fix(wechat): scrub image_url from sessions and harden vision handoff.

Strip image_url from persisted session payloads, keep image turns scoped to the active request, and align proxy/vision/page-data paths so WeChat image history does not leak into fresh sessions.

Co-authored-by: Cursor <cursoragent@cursor.com>

tkmind-proxy-vision.test.mjs

@@ -140,3 +140,35 @@ test('buildVisionPayload does not mark billable usage when vision analysis fails
   assert.equal(result?.billableImageCount, 0);
   assert.doesNotMatch(result?.userMessage?.content?.[0]?.text ?? '', /Qwen VL 图片描述/);
 });
+
+test('buildVisionPayload strips image_url content parts for text-only Goose providers', async () => {
+  const result = await buildVisionPayload({
+    userId: 'user-1',
+    publishLayout: { publicUrl: 'https://m.tkmind.cn/MindSpace/user-1' },
+    userMessage: {
+      content: [
+        { type: 'text', text: '这是什么' },
+        {
+          type: 'image_url',
+          image_url: { url: '/api/mindspace/v1/assets/asset-7/download?inline=1' },
+        },
+      ],
+      metadata: {
+        imageUrls: ['/api/mindspace/v1/assets/asset-7/download?inline=1'],
+      },
+    },
+    localFetchAsset: async () => ({
+      buffer: Buffer.from('fake-image'),
+      mimeType: 'image/png',
+    }),
+    llmProviderService: {
+      analyzeImagesWithVision: async () => '蓝色方块',
+    },
+  });
+
+  assert.equal(
+    (result?.userMessage?.content ?? []).some((item) => item?.type === 'image_url'),
+    false,
+  );
+  assert.match(result?.userMessage?.content?.[0]?.text ?? '', /蓝色方块/);
+});