc/docx）和 Excel（xls/xlsx）');
+  }
+  return {
+    filename: basename.slice(0, 160),
+    extension,
+    mimeType,
+  };
+}
+
+function ensureAttachmentWithinLimit(buffer, maxBytes) {
+  if (!Buffer.isBuffer(buffer) || buffer.length === 0) {
+    throw new Error('微信文件内容为空');
+  }
+  if (buffer.length > maxBytes) {
+    throw new Error(`文件超过大小限制（${maxBytes} bytes）`);
+  }
+}
+
 export async function downloadTemporaryMedia(accessToken, mediaId, { wechatFetch = undiciFetch } = {}) {
   if (!accessToken) throw new Error('缺少微信 access_token');
   if (!mediaId) throw new Error('缺少微信 mediaId');
@@ -154,3 +187,58 @@ export async function persistWechatImage(
     source,
   };
 }
+
+export async function persistWechatAttachment(
+  {
+    userId,
+    appId,
+    openid,
+    msgId,
+    mediaId,
+    filename,
+    publicBaseUrl,
+    maxFileBytes = DEFAULT_MAX_ATTACHMENT_BYTES,
+  },
+  {
+    wechatFetch = undiciFetch,
+    accessToken,
+    h5Root = __dirname,
+  } = {},
+) {
+  if (!userId) throw new Error('缺少 userId');
+  const resolved = sanitizeAttachmentFilename(filename);
+  const downloaded = await downloadTemporaryMedia(accessToken, mediaId, { wechatFetch });
+  ensureAttachmentWithinLimit(downloaded.buffer, maxFileBytes);
+
+  const publishDir = path.join(h5Root, PUBLISH_ROOT_DIR, String(userId), PUBLIC_ZONE_DIR, 'wechat-mp');
+  fs.mkdirSync(publishDir, { recursive: true });
+
+  const timestamp = Date.now();
+  const hash = crypto.createHash('sha1').update(downloaded.buffer).digest('hex').slice(0, 12);
+  const originalStem = path.basename(resolved.filename, resolved.extension)
+    .replace(/[^\p{L}\p{N}._-]+/gu, '_')
+    .replace(/^_+|_+$/g, '')
+    .slice(0, 80) || 'attachment';
+  const identity = [appId || 'wx', openid || 'openid', msgId || timestamp, mediaId || hash]
+    .filter(Boolean)
+    .join('-')
+    .replace(/[^a-zA-Z0-9._-]+/g, '_')
+    .slice(0, 100);
+  const publicFilename = `${originalStem}-${identity}-${hash}${resolved.extension}`;
+  const absolutePath = path.join(publishDir, publicFilename);
+  fs.writeFileSync(absolutePath, downloaded.buffer);
+
+  return {
+    absolutePath,
+    bytes: downloaded.buffer.length,
+    contentType: resolved.mimeType,
+    filename: resolved.filename,
+    publicFilename,
+    publicUrl: buildWechatImagePublicUrl({
+      publicBaseUrl,
+      publishKey: String(userId),
+      filename: publicFilename,
+    }),
+    source: 'wechat_media',
+  };
+}

wechat-mp-config.mjs

@@ -23,6 +23,13 @@ function deriveWechatEndpointFromUrl(baseUrl, suffix) {
   }
 }
 
+function parseCsvList(value = '') {
+  return String(value ?? '')
+    .split(',')
+    .map((item) => item.trim())
+    .filter(Boolean);
+}
+
 export function loadWechatMpConfig(env = process.env) {
   const appId = env.H5_WECHAT_MP_APP_ID?.trim() ?? env.H5_WECHAT_APP_ID?.trim() ?? '';
   const appSecret =
@@ -67,10 +74,13 @@ export function loadWechatMpConfig(env = process.env) {
     mediaPublicBaseUrl:
       env.H5_WECHAT_MP_MEDIA_PUBLIC_BASE_URL?.trim()?.replace(/\/$/, '') || publicBaseUrl,
     maxImageBytes: Math.max(1, Number(env.H5_WECHAT_MP_MAX_IMAGE_BYTES ?? 10 * 1024 * 1024)),
+    maxFileBytes: Math.max(1, Number(env.H5_WECHAT_MP_MAX_FILE_BYTES ?? 30 * 1024 * 1024)),
     acceptVoice: env.H5_WECHAT_MP_ACCEPT_VOICE !== '0',
     acceptImage: env.H5_WECHAT_MP_ACCEPT_IMAGE !== '0',
+    acceptFile: env.H5_WECHAT_MP_ACCEPT_FILE !== '0',
     acceptLocation: env.H5_WECHAT_MP_ACCEPT_LOCATION !== '0',
     acceptLink: env.H5_WECHAT_MP_ACCEPT_LINK !== '0',
+    mediaAnalysisGrayUsers: parseCsvList(env.H5_WECHAT_MP_MEDIA_GRAY_USERS),
     encodingAesKey: env.H5_WECHAT_MP_ENCODING_AES_KEY?.trim() ?? '',
   };
 }

wechat-mp.mjs

@@ -9,7 +9,11 @@ import { resolveSessionAccess } from './session-broker.mjs';
 import { isStubPublicHtmlContent, materializeMissingPublicHtmlWrites } from './mindspace-public-finish-sync.mjs';
 import { loadWechatMpConfig } from './wechat-mp-config.mjs';
 import { buildPublicUrl, PUBLISH_ROOT_DIR } from './user-publish.mjs';
-import { downloadTemporaryMedia, persistWechatImage } from './wechat-media.mjs';
+import {
+  downloadTemporaryMedia,
+  persistWechatAttachment,
+  persistWechatImage,
+} from './wechat-media.mjs';
 import { normalizeWechatName, resolveWechatAddressName } from './wechat/user/display-name.mjs';
 import { buildAckText } from './wechat/ack/ack-provider.mjs';
 import { guardScheduleConfirmationReply } from './wechat/handlers/schedule-guard.mjs';
@@ -46,6 +50,8 @@ const DEFAULT_WECHAT_CUSTOMER_SERVICE_URL =
   'https://api.weixin.qq.com/cgi-bin/message/custom/send';
 const DEFAULT_WECHAT_JSAPI_TICKET_URL = 'https://api.weixin.qq.com/cgi-bin/ticket/getticket';
 const DEFAULT_ASR_TARGET = process.env.H5_ASR_TARGET ?? 'https://asr.tkmind.cn';
+const WECHAT_RECENT_MEDIA_TTL_MS = 15 * 60 * 1000;
+const WECHAT_RECENT_IMAGE_MAX_COUNT = 10;
 export { loadWechatMpConfig };
 const PUBLIC_HTML_LINK_PATTERN =
   /https?:\/\/[^\s<>"')\]]+\/MindSpace\/([0-9a-f-]{36}|[a-z0-9._-]+)\/public\/([^\s<>"')\]]+\.html)/gi;
@@ -100,6 +106,8 @@ function parseWechatMessage(xml) {
     description: parseXmlField(xml, 'Description'),
     url: parseXmlField(xml, 'Url'),
     thumbMediaId: parseXmlField(xml, 'ThumbMediaId'),
+    fileName: parseXmlField(xml, 'FileName') || parseXmlField(xml, 'Filename'),
+    fileSize: parseXmlField(xml, 'FileSize'),
     event: parseXmlField(xml, 'Event').toLowerCase(),
     eventKey: parseXmlField(xml, 'EventKey'),
     latitude: parseXmlField(xml, 'Latitude'),
@@ -204,7 +212,14 @@ function pushMessage(messages, incoming) {
   return [...messages, incoming];
 }
 
