feat: 修复 AI 风险语义并隔离 Mock 数据

README.md

@@ -67,8 +67,8 @@
 
 ### 1.9 AI 拍照巡检闭环（计划 #5/#6/#8/#9）
 
-- `ai-service/`：FastAPI + ONNX Runtime（默认 mock 模式；`POST /detect` 返回框/类别/置信度，`POST /stream-detect` 拉流抽帧骨架，`GET /metrics` 监控指标）
+- `ai-service/`：FastAPI + ONNX Runtime（默认 mock 模式；`POST /detect` 返回框/类别/置信度、`modelVersion`、`isMock`、`abnormalProbability`，`POST /internal/stream-tasks` 受限拉流任务，`GET /metrics` 监控指标）
-- 拍照上传 → AI 检测 → 风险评分（0-100 分，绿/黄/橙/红四级）→ 巡检记录（`inspection_records`，`Idempotency-Key` 幂等）
+- 拍照上传 → AI 检测 → 风险评分（0-100 分，绿/黄/橙/红四级，只消费 AI 异常概率，缺失项不按 0 参与）→ 巡检记录（`inspection_records`，`Idempotency-Key` 幂等）
 - 小程序「拍照巡检」页；Web「巡检记录」页（技术员复查）
 
 ### 1.10 知识库与阶段风险提示（计划 #10/#13）
@@ -595,8 +595,8 @@ python -m venv .venv
 MODEL_MODE=mock .venv\Scripts\python -m uvicorn app.main:app --host 0.0.0.0 --port 8000
 ```
 
-- 接口：`GET /health`、`POST /detect`（multipart 图片）、`POST /stream-detect`（摄像头流拉帧骨架）、`GET /metrics`（含 GPU 信息）
+- 接口：`GET /health`、`POST /detect`（multipart 图片，返回 `abnormalProbability`）、`POST /internal/stream-tasks`（受限摄像头流拉帧任务）、`GET /metrics`（含 GPU 信息）
-- 默认 mock 模式；训练恢复后放 `models/best.onnx` 并设 `MODEL_MODE=onnx`（`MODEL_LABELS` 可配类别）
+- 默认 mock 模式且响应带 `isMock=true`；训练恢复后放 `models/best.onnx` 并设 `MODEL_MODE=onnx`（`MODEL_LABELS` 可配类别、`MODEL_VERSION` 必须显式配置）
 - 开发服务器部署：`/home/pan/ai-service`（venv + `start.sh`，:8000），服务器 pip 源已配置清华镜像
 
 ### 12.7 开发服务器部署摘要

ai-service/README.md

@@ -5,7 +5,7 @@ FastAPI + ONNX Runtime 的蚕病检测推理服务。YOLO 训练挂起期间以
 ## 接口
 
 - `GET /health` → `{"status":"ok","model":"mock|onnx"}`
-- `POST /detect`（multipart 字段 `file`）→ `{"model":"mock","detections":[{"bbox":{x,y,w,h},"class":"healthy|sick","confidence":0.95}]}`
+- `POST /detect`（multipart 字段 `file`）→ `{"model":"mock","modelVersion":"...","isMock":true,"status":"healthy|abnormal|unknown","abnormalProbability":0,"detections":[{"bbox":{x,y,w,h},"class":"healthy|sick","confidence":0.95}]}`
 - `POST /internal/stream-tasks`（需 `X-Internal-Key`）→ 创建受限拉流任务，不再接受客户端任意 URL
 
 ## 本地运行
@@ -33,6 +33,7 @@ powershell -ExecutionPolicy Bypass -File scripts/verify.ps1
 | `MODEL_MODE` | `mock` | `mock` / `onnx` |
 | `MODEL_PATH` | `models/best.onnx` | ONNX 模型路径 |
 | `MODEL_LABELS` | `healthy,sick` | 类别列表（逗号分隔） |
+| `MODEL_VERSION` | `mock-2026.08.14` / `best.onnx` | 模型版本；真实模型上线时必须显式配置 |
 | `MOCK_CLASS` | `healthy` | mock 返回类别 |
 | `MOCK_CONFIDENCE` | `0.95` | mock 返回置信度 |
 | `INTERNAL_API_KEY` | `silk-internal-2026` | 内部接口认证密钥 |
@@ -42,5 +43,6 @@ powershell -ExecutionPolicy Bypass -File scripts/verify.ps1
 
 ## 说明
 
-- 只做检测，风险评分在 Go 后端计算（#9：0.5×AI 置信度 + 0.2×环境 + 0.15×阶段 + 0.15×整齐度）。
+- 只做检测并返回 `abnormalProbability`；风险评分在 Go 后端计算（#9：0.5×AI 异常概率 + 0.2×环境 + 0.15×阶段 + 0.15×整齐度，缺失项归一化）。
+- `healthy` 高置信度不贡献异常概率；空检测或 `unknown` 返回 `unknown`，不自动视为健康。
 - ONNX 后处理按 YOLOv8 常见输出格式实现（含 letterbox 与 NMS），训练产物出来后需用真实模型校准验证。

ai-service/app/config.py

@@ -5,6 +5,7 @@ MODEL_MODE = os.getenv("MODEL_MODE", "mock")
 MODEL_PATH = os.getenv("MODEL_PATH", "models/best.onnx")
 # YOLO 类别（二分类训练基线：healthy/sick；7 类病种扩展后再调整）
 MODEL_LABELS = os.getenv("MODEL_LABELS", "healthy,sick").split(",")
