fix:完善Maisaka记忆写回链路

补齐聊天摘要自动写回、发送后同步与图存储清理逻辑，对齐 visual 新配置字段并补充相关回归测试，同时忽略 algorithm_redesign 设计目录。

src/config/legacy_migration.py

@@ -282,6 +282,10 @@ def try_migrate_legacy_bot_config_dict(data: dict[str, Any]) -> MigrationResult:
 
     personality = _as_dict(data.get("personality"))
     visual = _as_dict(data.get("visual"))
+    if visual is None and personality is not None and "visual_style" in personality:
+        visual = {}
+        data["visual"] = visual
+
     if visual is not None and personality is not None and "visual_style" in personality:
         if "visual_style" not in visual:
             visual["visual_style"] = personality["visual_style"]
@@ -289,15 +293,6 @@ def try_migrate_legacy_bot_config_dict(data: dict[str, Any]) -> MigrationResult:
         migrated_any = True
         reasons.append("personality.visual_style_moved_to_visual.visual_style")
 
-    if visual is not None and "multimodal_planner" in visual and "planner_mode" not in visual:
-        multimodal_planner = visual.pop("multimodal_planner")
-        if isinstance(multimodal_planner, bool):
-            visual["planner_mode"] = "multimodal" if multimodal_planner else "text"
-            migrated_any = True
-            reasons.append("visual.multimodal_planner_moved_to_visual.planner_mode")
-        else:
-            visual["multimodal_planner"] = multimodal_planner
-
     memory = _as_dict(data.get("memory"))
     if memory is not None and _migrate_target_item_list(memory, "global_memory_blacklist"):
         migrated_any = True

src/config/official_configs.py

@@ -145,23 +145,23 @@ class VisualConfig(ConfigBase):
     __ui_label__ = "视觉"
     __ui_icon__ = "image"
 
-    multimodal_planner: bool = Field(
-        default=True,
+    planner_mode: Literal["text", "multimodal", "auto"] = Field(
+        default="text",
         json_schema_extra={
-            "x-widget": "switch",
+            "x-widget": "select",
             "x-icon": "image",
         },
     )
-    """是否直接输入图片"""
+    """Planner 视觉模式：text 仅文本，multimodal 强制多模态，auto 按模型能力自动选择"""
 
-    multimodal_replyer: bool = Field(
-        default=False,
+    replyer_mode: Literal["text", "multimodal", "auto"] = Field(
+        default="text",
         json_schema_extra={
-            "x-widget": "switch",
+            "x-widget": "select",
             "x-icon": "git-branch",
         },
     )
-    """是否启用 Maisaka 多模态 replyer 生成器"""
+    """Replyer 视觉模式：text 仅文本，multimodal 强制多模态，auto 按模型能力自动选择"""
 
     visual_style: str = Field(
         default="请用中文描述这张图片的内容。如果有文字，请把文字描述概括出来，请留意其主题，直观感受，输出为一段平文本，最多30字，请注意不要分点，就输出一段文本",
@@ -424,6 +424,36 @@ class MemoryConfig(ConfigBase):
     )
     """是否在发送回复后自动提取并写回人物事实到长期记忆"""
 
+    chat_summary_writeback_enabled: bool = Field(
+        default=True,
+        json_schema_extra={
+            "x-widget": "switch",
+            "x-icon": "scroll-text",
+        },
+    )
+    """是否在 Maisaka 聊天过程中按消息窗口自动写回聊天摘要到长期记忆"""
+
+    chat_summary_writeback_message_threshold: int = Field(
+        default=12,
+        ge=1,
+        json_schema_extra={
+            "x-widget": "input",
+            "x-icon": "messages-square",
+        },
+    )
+    """自动写回聊天摘要的消息窗口阈值"""
+
+    chat_summary_writeback_context_length: int = Field(
+        default=50,
+        ge=1,
+        le=500,
+        json_schema_extra={
+            "x-widget": "input",
+            "x-icon": "rows-3",
+        },
+    )
+    """自动写回聊天摘要时，从聊天流中回看的消息条数"""
+
     feedback_correction_enabled: bool = Field(
         default=False,
         json_schema_extra={