refactor(core): 统一配置系统，移除 config.json

- 移除 config.json，所有配置统一从 agents.json 和 providers.json 读取
- config-loader.ts 从全局目录 ~/.ai-terminal-assistant/ 加载配置
- loadConfig() 从 agentRegistry.getGlobalConfig() 获取 defaults.model
- 添加 loadVisionConfig() 支持 Vision 模型配置
- Tavily API Key 仅从环境变量读取
- UI AgentDefaultsEditor 添加 Vision 模型配置界面
- 更新相关测试

packages/ui/src/api/types.ts

@@ -496,6 +496,8 @@ export interface AgentDefaults {
   maxSteps?: number;
   /** 模型配置 */
   model?: AgentModelConfig;
+  /** Vision 模型配置（用于图片理解） */
+  vision?: AgentModelConfig;
   /** 权限配置 */
   permission?: AgentPermission;
 }

packages/ui/src/components/AgentDefaultsEditor.tsx

@@ -44,6 +44,10 @@ export function AgentDefaultsEditor({
   const [temperature, setTemperature] = useState<number | undefined>(undefined);
   const [maxTokens, setMaxTokens] = useState<number | undefined>(undefined);
 
+  // Vision 模型配置
+  const [visionProvider, setVisionProvider] = useState<string>('');
+  const [visionModel, setVisionModel] = useState('');
+
   // UI 状态
   const [loading, setLoading] = useState(true);
   const [saving, setSaving] = useState(false);
@@ -66,6 +70,11 @@ export function AgentDefaultsEditor({
             setTemperature(defaults.model.temperature);
             setMaxTokens(defaults.model.maxTokens);
           }
+
+          if (defaults.vision) {
+            setVisionProvider(defaults.vision.provider || '');
+            setVisionModel(defaults.vision.model || '');
+          }
         }
       } catch (err) {
         setError(err instanceof Error ? err.message : 'Failed to load defaults');
@@ -103,6 +112,18 @@ export function AgentDefaultsEditor({
       defaults.model = model;
     }
 
+    // Vision 配置
+    const vision: AgentModelConfig = {};
+    if (visionProvider) {
+      vision.provider = visionProvider as 'anthropic' | 'deepseek' | 'openai';
+    }
+    if (visionModel) {
+      vision.model = visionModel;
+    }
+    if (Object.keys(vision).length > 0) {
+      defaults.vision = vision;
+    }
+
     return defaults;
   };
 
@@ -225,6 +246,9 @@ export function AgentDefaultsEditor({
                 {/* Model Configuration */}
                 <div className="space-y-4">
                   <h3 className="text-sm font-medium text-fg-secondary">Default Model</h3>
+                  <p className="text-xs text-fg-subtle -mt-2">
+                    The main model used for all conversations
+                  </p>
                   <div className="grid grid-cols-2 gap-4">
                     {/* Provider */}
                     <div>
@@ -287,6 +311,41 @@ export function AgentDefaultsEditor({
                   </div>
                 </div>
 
+                {/* Vision Model Configuration */}
+                <div className="space-y-4">
+                  <h3 className="text-sm font-medium text-fg-secondary">Vision Model</h3>
+                  <p className="text-xs text-fg-subtle -mt-2">
+                    Used for image understanding when the default model doesn't support vision
+                  </p>
+                  <div className="grid grid-cols-2 gap-4">
+                    {/* Vision Provider */}
+                    <div>
+                      <label className="block text-xs text-fg-muted mb-1">Provider</label>
+                      <select
+                        value={visionProvider}
+                        onChange={(e) => setVisionProvider(e.target.value)}
+                        className="w-full px-3 py-2 bg-surface-base border border-line rounded-lg text-sm focus:outline-none focus:border-primary-500"
+                      >
+                        <option value="">Not configured</option>
+                        <option value="anthropic">Anthropic</option>
+                        <option value="openai">OpenAI</option>
+                      </select>
+                    </div>
+
+                    {/* Vision Model */}
+                    <div>
+                      <label className="block text-xs text-fg-muted mb-1">Model</label>
+                      <input
+                        type="text"
+                        value={visionModel}
+                        onChange={(e) => setVisionModel(e.target.value)}
+                        placeholder="gpt-4o"
+                        className="w-full px-3 py-2 bg-surface-base border border-line rounded-lg text-sm focus:outline-none focus:border-primary-500"
+                      />
+                    </div>
+                  </div>
+                </div>
+
                 {/* Info */}
                 <div className="p-3 bg-blue-500/10 border border-blue-500/30 rounded-lg text-blue-400 text-xs">
                   <p>