-async function executeSessionReply(apiFetch, sessionId, requestId, prompt, metadata = {}) {
+async function executeSessionReply(
+  apiFetch,
+  sessionId,
+  requestId,
+  prompt,
+  metadata = {},
+  { submitReply = null } = {},
+) {
   const eventsResponse = await apiFetch(`/sessions/${sessionId}/events`, {
     method: 'GET',
     headers: { Accept: 'text/event-stream' },
@@ -214,18 +229,23 @@ async function executeSessionReply(apiFetch, sessionId, requestId, prompt, metad
     throw new Error(text || '无法建立公众号消息事件流');
   }
 
-  const replyResponse = await apiFetch(`/sessions/${sessionId}/reply`, {
-    method: 'POST',
-    body: JSON.stringify({
-      request_id: requestId,
-      user_message: createUserMessage(prompt, metadata),
-    }),
-  });
-  if (!replyResponse.ok) {
-    const text = await replyResponse.text().catch(() => '');
-    throw new Error(text || 'Agent reply 请求失败');
+  const userMessage = createUserMessage(prompt, metadata);
+  if (submitReply) {
+    await submitReply({ sessionId, requestId, userMessage });
+  } else {
+    const replyResponse = await apiFetch(`/sessions/${sessionId}/reply`, {
+      method: 'POST',
+      body: JSON.stringify({
+        request_id: requestId,
+        user_message: userMessage,
+      }),
+    });
+    if (!replyResponse.ok) {
+      const text = await replyResponse.text().catch(() => '');
+      throw new Error(text || 'Agent reply 请求失败');
+    }
+    replyResponse.body?.cancel().catch?.(() => {});
   }
-  replyResponse.body?.cancel().catch?.(() => {});
 
   const reader = eventsResponse.body.getReader();
   const decoder = new TextDecoder();
@@ -994,6 +1014,8 @@ export function isRecoverableWechatAgentSessionError(message) {
   if (/stale_session_poisoned_completion/i.test(normalized)) return true;
   if (/403|404|not found|无权访问/i.test(normalized)) return true;
   if (/tool_calls|tool_call_id|insufficient tool messages/i.test(normalized)) return true;
+  if (/session already has an active request|active request.*cancel/i.test(normalized)) return true;
+  if (/wechat_agent_incomplete_reply/i.test(normalized)) return true;
   return false;
 }
 
@@ -1024,6 +1046,13 @@ export function findRecoverableWechatAgentErrorInReply(reply) {
 export function assertWechatAgentReplyIsSendable(reply) {
   const recoverable = findRecoverableWechatAgentErrorInReply(reply);
   if (recoverable) throw new Error(recoverable);
+  const text = String(reply?.text ?? '').trim();
+  if (
+    /^(?:let me|i(?:'ll| will))\s+(?:first\s+)?(?:look|check|inspect|analy[sz]e)(?:\s+at)?\s+(?:the\s+)?image(?:\s+first)?[.!]?$/iu.test(text) ||
+    /^(?:让我|我先)(?:先)?(?:看|查看|检查|分析)(?:一下)?(?:这张|该张|这个)?图片[。！!]?$/u.test(text)
+  ) {
+    throw new Error('wechat_agent_incomplete_reply');
+  }
 }
 
 export function isWechatAgentApiErrorText(message) {
@@ -1164,6 +1193,24 @@ function normalizeNumber(value) {
   return Number.isFinite(num) ? num : null;
 }
 
+function isWechatMediaGrayUser(user, configuredUsers = []) {
+  const allowlist = Array.isArray(configuredUsers)
+    ? configuredUsers.map((value) => String(value ?? '').trim().toLowerCase()).filter(Boolean)
+    : [];
+  if (allowlist.length === 0) return false;
+  if (allowlist.includes('*')) return true;
+  const identities = [
+    user?.userId,
+    user?.username,
+    user?.slug,
+    user?.displayName,
+    user?.nickname,
+  ]
+    .map((value) => String(value ?? '').trim().toLowerCase())
+    .filter(Boolean);
+  return identities.some((identity) => allowlist.includes(identity));
+}
+
 function normalizeWechatInboundIntent(inbound) {
   const msgType = String(inbound?.msgType ?? '').toLowerCase();
   const base = {
@@ -1210,6 +1257,22 @@ function normalizeWechatInboundIntent(inbound) {
     };
   }
 
+  if (msgType === 'file') {
+    const filename = String(inbound?.fileName ?? '').trim();
+    return {
+      ...base,
+      displayText: filename ? `收到文件：${filename}` : '收到文件',
+      agentText: '',
+      media: {
+        mediaId: inbound?.mediaId || '',
+      },
+      attachment: {
+        filename,
+        sizeBytes: normalizeNumber(inbound?.fileSize),
+      },
+    };
+  }
+
   if (msgType === 'location') {
     const latitude = normalizeNumber(inbound?.locationX);
     const longitude = normalizeNumber(inbound?.locationY);
@@ -1409,6 +1472,7 @@ export function createWechatMpService({
   apiFetch,
   startAgentSession = null,
   sessionApiFetch = null,
+  submitSessionReply = null,
   scheduleService = null,
   wechatScheduleLlmConfigService = null,
   llmProviderService = null,
@@ -1434,10 +1498,15 @@ export function createWechatMpService({
     jsapiTicketUrl: config.jsapiTicketUrl || DEFAULT_WECHAT_JSAPI_TICKET_URL,
     mediaPublicBaseUrl: config.mediaPublicBaseUrl || config.publicBaseUrl,
     maxImageBytes: Math.max(1, Number(config.maxImageBytes ?? 10 * 1024 * 1024)),
+    maxFileBytes: Math.max(1, Number(config.maxFileBytes ?? 30 * 1024 * 1024)),
     acceptVoice: config.acceptVoice !== false,
     acceptImage: config.acceptImage !== false,
+    acceptFile: config.acceptFile !== false,
     acceptLocation: config.acceptLocation !== false,
     acceptLink: config.acceptLink !== false,
+    mediaAnalysisGrayUsers: Array.isArray(config.mediaAnalysisGrayUsers)
+      ? config.mediaAnalysisGrayUsers
+      : [],
     asrTarget: config.asrTarget || DEFAULT_ASR_TARGET,
   };
 
