feat: add WeChat media analysis adapters

wechat-mp.test.mjs

@@ -3686,6 +3686,7 @@ test('wechat mp service persists image and routes image url into agent prompt',
   const nonce = 'nonce';
   const testUserId = 'test-user-image';
   const prompts = [];
+  const metadataCalls = [];
   const detailCalls = [];
   const service = createWechatMpService({
     config: {
@@ -3747,6 +3748,7 @@ 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 +3811,170 @@ 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 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,
+    },
+    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 },
+    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 accepts full voice xml payload and routes recognition text into agent', async () => {
   const token = 'token';
   const timestamp = '1710000000';