v2.1.1: Auto配置改为模型粒度拖动排序 + 历史缓存复用 + 对话生图

- Auto配置编辑: 粒度从提供商改为具体模型, 原生HTML5拖动排序, 修复已保存模型不显示勾选(字符串/对象混拼bug)
- 历史缓存: 带历史上下文的auto请求按首条消息识别会话, 优先复用上次模型命中上游前缀缓存
  (可配置: prefer_cache_model + cache_ttl_seconds, 系统配置页开关 + /api/admin/routing)
- 对话生图: chat页选 auto-image/Qwen-Image 直接调生图接口并在对话内展示图片
- 修复: Flask模板缓存导致改模板不生效 -> TEMPLATES_AUTO_RELOAD
- 修复: 自动回退/显式有序模型列表路由, 空列表回退提供商能力选择

data/config.json

@@ -131,8 +131,16 @@
   "auto_profiles": {
     "auto": {
       "name": "默认Auto",
-      "description": "文本推理 - 自动选择可用提供商",
+      "description": "文本推理 - 按模型列表顺序自动选择",
       "capability": "text",
+      "models": [
+        "unsloth/Qwen3.8-27B-Q6_K",
+        "unsloth/Qwen3.8-27B-Q4_K_M",
+        "deepseek-ai/DeepSeek-V4-Flash",
+        "meituan-longcat/LongCat-2.0",
+        "qwen3.6-plus",
+        "GLM-5.3-flash"
+      ],
       "providers": [
         "*"
       ],
@@ -140,8 +148,16 @@
     },
     "auto-text": {
       "name": "文本推理",
-      "description": "纯文本推理，自动选择文本模型",
+      "description": "纯文本推理，按模型列表顺序自动选择",
       "capability": "text",
+      "models": [
+        "unsloth/Qwen3.8-27B-Q6_K",
+        "unsloth/Qwen3.8-27B-Q4_K_M",
+        "deepseek-ai/DeepSeek-V4-Flash",
+        "meituan-longcat/LongCat-2.0",
+        "qwen3.6-plus",
+        "GLM-5.3-flash"
+      ],
       "providers": [
         "*"
       ],
@@ -149,8 +165,14 @@
     },
     "auto-vision": {
       "name": "视觉能力",
-      "description": "多模态视觉理解，自动选择视觉模型",
+      "description": "多模态视觉理解，按模型列表顺序自动选择",
       "capability": "vision",
+      "models": [
+        "unsloth/Qwen3.8-27B-Q6_K",
+        "unsloth/Qwen3.8-27B-Q4_K_M",
+        "qwen3.6-plus",
+        "GLM-5.3-flash"
+      ],
       "providers": [
         "*"
       ],
@@ -158,8 +180,11 @@
     },
     "auto-image": {
       "name": "图片生成",
-      "description": "文生图，自动选择生图模型",
+      "description": "文生图，按模型列表顺序自动选择",
       "capability": "image_gen",
+      "models": [
+        "Qwen-Image"
+      ],
       "providers": [
         "*"
       ],
@@ -167,8 +192,9 @@
     },
     "auto-voice-out": {
       "name": "语音输出",
-      "description": "语音合成(TTS)，自动选择语音输出模型",
+      "description": "语音合成(TTS)，按模型列表顺序自动选择",
       "capability": "audio_out",
+      "models": [],
       "providers": [
         "*"
       ],
@@ -176,8 +202,9 @@
     },
     "auto-voice-in": {
       "name": "语音输入",
-      "description": "语音识别(ASR)，自动选择语音输入模型",
+      "description": "语音识别(ASR)，按模型列表顺序自动选择",
       "capability": "audio_in",
+      "models": [],
       "providers": [
         "*"
       ],
@@ -185,12 +212,17 @@
     },
     "auto-video": {
       "name": "视频生成",
-      "description": "文生视频，自动选择视频生成模型",
+      "description": "文生视频，按模型列表顺序自动选择",
       "capability": "video_gen",
+      "models": [],
       "providers": [
         "*"
       ],
       "strategy": "priority"
     }
+  },
+  "routing_config": {
+    "prefer_cache_model": true,
+    "cache_ttl_seconds": 3600
   }
 }