@@ -1446,6 +1515,8 @@ export function createWechatMpService({
     expiresAt: 0,
   };
   const rememberedWechatContexts = new Map();
+  const messageTasksByOpenid = new Map();
+  const recentMediaByOpenid = new Map();
   let jsapiTicketCache = {
     ticket: null,
     expiresAt: 0,
@@ -1454,6 +1525,86 @@ export function createWechatMpService({
   const fetchForSession = (sessionId, pathname, init) =>
     sessionApiFetch ? sessionApiFetch(sessionId, pathname, init) : apiFetch(pathname, init);
 
+  const enqueueMessageTask = (openid, taskFactory) => {
+    const key = String(openid ?? '').trim();
+    const previous = messageTasksByOpenid.get(key) ?? Promise.resolve();
+    const task = previous.catch(() => undefined).then(taskFactory);
+    messageTasksByOpenid.set(key, task);
+    void task
+      .finally(() => {
+        if (messageTasksByOpenid.get(key) === task) messageTasksByOpenid.delete(key);
+      })
+      .catch(() => {});
+    return task;
+  };
+
+  const rememberRecentMedia = (openid, intent) => {
+    const publicUrl = String(intent?.media?.publicUrl ?? '').trim();
+    if (!publicUrl) return;
+    const key = String(openid ?? '').trim();
+    const now = Date.now();
+    const item = {
+      media: { ...(intent.media ?? {}) },
+      attachment: intent.attachment ? { ...intent.attachment } : null,
+    };
+    const current = recentMediaByOpenid.get(key);
+    const canAppendImage =
+      intent?.msgType === 'image' &&
+      current &&
+      !current.claimed &&
+      now - current.rememberedAt <= WECHAT_RECENT_MEDIA_TTL_MS &&
+      current.items.every((recentItem) => !recentItem.attachment);
+    const candidates = canAppendImage ? [...current.items, item] : [item];
+    const deduped = candidates.filter(
+      (candidate, index, values) =>
+        values.findIndex(
+          (value) => String(value?.media?.publicUrl ?? '') === String(candidate?.media?.publicUrl ?? ''),
+        ) === index,
+    );
+    recentMediaByOpenid.set(key, {
+      items: deduped.slice(-WECHAT_RECENT_IMAGE_MAX_COUNT),
+      rememberedAt: now,
+      batchId: crypto.randomUUID(),
+      claimed: false,
+    });
+  };
+
+  const attachRecentMediaForFollowup = (openid, intent, mediaAnalysisEnabled) => {
+    if (!mediaAnalysisEnabled || intent?.msgType !== 'text' || intent?.media?.publicUrl) return;
+    const text = String(intent?.agentText ?? '');
+    if (!/(?:刚才|之前|上一|这张|这份|图片|图像|照片|文件|文档|表格|excel|word)/iu.test(text)) return;
+    const key = String(openid ?? '').trim();
+    const recent = recentMediaByOpenid.get(key);
+    if (
+      !recent ||
+      recent.claimed ||
+      Date.now() - recent.rememberedAt > WECHAT_RECENT_MEDIA_TTL_MS ||
+      recent.items.length === 0
+    ) {
+      return;
+    }
+    const items = recent.items.map((item) => ({
+      media: { ...(item.media ?? {}), source: 'wechat_recent_media' },
+      attachment: item.attachment ? { ...item.attachment } : null,
+    }));
+    const primary = items.at(-1);
+    intent.media = { ...(primary?.media ?? {}) };
+    if (primary?.attachment) intent.attachment = { ...primary.attachment };
+    intent.recentMediaItems = items;
+    intent.recentMediaBatchId = recent.batchId;
+    recent.claimed = true;
+  };
+
+  const settleRecentMediaBatch = (openid, intent, { succeeded }) => {
+    const batchId = String(intent?.recentMediaBatchId ?? '').trim();
+    if (!batchId) return;
+    const key = String(openid ?? '').trim();
+    const recent = recentMediaByOpenid.get(key);
+    if (!recent || recent.batchId !== batchId) return;
+    if (succeeded) recentMediaByOpenid.delete(key);
+    else recent.claimed = false;
+  };
+
   const resolveWechatBillingTokenState = (sessionId, tokenState) =>
     resolveBillingTokenState(tokenState, {
       sessionId,
@@ -1872,16 +2023,40 @@ export function createWechatMpService({
     }
   };
 
-  const buildIntentMetadata = (intent) => ({
-    source: 'wechat_mp',
-    msgType: intent.msgType,
-    originalMsgId: intent.msgId || null,
-    displayText: intent.displayText || '',
-    mediaPublicUrl: intent.media?.publicUrl || null,
-    recognition: intent.msgType === 'voice' ? intent.agentText || null : null,
-    location: intent.location || null,
-    link: intent.link || null,
-  });
+  const buildIntentMetadata = (intent, { mediaAnalysisEnabled = false } = {}) => {
+    const mediaPublicUrl = intent.media?.publicUrl || null;
+    const mediaItems = Array.isArray(intent.recentMediaItems) && intent.recentMediaItems.length > 0
+      ? intent.recentMediaItems
+      : mediaPublicUrl
+        ? [{ media: intent.media, attachment: intent.attachment ?? null }]
+        : [];
+    const imageUrls = mediaItems
+      .filter((item) => !item.attachment)
+      .map((item) => String(item?.media?.publicUrl ?? '').trim())
+      .filter((url, index, values) => url && values.indexOf(url) === index);
+    const fileAttachments = mediaItems
+      .filter((item) => item.attachment?.filename && item.media?.publicUrl)
+      .map((item) => ({
+        assetId: '',
+        downloadUrl: item.media.publicUrl,
+        filename: item.attachment.filename,
+        mimeType: item.attachment.mimeType || 'application/octet-stream',
+      }));
+    return {
+      source: 'wechat_mp',
+      msgType: intent.msgType,
+      originalMsgId: intent.msgId || null,
+      displayText: intent.displayText || '',
+      mediaPublicUrl,
+      ...(mediaAnalysisEnabled && imageUrls.length > 0
+        ? { imageUrls }
+        : {}),
+      ...(mediaAnalysisEnabled && fileAttachments.length > 0 ? { fileAttachments } : {}),
+      recognition: intent.msgType === 'voice' ? intent.agentText || null : null,
+      location: intent.location || null,
+      link: intent.link || null,
+    };
+  };
 
   const persistIntentDetail = async ({ intent, userId = null, rawXmlHash = '' }) => {
     if (typeof userAuth.insertWechatMpMessageDetail !== 'function') return;
@@ -1921,6 +2096,7 @@ export function createWechatMpService({
 