+MODEL_VERSION = os.getenv("MODEL_VERSION", "mock-2026.08.14" if MODEL_MODE == "mock" else "best.onnx")
 # mock 模式返回的固定结果
 MOCK_CLASS = os.getenv("MOCK_CLASS", "healthy")
 MOCK_CONFIDENCE = float(os.getenv("MOCK_CONFIDENCE", "0.95"))

ai-service/app/detector.py

@@ -5,6 +5,52 @@ from abc import ABC, abstractmethod
 from PIL import Image
 
 
+def _label_set(labels) -> set[str]:
+    return {label.strip().lower() for label in labels if label.strip()}
+
+
+def abnormal_probability(detections: list[dict], labels: tuple[str, ...] | list[str]) -> float:
+    """单帧异常概率：取异常类检测的最高置信度，避免多框求和造成虚高。"""
+    abnormal = _label_set(labels) - {"healthy"}
+    best = 0.0
+    for detection in detections:
+        class_name = str(detection.get("class_name", "")).strip().lower()
+        if class_name in ("", "healthy", "unknown"):
+            continue
+        if abnormal and class_name not in abnormal:
+            continue
+        try:
+            confidence = float(detection.get("confidence", 0))
+        except (TypeError, ValueError):
+            confidence = 0.0
+        best = max(best, min(1.0, max(0.0, confidence)))
+    return best
+
+
+def detection_status(detections: list[dict], labels: tuple[str, ...] | list[str]) -> str:
+    """空检测或 unknown 类不再被当作 healthy。"""
+    if not detections:
+        return "unknown"
+    abnormal = _label_set(labels) - {"healthy"}
+    healthy_seen = False
+    unknown_seen = False
+    for detection in detections:
+        class_name = str(detection.get("class_name", "")).strip().lower()
+        if abnormal and class_name in abnormal:
+            return "abnormal"
+        if not abnormal and class_name not in ("", "healthy", "unknown"):
+            return "abnormal"
+        if class_name == "healthy":
+            healthy_seen = True
+        elif class_name in ("", "unknown"):
+            unknown_seen = True
+        else:
+            unknown_seen = True
+    if healthy_seen and not unknown_seen:
+        return "healthy"
+    return "unknown"
+
+
 class Detector(ABC):
     @abstractmethod
     def detect(self, image_bytes: bytes) -> list[dict]:

ai-service/app/main.py

@@ -6,7 +6,7 @@ import time
 from fastapi import FastAPI, File, HTTPException, Request, UploadFile
 
 from . import config
-from .detector import MockDetector, ONNXDetector
+from .detector import MockDetector, ONNXDetector, abnormal_probability, detection_status
 from .stream_tasks import StreamTaskWorker, is_allowed_stream_ref
 
 app = FastAPI(title="Silk AI Service", version="0.1.0")
@@ -36,7 +36,12 @@ else:
 
 @app.get("/health")
 def health():
-    return {"status": "ok", "model": config.MODEL_MODE}
+    return {
+        "status": "ok",
+        "model": config.MODEL_MODE,
+        "modelVersion": config.MODEL_VERSION,
+        "isMock": isinstance(detector, MockDetector),
+    }
 
 
 @app.post("/detect")
@@ -50,6 +55,10 @@ async def detect(file: UploadFile = File(...)):
         raise HTTPException(status_code=400, detail=str(exc)) from exc
     return {
         "model": config.MODEL_MODE,
+        "modelVersion": config.MODEL_VERSION,
