feat: persist photos and enrich conversation history

backend/tests/test_user_input.py

@@ -32,7 +32,7 @@ class UserInputParserTests(unittest.TestCase):
         self.assertIsNotNone(value)
         self.assertEqual(value.text, "帮我看看")
         self.assertTrue(value.has_camera_frame)
-        self.assertEqual(value.transcript_text, "帮我看看\n已发送一张图片")
+        self.assertEqual(value.transcript_text, "帮我看看")
 
     def test_rejects_legacy_and_unsupported_image_sources(self):
         self.assertIsNone(parse_user_input({"type": "user-text", "text": "旧协议"}))
@@ -54,6 +54,25 @@ class UserInputParserTests(unittest.TestCase):
                 }
             )
 
+    def test_image_only_input_has_no_synthetic_chat_text(self):
+        value = parse_user_input(
+            {
+                "type": "user-input",
+                "schema_version": 1,
+                "input_id": "input_photo",
+                "parts": [
+                    {
+                        "type": "input_image",
+                        "source": {"type": "camera_frame", "frame": "current"},
+                    }
+                ],
+            }
+        )
+
+        self.assertIsNotNone(value)
+        self.assertEqual(value.transcript_text, "")
+        self.assertEqual(value.prompt_text, "请根据用户刚提交的图片进行回复。")
+
     def test_native_image_uses_the_standard_multimodal_user_turn_path(self):
         image = UserImageRawFrame(
             image=bytes([220, 40, 40] * 16 * 16),