   const runIntentMessage = async ({ inbound, intent, user }) => {
     const wechatIntent = classifyWechatIntent(intent);
+    const mediaAnalysisEnabled = isWechatMediaGrayUser(user, config.mediaAnalysisGrayUsers);
     const resetCandidate =
       intent.msgType === 'text' || intent.msgType === 'voice' ? intent.agentText : '';
     // Page Data delivery owns persistent files, datasets and two publication
@@ -1970,7 +2146,18 @@ export function createWechatMpService({
         sessionId,
         requestId,
         agentPrompt,
-        buildIntentMetadata(intent),
+        buildIntentMetadata(intent, { mediaAnalysisEnabled }),
+        {
+          submitReply: submitSessionReply
+            ? ({ requestId: replyRequestId, userMessage }) =>
+                submitSessionReply({
+                  userId: user.userId,
+                  sessionId,
+                  requestId: replyRequestId,
+                  userMessage,
+                })
+            : null,
+        },
       );
       const workingDir = publishLayout?.publishDir ?? (await userAuth.resolveWorkingDir(user.userId));
       const linkExistsForRequest = resolveLinkExistsForWorkingDir(workingDir, linkExists);
@@ -2129,7 +2316,18 @@ export function createWechatMpService({
           sessionId,
           retryId,
           retryPrompt,
-          buildIntentMetadata(intent),
+          buildIntentMetadata(intent, { mediaAnalysisEnabled }),
+          {
+            submitReply: submitSessionReply
+              ? ({ requestId: replyRequestId, userMessage }) =>
+                  submitSessionReply({
+                    userId: user.userId,
+                    sessionId,
+                    requestId: replyRequestId,
+                    userMessage,
+                  })
+              : null,
+          },
         );
         const workingDir = publishLayout?.publishDir ?? (await userAuth.resolveWorkingDir(user.userId));
         const linkExistsForRequest = resolveLinkExistsForWorkingDir(workingDir, linkExists);
@@ -2302,6 +2500,7 @@ export function createWechatMpService({
     const supportedByConfig =
       (intent.msgType === 'voice' && config.acceptVoice) ||
       (intent.msgType === 'image' && config.acceptImage) ||
+      (intent.msgType === 'file' && config.acceptFile) ||
       (intent.msgType === 'location' && config.acceptLocation) ||
       (intent.msgType === 'link' && config.acceptLink) ||
       intent.msgType === 'text' ||
@@ -2350,6 +2549,26 @@ export function createWechatMpService({
       };
     }
 
+    const mediaAnalysisEnabled = isWechatMediaGrayUser(
+      boundUser,
+      config.mediaAnalysisGrayUsers,
+    );
+    attachRecentMediaForFollowup(inbound.fromUserName, intent, mediaAnalysisEnabled);
+
+    if (intent.msgType === 'file' && !mediaAnalysisEnabled) {
+      await persistIntentDetail({ intent, userId: boundUser.userId, rawXmlHash });
+      return {
+        ok: true,
+        status: 200,
+        contentType: 'application/xml; charset=utf-8',
+        body: buildWechatTextReply({
+          toUserName: inbound.fromUserName,
+          fromUserName: inbound.toUserName,
+          content: '当前账号尚未开启服务号文件分析灰度，请先通过 H5 上传文件。',
+        }),
+      };
+    }
+
     if (intent.msgType === 'voice' && !intent.agentText.trim() && intent.media?.mediaId) {
       try {
         const fallbackText = await transcribeWechatVoiceMedia(intent.media.mediaId, intent.media.format);
@@ -2418,6 +2637,7 @@ export function createWechatMpService({
           source: persisted.source,
         };
         intent.agentText = `[图片1]: ${persisted.publicUrl}`;
+        rememberRecentMedia(inbound.fromUserName, intent);
       } catch (error) {
         await persistIntentDetail({ intent, userId: boundUser.userId, rawXmlHash });
         return {
@@ -2433,6 +2653,57 @@ export function createWechatMpService({
       }
     }
 
+    if (intent.msgType === 'file') {
+      try {
+        const accessToken = await getStableAccessToken();
+        const persisted = await persistWechatAttachment(
+          {
+            userId: boundUser.userId,
+            appId: config.appId,
+            openid: inbound.fromUserName,
+            msgId: inbound.msgId,
+            mediaId: inbound.mediaId,
+            filename: inbound.fileName,
+            publicBaseUrl: config.mediaPublicBaseUrl,
+            maxFileBytes: config.maxFileBytes,
+          },
+          {
+            wechatFetch,
+            accessToken,
+          },
+        );
+        intent.media = {
+          ...intent.media,
+          mediaId: inbound.mediaId || intent.media?.mediaId || '',
+          publicUrl: persisted.publicUrl,
+          format: persisted.contentType,
+          source: persisted.source,
+        };
+        intent.attachment = {
+          ...intent.attachment,
+          filename: persisted.filename,
+          publicUrl: persisted.publicUrl,
+          mimeType: persisted.contentType,
+          sizeBytes: persisted.bytes,
+        };
+        intent.agentText = `[文件1: ${persisted.filename}]: ${persisted.publicUrl}`;
+        intent.displayText = `文件：${persisted.filename}`;
+        rememberRecentMedia(inbound.fromUserName, intent);
+      } catch (error) {
+        await persistIntentDetail({ intent, userId: boundUser.userId, rawXmlHash });
+        return {
+          ok: true,
+          status: 200,
+          contentType: 'application/xml; charset=utf-8',
+          body: buildWechatTextReply({
+            toUserName: inbound.fromUserName,
+            fromUserName: inbound.toUserName,
+            content: error instanceof Error ? error.message : '文件处理失败，请稍后重试。',
+          }),
+        };
+      }
+    }
+
     if (
       intent.msgType === 'location' &&
       (intent.location?.latitude == null || intent.location?.longitude == null)
@@ -2555,12 +2826,15 @@ export function createWechatMpService({
       };
     }
 
-    const task = runIntentMessage({
-      inbound,
-      intent,
-      user: boundUser,
-    })
+    const task = enqueueMessageTask(inbound.fromUserName, () =>
+      runIntentMessage({
+        inbound,
+        intent,
+        user: boundUser,
+      }),
+    )
       .then(async ({ sessionId } = {}) => {
+        settleRecentMediaBatch(inbound.fromUserName, intent, { succeeded: true });
         if (inbound.msgId && typeof userAuth.finishWechatMpMessage === 'function') {
           await userAuth.finishWechatMpMessage({
             appId: config.appId,
@@ -2572,6 +2846,7 @@ export function createWechatMpService({
         }
       })
       .catch(async (err) => {
+        settleRecentMediaBatch(inbound.fromUserName, intent, { succeeded: false });
         if (inbound.msgId && typeof userAuth.finishWechatMpMessage === 'function') {
           await userAuth.finishWechatMpMessage({
             appId: config.appId,

wechat-mp.test.mjs

@@ -79,6 +79,7 @@ function createBoundWechatService({
   config = {},
   scheduleService = null,
   applySessionLlmProvider = null,
+  submitSessionReply = null,
 }) {
   return createWechatMpService({
     config: {
@@ -128,6 +129,7 @@ function createBoundWechatService({
     },
     startAgentSession,
     sessionApiFetch,
+    submitSessionReply,
     scheduleService,
     applySessionLlmProvider,
     wechatFetch,
@@ -2712,6 +2714,10 @@ test('isRecoverableWechatAgentSessionError detects poisoned tool_calls history',
   );
   assert.equal(isRecoverableWechatAgentSessionError('stale_session_poisoned_completion'), true);
   assert.equal(isRecoverableWechatAgentSessionError('无权访问该会话'), true);
+  assert.equal(
+    isRecoverableWechatAgentSessionError('Session already has an active request. Cancel it first.'),
+    true,
+  );
   assert.equal(isRecoverableWechatAgentSessionError('network timeout'), false);
 });
 
