feat: 问诊对话+建档引导+报告+日历+趋势页重写+视觉统一

- 问诊对话页: 新建consultation_provider, 完整AI分身聊天UI
- 首次建档引导: OnboardingPrompt+ManageArchiveTool+前端卡片
- 报告模块: POST端点 + report_agent_handler接VLM
- 健康日历: GET /api/calendar端点 + 前端接API
- 趋势页重写: 删除Mock数据 + 真实API + 手动录入
- 饮食保存: submit按钮接后端API
- 侧边栏: 健康概览横排 + 统一紫色配色
- 运动计划: 修复JSON反序列化(record→手动解析)
- VLM: prompt优化 + 模型切qwen3-vl-plus + 1280px压缩
- UI颜色: 加深主色 8B9CF7→6C5CE7
- 个人信息页精简 + 健康档案可编辑重设计
- 保洁: 删除无用agent_bar + 删除意见反馈/关于我们
- 新增AI提示词文档

backend/src/Health.WebApi/Endpoints/ai_chat_endpoints.cs

@@ -28,6 +28,7 @@ public static class AiChatEndpoints
             AppDbContext db,
             DeepSeekClient llmClient,
             PromptManager promptManager,
+            VisionClient visionClient,
             CancellationToken ct) =>
         {
             // 支持 token 通过 query string（浏览器 EventSource）或 header 传递
@@ -153,7 +154,7 @@ public static class AiChatEndpoints
                         object toolResult;
                         try
                         {
-                            toolResult = await ExecuteToolCall(tc.Function.Name, tc.Function.Arguments, db, userId.Value);
+                            toolResult = await ExecuteToolCall(tc.Function.Name, tc.Function.Arguments, db, userId.Value, visionClient);
                         }
                         catch (Exception ex)
                         {
@@ -265,24 +266,23 @@ public static class AiChatEndpoints
                     await file.CopyToAsync(stream, ct);
 
                 var compressedPath = Path.Combine(uploadsDir, $"compressed_{safeName}");
-                CompressImage(filePath, compressedPath, maxWidth: 1280, quality: 75L);
+                CompressImage(filePath, compressedPath, maxWidth: 1280, quality: 85L);
                 var compressedBytes = await File.ReadAllBytesAsync(compressedPath, ct);
                 var base64 = Convert.ToBase64String(compressedBytes);
                 imageUrls.Add($"data:image/jpeg;base64,{base64}");
             }
 
             var prompt = """
-    识别图片中的食物，以 JSON 数组格式返回，每项包含：
-    - name: 食物名称
-    - portion: 份量描述（如"约1碗"、"约200g"）
-    - calories: 估算热量（千卡，整数）
-    只返回 JSON 数组，不要任何其他文字。
-    示例：[{"name":"米饭","portion":"约1碗","calories":150},{"name":"番茄炒蛋","portion":"约1份","calories":200},{...}]
+    Describe everything you see in this image in detail. What objects are present? What are their colors, shapes, labels?
+    If there are any food items, drinks, or beverages, identify them specifically by name.
+    Then, for each food/drink item, estimate the portion size and calories (kcal).
+    Return ONLY a JSON array: [{"name":"item name","portion":"estimated portion","calories":estimated_calories}]
+    If you cannot identify something, say so in the name field.
     """;
 
             try
             {
-                var response = await visionClient.VisionAsync(prompt, imageUrls, userText: "请看图识别食物", ct: ct);
+                var response = await visionClient.VisionAsync(prompt, imageUrls, userText: "请看图识别食物", maxTokens: 8192, ct: ct);
                 var result = response.Choices?.FirstOrDefault()?.Message?.Content ?? "{}";
                 return Results.Ok(new { code = 0, data = result, message = (string?)null });
             }
@@ -328,10 +328,11 @@ public static class AiChatEndpoints
         AgentType.Consultation => ConsultationAgentHandler.Tools,
         AgentType.Report => ReportAgentHandler.Tools,
         AgentType.Exercise => ExerciseAgentHandler.Tools,
+        AgentType.Onboarding => [CommonAgentHandler.ManageArchiveTool, CommonAgentHandler.CheckArchiveTool],
         _ => CommonAgentHandler.Tools,
     };
 
-    private static Task<object> ExecuteToolCall(string toolName, string arguments, AppDbContext db, Guid userId)
+    private static Task<object> ExecuteToolCall(string toolName, string arguments, AppDbContext db, Guid userId, VisionClient? visionClient = null)
     {
         using var jsonDoc = JsonDocument.Parse(arguments);
         var root = jsonDoc.RootElement;
@@ -343,8 +344,11 @@ public static class AiChatEndpoints
             "check_archive" => CommonAgentHandler.Execute(toolName, root, db, userId),
             "manage_medication" => MedicationAgentHandler.Execute(toolName, root, db, userId),
             "estimate_food_text" => DietAgentHandler.Execute(toolName, root, db, userId),
-            "analyze_report" => ReportAgentHandler.Execute(toolName, root, db, userId),
+            "analyze_report" => visionClient != null
+                ? ReportAgentHandler.AnalyzeReport(db, userId, root, visionClient)
+                : Task.FromResult<object>(new { success = false, message = "VLM服务未配置" }),
             "manage_exercise" => ExerciseAgentHandler.Execute(toolName, root, db, userId),
+            "manage_archive" => CommonAgentHandler.ExecuteManageArchive(db, userId, root),
             "request_doctor" => ConsultationAgentHandler.Execute(toolName, root, db, userId),
             _ => Task.FromResult<object>(new { success = false, message = $"未知工具: {toolName}" })
         };