+        "isMock": isinstance(detector, MockDetector),
+        "status": detection_status(detections, config.MODEL_LABELS),
+        "abnormalProbability": abnormal_probability(detections, config.MODEL_LABELS),
         "detections": [
             {
                 "bbox": d["bbox"],
@@ -112,6 +121,8 @@ def metrics():
             gpu = None
     return {
         "model": config.MODEL_MODE,
+        "modelVersion": config.MODEL_VERSION,
+        "isMock": isinstance(detector, MockDetector),
         "uptimeSeconds": int(time.time() - START_TIME),
         "requests": reqs,
         "avgLatencyMs": round(avg, 2),

ai-service/tests/test_api.py

@@ -6,7 +6,9 @@ os.environ.setdefault("ALLOWED_STREAM_HOSTS", "localhost,127.0.0.1,100.83.103.1"
 
 from fastapi.testclient import TestClient
 
+from app import main as main_module
 from app.main import app
+from app.detector import MockDetector
 
 # 1x1 透明 PNG
 TINY_PNG = base64.b64decode(
@@ -22,6 +24,8 @@ def test_health():
     body = r.json()
     assert body["status"] == "ok"
     assert body["model"] in ("mock", "onnx")
+    assert body["modelVersion"]
+    assert body["isMock"] is True
 
 
 def test_detect_ok():
@@ -29,6 +33,10 @@ def test_detect_ok():
     assert r.status_code == 200
     body = r.json()
     assert body["model"] == "mock"
+    assert body["modelVersion"]
+    assert body["isMock"] is True
+    assert body["status"] == "healthy"
+    assert body["abnormalProbability"] == 0
     assert len(body["detections"]) >= 1
     d = body["detections"][0]
     assert d["class"] in ("healthy", "sick")
@@ -36,6 +44,28 @@ def test_detect_ok():
     assert d["bbox"]["w"] > 0
 
 
+def test_detect_uses_abnormal_class_confidence(monkeypatch):
+    monkeypatch.setattr(main_module, "detector", MockDetector(class_name="sick", confidence=0.92))
+    r = client.post("/detect", files={"file": ("a.png", TINY_PNG, "image/png")})
+    assert r.status_code == 200
+    body = r.json()
+    assert body["status"] == "abnormal"
+    assert body["abnormalProbability"] == 0.92
+
+
+def test_detect_empty_result_is_unknown(monkeypatch):
+    class EmptyDetector:
+        def detect(self, image_bytes):
+            return []
+
+    monkeypatch.setattr(main_module, "detector", EmptyDetector())
+    r = client.post("/detect", files={"file": ("a.png", TINY_PNG, "image/png")})
+    assert r.status_code == 200
+    body = r.json()
+    assert body["status"] == "unknown"
+    assert body["abnormalProbability"] == 0
+
+
 def test_detect_empty_file_rejected():
     r = client.post("/detect", files={"file": ("a.png", b"", "image/png")})
     assert r.status_code == 400

ai-service/tests/test_detector.py

@@ -2,7 +2,7 @@ import base64
 
 import pytest
 
-from app.detector import MockDetector
+from app.detector import MockDetector, abnormal_probability, detection_status
 
 # 1x1 透明 PNG
 TINY_PNG = base64.b64decode(
@@ -23,3 +23,22 @@ def test_mock_detector_rejects_invalid_image():
     det = MockDetector()
     with pytest.raises(ValueError):
         det.detect(b"not an image")
+
+
+def test_abnormal_probability_ignores_healthy_and_unknown():