@@ -2752,6 +2758,17 @@ test('sanitizeWechatAgentOutboundText replaces raw api errors with friendly text
   );
 });
 
+test('assertWechatAgentReplyIsSendable rejects image inspection placeholders', () => {
+  assert.throws(
+    () => assertWechatAgentReplyIsSendable({ text: 'Let me look at the image first.' }),
+    /wechat_agent_incomplete_reply/,
+  );
+  assert.throws(
+    () => assertWechatAgentReplyIsSendable({ text: '我先看一下这张图片。' }),
+    /wechat_agent_incomplete_reply/,
+  );
+});
+
 test('wechat mp service recreates dedicated session when tool_calls error arrives via Finish assistant text', async () => {
   const token = 'token';
   const timestamp = '1710000000';
@@ -3680,12 +3697,13 @@ test('wechat mp service falls back to ASR when voice recognition is empty', asyn
   }
 });
 
-test('wechat mp service persists image and routes image url into agent prompt', async () => {
+test('wechat mp wildcard media access persists image and routes image url into agent prompt', async () => {
   const token = 'token';
   const timestamp = '1710000000';
   const nonce = 'nonce';
   const testUserId = 'test-user-image';
   const prompts = [];
+  const metadataCalls = [];
   const detailCalls = [];
   const service = createWechatMpService({
     config: {
@@ -3701,6 +3719,7 @@ test('wechat mp service persists image and routes image url into agent prompt',
       unboundTextPrefix: '请先绑定',
       progressDelayMs: 0,
       maxImageBytes: 1024 * 1024,
+      mediaAnalysisGrayUsers: ['*'],
     },
     userAuth: {
       async findWechatUserByOpenid() {
@@ -3747,6 +3766,7 @@ test('wechat mp service persists image and routes image url into agent prompt',
       if (pathname === '/sessions/session-1/reply') {
         const body = JSON.parse(init.body);
         prompts.push(body.user_message.content[0].text);
+        metadataCalls.push(body.user_message.metadata);
         return new Response('{}', { status: 200, headers: { 'Content-Type': 'application/json' } });
       }
       if (pathname === '/agent/harness_remember' || pathname === '/agent/harness_bootstrap') {
@@ -3809,10 +3829,543 @@ test('wechat mp service persists image and routes image url into agent prompt',
     prompts[0],
     /\[图片1\]: https:\/\/example\.com\/MindSpace\/test-user-image\/public\/wechat-mp\//,
   );
+  assert.equal(metadataCalls.length, 1);
+  assert.equal(metadataCalls[0].source, 'wechat_mp');
+  assert.equal(metadataCalls[0].msgType, 'image');
+  assert.equal(metadataCalls[0].imageUrls.length, 1);
+  assert.match(metadataCalls[0].imageUrls[0], /\/public\/wechat-mp\//);
   assert.equal(detailCalls.length, 1);
   assert.match(detailCalls[0].mediaPublicUrl, /\/wechat-mp\//);
 });
 
+test('wechat mp image submission reuses the H5 prepared reply path', async () => {
+  const token = 'token';
+  const timestamp = '1710000000';
+  const nonce = 'nonce';
+  const testUserId = 'test-user-image-prepared';
+  const submitCalls = [];
+  const service = createBoundWechatService({
+    token,
+    config: {
+      mediaAnalysisGrayUsers: [testUserId],
+    },
+    userAuth: {
+      async findWechatUserByOpenid() {
+        return { userId: testUserId, status: 'active', nickname: '唐' };
+      },
+    },
+    sessionApiFetch: async (_sessionId, pathname) => {
+      if (pathname === '/sessions/session-1/events') {
+        return new Response(
+          [
+            'data: {"type":"Message","message":{"id":"assistant-1","role":"assistant","metadata":{"userVisible":true},"content":[{"type":"text","text":"图片内容已识别。"}]}}\n\n',
+            'data: {"type":"Finish"}\n\n',
+          ].join(''),
+          { status: 200, headers: { 'Content-Type': 'text/event-stream' } },
+        );
+      }
+      if (pathname === '/agent/harness_remember' || pathname === '/agent/harness_bootstrap') {
+        return new Response('{}', { status: 200, headers: { 'Content-Type': 'application/json' } });
+      }
+      throw new Error(`unexpected api path: ${pathname}`);
+    },
+    submitSessionReply: async (input) => {
+      submitCalls.push(input);
+      return { ok: true };
+    },
+    wechatFetch: async (url) => {
+      if (String(url).includes('/cgi-bin/stable_token')) {
+        return new Response(JSON.stringify({ access_token: 'access-1', expires_in: 7200 }), {
+          status: 200,
+          headers: { 'Content-Type': 'application/json' },
+        });
+      }
+      if (String(url).includes('/cgi-bin/media/get')) {
+        return new Response(Buffer.from([0x89, 0x50, 0x4e, 0x47]), {
+          status: 200,
+          headers: { 'Content-Type': 'image/png' },
+        });
+      }
+      if (String(url).includes('/cgi-bin/message/custom/send')) {
+        return new Response(JSON.stringify({ errcode: 0, errmsg: 'ok' }), {
+          status: 200,
+          headers: { 'Content-Type': 'application/json' },
+        });
+      }
+      throw new Error(`unexpected wechat url: ${url}`);
+    },
+  });
+
+  try {
+    const result = await service.handleInboundMessage(
+      inboundXml({
+        msgType: 'image',
+        content: '',
+        extraFields: { MediaId: 'media-prepared', PicUrl: 'https://wx.example.com/image.png' },
+      }),
+      { timestamp, nonce, signature: signatureFor(token, timestamp, nonce) },
+    );
+    await result.task;
+  } finally {
+    fs.rmSync(path.join(process.cwd(), 'MindSpace', testUserId), { recursive: true, force: true });
+  }
+
+  assert.equal(submitCalls.length, 1);
+  assert.equal(submitCalls[0].userId, testUserId);
+  assert.equal(submitCalls[0].sessionId, 'session-1');
+  assert.equal(submitCalls[0].userMessage.metadata.imageUrls.length, 1);
+  assert.match(submitCalls[0].userMessage.metadata.imageUrls[0], /\/public\/wechat-mp\//);
+});
+
+test('wechat mp serializes image and follow-up text and reattaches recent image', async () => {
+  const token = 'token';
+  const timestamp = '1710000000';
+  const nonce = 'nonce';
+  const testUserId = 'test-user-image-followup';
+  const submitCalls = [];
+  let eventCall = 0;
+  let releaseFirst = null;
+  const service = createBoundWechatService({
+    token,
+    config: {
+      mediaAnalysisGrayUsers: [testUserId],
+    },
+    userAuth: {
+      async findWechatUserByOpenid() {
+        return { userId: testUserId, status: 'active', nickname: '唐' };
+      },
+    },
+    sessionApiFetch: async (_sessionId, pathname) => {
+      if (pathname === '/sessions/session-1/events') {
+        eventCall += 1;
+        if (eventCall === 1) {
+          return new Response(
+            new ReadableStream({
+              start(controller) {
+                releaseFirst = () => {
+                  controller.enqueue(
+                    new TextEncoder().encode(
+                      'data: {"type":"Message","message":{"id":"assistant-image","role":"assistant","metadata":{"userVisible":true},"content":[{"type":"text","text":"图片已识别。"}]}}\n\n' +
+                        'data: {"type":"Finish"}\n\n',
+                    ),
+                  );
+                  controller.close();
+                };
+              },
+            }),
+            { status: 200, headers: { 'Content-Type': 'text/event-stream' } },
+          );
+        }
+        return new Response(
+          [
+            'data: {"type":"Message","message":{"id":"assistant-followup","role":"assistant","metadata":{"userVisible":true},"content":[{"type":"text","text":"已结合刚才图片分析主题。"}]}}\n\n',
+            'data: {"type":"Finish"}\n\n',
+          ].join(''),
+          { status: 200, headers: { 'Content-Type': 'text/event-stream' } },
+        );
+      }
+      if (pathname === '/agent/harness_remember' || pathname === '/agent/harness_bootstrap') {
+        return new Response('{}', { status: 200, headers: { 'Content-Type': 'application/json' } });
+      }
+      throw new Error(`unexpected api path: ${pathname}`);
+    },
+    submitSessionReply: async (input) => {
+      submitCalls.push(input);
+      return { ok: true };
+    },
+    wechatFetch: async (url) => {
+      if (String(url).includes('/cgi-bin/stable_token')) {
+        return new Response(JSON.stringify({ access_token: 'access-1', expires_in: 7200 }), {
+          status: 200,
+          headers: { 'Content-Type': 'application/json' },
+        });
+      }
+      if (String(url).includes('/cgi-bin/media/get')) {
+        return new Response(Buffer.from([0x89, 0x50, 0x4e, 0x47]), {
+          status: 200,
+          headers: { 'Content-Type': 'image/png' },
+        });
+      }
+      if (String(url).includes('/cgi-bin/message/custom/send')) {
+        return new Response(JSON.stringify({ errcode: 0, errmsg: 'ok' }), {
+          status: 200,
+          headers: { 'Content-Type': 'application/json' },
+        });
+      }
+      throw new Error(`unexpected wechat url: ${url}`);
+    },
+  });
+
+  try {
+    const imageResult = await service.handleInboundMessage(
+      inboundXml({
+        msgType: 'image',
+        content: '',
+        extraFields: { MediaId: 'media-followup', PicUrl: 'https://wx.example.com/image.png' },
+      }),
+      { timestamp, nonce, signature: signatureFor(token, timestamp, nonce) },
+    );
+    while (submitCalls.length === 0) await new Promise((resolve) => setTimeout(resolve, 0));
+
+    const followupResult = await service.handleInboundMessage(
+      inboundXml({ msgType: 'text', content: '请根据刚才图片分析主题' }),
+      { timestamp, nonce, signature: signatureFor(token, timestamp, nonce) },
+    );
+    await new Promise((resolve) => setTimeout(resolve, 0));
+    assert.equal(submitCalls.length, 1);
+
+    releaseFirst();
+    await imageResult.task;
+    await followupResult.task;
+  } finally {
+    fs.rmSync(path.join(process.cwd(), 'MindSpace', testUserId), { recursive: true, force: true });
+  }
+
+  assert.equal(submitCalls.length, 2);
+  assert.deepEqual(
+    submitCalls[1].userMessage.metadata.imageUrls,
+    submitCalls[0].userMessage.metadata.imageUrls,
+  );
+  assert.equal(submitCalls[1].userMessage.metadata.msgType, 'text');
+});
+
+test('wechat mp groups consecutive images for one follow-up and clears the consumed batch', async () => {
+  const token = 'token';
+  const timestamp = '1710000000';
+  const nonce = 'nonce';
+  const testUserId = 'test-user-multi-image-followup';
+  const submitCalls = [];
+  let eventCall = 0;
+  let releaseFirst = null;
+  const service = createBoundWechatService({
+    token,
+    config: {
+      mediaAnalysisGrayUsers: [testUserId],
+    },
+    userAuth: {
+      async findWechatUserByOpenid() {
+        return { userId: testUserId, status: 'active', nickname: '唐' };
+      },
+    },
+    sessionApiFetch: async (_sessionId, pathname) => {
+      if (pathname === '/sessions/session-1/events') {
+        eventCall += 1;
+        if (eventCall === 1) {
+          return new Response(
+            new ReadableStream({
+              start(controller) {
+                releaseFirst = () => {
+                  controller.enqueue(
+                    new TextEncoder().encode(
+                      'data: {"type":"Message","message":{"id":"assistant-first","role":"assistant","metadata":{"userVisible":true},"content":[{"type":"text","text":"第一张处理完成。"}]}}\n\n' +
+                        'data: {"type":"Finish"}\n\n',
+                    ),
+                  );
+                  controller.close();
+                };
+              },
+            }),
+            { status: 200, headers: { 'Content-Type': 'text/event-stream' } },
+          );
+        }
+        return new Response(
+          [
+            'data: {"type":"Message","message":{"id":"assistant-ok","role":"assistant","metadata":{"userVisible":true},"content":[{"type":"text","text":"处理完成。"}]}}\n\n',
+            'data: {"type":"Finish"}\n\n',
+          ].join(''),
+          { status: 200, headers: { 'Content-Type': 'text/event-stream' } },
+        );
+      }
+      if (pathname === '/agent/harness_remember' || pathname === '/agent/harness_bootstrap') {
+        return new Response('{}', { status: 200, headers: { 'Content-Type': 'application/json' } });
+      }
+      throw new Error(`unexpected api path: ${pathname}`);
+    },
+    submitSessionReply: async (input) => {
+      submitCalls.push(input);
+      return { ok: true };
+    },
+    wechatFetch: async (url) => {
+      if (String(url).includes('/cgi-bin/stable_token')) {
+        return new Response(JSON.stringify({ access_token: 'access-1', expires_in: 7200 }), {
+          status: 200,
+          headers: { 'Content-Type': 'application/json' },
+        });
+      }
+      if (String(url).includes('/cgi-bin/media/get')) {
+        return new Response(Buffer.from([0x89, 0x50, 0x4e, 0x47]), {
+          status: 200,
+          headers: { 'Content-Type': 'image/png' },
+        });
+      }
+      if (String(url).includes('/cgi-bin/message/custom/send')) {
+        return new Response(JSON.stringify({ errcode: 0, errmsg: 'ok' }), {
+          status: 200,
+          headers: { 'Content-Type': 'application/json' },
+        });
+      }
+      throw new Error(`unexpected wechat url: ${url}`);
+    },
+  });
+
+  try {
+    const firstImageResult = await service.handleInboundMessage(
+      inboundXml({
+        msgType: 'image',
+        content: '',
+        extraFields: {
+          MsgId: '10011',
+          MediaId: 'media-first',
+          PicUrl: 'https://wx.example.com/media-first.png',
+        },
+      }),
+      { timestamp, nonce, signature: signatureFor(token, timestamp, nonce) },
+    );
+    while (submitCalls.length === 0) await new Promise((resolve) => setTimeout(resolve, 0));
+
+    const secondImageResult = await service.handleInboundMessage(
+      inboundXml({
+        msgType: 'image',
+        content: '',
+        extraFields: {
+          MsgId: '10012',
+          MediaId: 'media-second',
+          PicUrl: 'https://wx.example.com/media-second.png',
+        },
+      }),
+      { timestamp, nonce, signature: signatureFor(token, timestamp, nonce) },
+    );
+
+    const followupResult = await service.handleInboundMessage(
+      inboundXml({
+        content: '请结合刚才两张图片分析主题',
+        extraFields: { MsgId: '10013' },
+      }),
+      { timestamp, nonce, signature: signatureFor(token, timestamp, nonce) },
+    );
+    await new Promise((resolve) => setTimeout(resolve, 0));
+    assert.equal(submitCalls.length, 1);
+
+    releaseFirst();
+    await firstImageResult.task;
+    await secondImageResult.task;
+    await followupResult.task;
+
+    const laterResult = await service.handleInboundMessage(
+      inboundXml({
+        content: '请再次分析刚才图片',
+        extraFields: { MsgId: '10014' },
+      }),
+      { timestamp, nonce, signature: signatureFor(token, timestamp, nonce) },
+    );
+    await laterResult.task;
+  } finally {
+    fs.rmSync(path.join(process.cwd(), 'MindSpace', testUserId), { recursive: true, force: true });
+  }
+
+  assert.equal(submitCalls.length, 4);
+  assert.equal(submitCalls[0].userMessage.metadata.imageUrls.length, 1);
+  assert.equal(submitCalls[1].userMessage.metadata.imageUrls.length, 1);
+  assert.equal(submitCalls[2].userMessage.metadata.imageUrls.length, 2);
+  assert.match(submitCalls[2].userMessage.metadata.imageUrls[0], /media-first/);
+  assert.match(submitCalls[2].userMessage.metadata.imageUrls[1], /media-second/);
+  assert.equal(submitCalls[2].userMessage.metadata.msgType, 'text');
+  assert.equal(submitCalls[3].userMessage.metadata.imageUrls, undefined);
+});
+
+test('wechat mp service persists Word and Excel files in user public area and reuses H5 attachment metadata', async () => {
+  const token = 'token';
+  const timestamp = '1710000000';
+  const nonce = 'nonce';
+  const testUserId = 'test-user-wechat-file';
+  const userMessages = [];
+  const service = createBoundWechatService({
+    token,
+    config: {
+      maxFileBytes: 1024 * 1024,
+      acceptFile: true,
+      mediaAnalysisGrayUsers: [testUserId],
+    },
+    userAuth: {
+      async findWechatUserByOpenid() {
+        return { userId: testUserId, status: 'active', nickname: '毕升' };
+      },
+    },
+    sessionApiFetch: async (_sessionId, pathname, init = {}) => {
+      if (pathname === '/sessions/session-1/events') {
+        return new Response(
+          [
+            'data: {"type":"Message","request_id":"req-file","message":{"id":"assistant-1","role":"assistant","metadata":{"userVisible":true},"content":[{"type":"text","text":"文件已解析。"}]}}\n\n',
+            'data: {"type":"Finish","request_id":"req-file","token_state":{"inputTokens":1,"outputTokens":2}}\n\n',
+          ].join(''),
+          { status: 200, headers: { 'Content-Type': 'text/event-stream' } },
+        );
+      }
+      if (pathname === '/sessions/session-1/reply') {
+        const body = JSON.parse(init.body);
+        userMessages.push(body.user_message);
+        return new Response('{}', { status: 200, headers: { 'Content-Type': 'application/json' } });
+      }
+      if (pathname === '/agent/harness_remember' || pathname === '/agent/harness_bootstrap') {
+        return new Response('{}', { status: 200, headers: { 'Content-Type': 'application/json' } });
+      }
+      throw new Error(`unexpected api path: ${pathname}`);
+    },
+    wechatFetch: async (url) => {
+      if (String(url).includes('/cgi-bin/stable_token')) {
+        return new Response(JSON.stringify({ access_token: 'access-1', expires_in: 7200 }), {
+          status: 200,
+          headers: { 'Content-Type': 'application/json' },
+        });
+      }
+      if (String(url).includes('/cgi-bin/media/get')) {
+        return new Response(Buffer.from('fake-docx-content'), {
+          status: 200,
+          headers: { 'Content-Type': 'application/octet-stream' },
+        });
+      }
+      if (String(url).includes('/cgi-bin/message/custom/send')) {
+        return new Response(JSON.stringify({ errcode: 0, errmsg: 'ok' }), {
+          status: 200,
+          headers: { 'Content-Type': 'application/json' },
+        });
+      }
+      throw new Error(`unexpected wechat url: ${url}`);
+    },
+  });
+
+  const originalRandomUuid = crypto.randomUUID;
+  crypto.randomUUID = () => 'req-file';
+  try {
+    for (const [index, filename] of ['季度分析.docx', '销售数据.xlsx'].entries()) {
+      const result = await service.handleInboundMessage(
+        inboundXml({
+          msgType: 'file',
+          content: '',
+          extraFields: {
+            MediaId: `file-media-${index + 1}`,
+            FileName: filename,
+            FileSize: 17,
+          },
+        }),
+        { timestamp, nonce, signature: signatureFor(token, timestamp, nonce) },
+      );
+      assert.equal(result.status, 200);
+      await result.task;
+      const attachment = userMessages.at(-1)?.metadata?.fileAttachments?.[0];
+      const publicFilename = decodeURIComponent(new URL(attachment.downloadUrl).pathname.split('/').at(-1));
+      assert.equal(
+        fs.existsSync(
+          path.join(process.cwd(), 'MindSpace', testUserId, 'public', 'wechat-mp', publicFilename),
+        ),
+        true,
+      );
+    }
+  } finally {
+    crypto.randomUUID = originalRandomUuid;
+    fs.rmSync(path.join(process.cwd(), 'MindSpace', testUserId), {
+      recursive: true,
+      force: true,
+    });
+  }
+
+  assert.equal(userMessages.length, 2);
+  assert.match(userMessages[0].content[0].text, /【微信服务号文件消息】/);
+  assert.match(userMessages[0].content[0].text, /\[文件1: 季度分析\.docx\]:/);
+  assert.equal(userMessages[0].metadata.msgType, 'file');
+  assert.equal(userMessages[0].metadata.fileAttachments.length, 1);
+  assert.equal(userMessages[0].metadata.fileAttachments[0].filename, '季度分析.docx');
+  assert.equal(
+    userMessages[0].metadata.fileAttachments[0].mimeType,
+    'application/vnd.openxmlformats-officedocument.wordprocessingml.document',
+  );
+  assert.match(userMessages[0].metadata.fileAttachments[0].downloadUrl, /\/public\/wechat-mp\//);
+  assert.match(userMessages[1].content[0].text, /\[文件1: 销售数据\.xlsx\]:/);
+  assert.equal(userMessages[1].metadata.fileAttachments[0].filename, '销售数据.xlsx');
+  assert.equal(
+    userMessages[1].metadata.fileAttachments[0].mimeType,
+    'application/vnd.openxmlformats-officedocument.spreadsheetml.sheet',
+  );
+  assert.match(userMessages[1].metadata.fileAttachments[0].downloadUrl, /\/public\/wechat-mp\//);
+});
+
+test('wechat mp service rejects unsupported public file types without entering the agent flow', async () => {
+  let mediaDownloads = 0;
+  let sessionCalls = 0;
+  const service = createBoundWechatService({
+    config: { acceptFile: true, mediaAnalysisGrayUsers: ['user-1'] },
+    sessionApiFetch: async () => {
+      sessionCalls += 1;
+      throw new Error('session should not be called');
+    },
+    wechatFetch: async (url) => {
+      if (String(url).includes('/cgi-bin/stable_token')) {
+        return new Response(JSON.stringify({ access_token: 'access-1', expires_in: 7200 }), {
+          status: 200,
+          headers: { 'Content-Type': 'application/json' },
+        });
+      }
+      if (String(url).includes('/cgi-bin/media/get')) mediaDownloads += 1;
+      throw new Error(`unexpected wechat url: ${url}`);
+    },
+  });
+
+  const result = await service.handleInboundMessage(
+    inboundXml({
+      msgType: 'file',
+      content: '',
+      extraFields: { MediaId: 'file-media-2', FileName: 'payload.exe' },
+    }),
+    {
+      timestamp: '1710000000',
+      nonce: 'nonce',
+      signature: signatureFor('token', '1710000000', 'nonce'),
+    },
+  );
+
+  assert.equal(result.status, 200);
+  assert.match(result.body, /仅支持 Word/);
+  assert.equal(mediaDownloads, 0);
+  assert.equal(sessionCalls, 0);
+});
+
+test('wechat mp service keeps file analysis disabled outside the media gray allowlist', async () => {
+  let mediaDownloads = 0;
+  let sessionCalls = 0;
+  const service = createBoundWechatService({
+    config: {
+      acceptFile: true,
+      mediaAnalysisGrayUsers: ['唐'],
+    },
+    sessionApiFetch: async () => {
+      sessionCalls += 1;
+      throw new Error('session should not be called');
+    },
+    wechatFetch: async (url) => {
+      if (String(url).includes('/cgi-bin/media/get')) mediaDownloads += 1;
+      throw new Error(`unexpected wechat url: ${url}`);
+    },
+  });
+
+  const result = await service.handleInboundMessage(
+    inboundXml({
+      msgType: 'file',
+      content: '',
+      extraFields: { MediaId: 'file-media-gray', FileName: '测试.docx' },
+    }),
+    {
+      timestamp: '1710000000',
+      nonce: 'nonce',
+      signature: signatureFor('token', '1710000000', 'nonce'),
+    },
+  );
+
+  assert.equal(result.status, 200);
+  assert.match(result.body, /尚未开启服务号文件分析灰度/);
+  assert.equal(mediaDownloads, 0);
+  assert.equal(sessionCalls, 0);
+});
+
 test('wechat mp service accepts full voice xml payload and routes recognition text into agent', async () => {
   const token = 'token';
   const timestamp = '1710000000';

wechat/ack/ack-builders.mjs

@@ -21,6 +21,12 @@ const ImageBuilder = {
   build: (ctx) => buildText(TEMPLATES.image, ctx),
 };
 
+const FileBuilder = {
+  priority: 80,
+  support: (ctx) => ctx.msgType === 'file',
+  build: (ctx) => buildText(TEMPLATES.file, ctx),
+};
+
 const VoiceBuilder = {
   priority: 80,
   support: (ctx) => ctx.msgType === 'voice',
@@ -54,7 +60,7 @@ const DefaultBuilder = {
   build: (ctx) => buildText(TEMPLATES.default, ctx),
 };
 
-const BUILDERS = [ImageBuilder, VoiceBuilder, LocationBuilder, LinkBuilder, IntentBuilder, DefaultBuilder]
+const BUILDERS = [ImageBuilder, FileBuilder, VoiceBuilder, LocationBuilder, LinkBuilder, IntentBuilder, DefaultBuilder]
   .sort((a, b) => b.priority - a.priority);
 
 export function selectBuilder(ctx) {

wechat/ack/ack-intent.mjs

@@ -13,6 +13,7 @@ const INTENT_RULES = [
 
 export function resolveIntent(msgType, text) {
   if (msgType === 'image') return { task: 'image_analysis' };
+  if (msgType === 'file') return { task: 'file_analysis' };
   if (msgType === 'voice') return { task: 'voice_analysis' };
   if (msgType === 'location') return { task: 'location' };
   if (msgType === 'link') return { task: 'link' };

wechat/ack/ack-provider.test.mjs

@@ -26,6 +26,11 @@ describe('buildAckText', () => {
     assert.equal(result, '图片收到，我先看看。');
   });
 
+  it('file → file template', () => {
+    const result = buildAckText({ intent: intent('file'), nickname: '', config: cfg, fallbackText: '' });
+    assert.equal(result, '文件收到，我先读取分析。');
+  });
+
   it('voice → voice template', () => {
     const result = buildAckText({ intent: intent('voice'), nickname: '', config: cfg, fallbackText: '' });
     assert.equal(result, '收到语音，我先听一下。');

wechat/ack/ack-templates.mjs

@@ -12,6 +12,7 @@ export const TEMPLATES = {
     '让我看看这张图片。',
     '图片已收到，我来处理。',
   ],
+  file: ['文件收到，我先读取分析。'],
   voice: [
     '收到语音，我先听一下。',
     '听到啦，我马上处理。',

wechat/prompts/chat-general.mjs

@@ -81,6 +81,20 @@ export function buildWechatAgentPrompt(intent, { grantedSkills = [] } = {}) {
       .filter(Boolean)
       .join('\n');
   }
+  if (msgType === 'file') {
+    const attachment = intent?.attachment ?? {};
+    return [
+      currentTimeHint,
+      '【微信服务号文件消息】用户发送了需要解析的 Office 文件。',
+      attachment.filename ? `文件名：${attachment.filename}` : '',
+      attachment.publicUrl ? `文件链接：${attachment.publicUrl}` : '',
+      '请复用 H5 附件分析结果回答；Excel 在专用工具可用时必须优先使用 Excel 工具读取完整工作簿。',
+      '',
+      String(agentText).trim(),
+    ]
+      .filter(Boolean)
+      .join('\n');
+  }
   if (msgType === 'location') {
     const location = intent?.location ?? {};
     return [