+    detections = [
+        {"class_name": "healthy", "confidence": 0.95},
+        {"class_name": "unknown", "confidence": 0.8},
+        {"class_name": "sick", "confidence": 0.72},
+    ]
+    assert abnormal_probability(detections, ("healthy", "sick")) == 0.72
+
+
+def test_detection_status_returns_unknown_for_empty_or_unknown():
+    assert detection_status([], ("healthy", "sick")) == "unknown"
+    assert detection_status([{"class_name": "unknown", "confidence": 0.8}], ("healthy", "sick")) == "unknown"
+    assert detection_status([{"class_name": "healthy", "confidence": 0.95}], ("healthy", "sick")) == "healthy"
+    assert detection_status(
+        [{"class_name": "white_muscardine", "confidence": 0.7}],
+        ("healthy", "white_muscardine", "nuclear_polyhedrosis"),
+    ) == "abnormal"

miniapp/src/pages/inspection/index.tsx

@@ -113,6 +113,7 @@ const InspectionPage: React.FC = () => {
                 ? '检测到疑似异常'
                 : '未见明显异常'
               : 'AI 检测失败'}
+            {result.isMock ? '（联调 Mock）' : ''}
           </Text>
           {result.aiStatus === 'done' && result.detections && result.detections.length > 0 ? (
             <View className={styles.detectionList}>
@@ -145,6 +146,7 @@ const InspectionPage: React.FC = () => {
                     ? '疑似异常'
                     : '正常'
                   : '检测失败'}
+                {rec.isMock ? '（Mock）' : ''}
               </Text>
               <Text className={styles.historyTime}>{formatRelativeTime(rec.createdAt)}</Text>
             </View>

miniapp/src/types/index.ts

@@ -223,6 +223,17 @@ export interface InspectionRecord {
   detections?: AIDetection[];
   riskScore?: number;
   riskLevel?: string;
+  riskAssessment?: {
+    score: number;
+    level: string;
+    confidence: string;
+    modelVersion: string;
+    ruleVersion: string;
+    components: Record<string, number | null>;
+    missing: string[];
+  };
+  modelVersion?: string;
+  isMock?: boolean;
   aiStatus: string;
   idempotencyKey?: string;
   createdAt?: string;

scripts/risk_historical_review.sql

@@ -0,0 +1,27 @@
+-- Task 6 Step 5: 只读历史数据评估报告
+-- 仅统计，不更新数据；未经人工确认不得批量重算或覆盖历史 risk_score/risk_level。
+
+WITH real_inspections AS (
+    SELECT
+        id,
+        risk_level,
+        risk_score,
+        detections
+    FROM inspection_records
+    WHERE ai_status = 'done'
+      AND (is_mock IS NULL OR is_mock = false)
+)
+SELECT
+    risk_level,
+    count(*) AS records,
+    count(*) FILTER (
+        WHERE EXISTS (
+            SELECT 1
+            FROM jsonb_array_elements(detections) AS d
+            WHERE d->>'class' = 'healthy'
+              AND COALESCE((d->>'confidence')::numeric, 0) >= 0.80
+        )
+    ) AS healthy_high_conf_records
+FROM real_inspections
+GROUP BY risk_level
+ORDER BY risk_level;

server-go/cmd/server/main.go

@@ -111,7 +111,7 @@ func main() {
 	handler.RegisterVideoRecordRoutes(api, db, mediaSvc, cfg)
 	handler.RegisterStorageRoutes(api, db)
 	handler.RegisterKnowledgeRoutes(api, db, s3Svc, cfg.S3BucketImages)
-	handler.RegisterInspectionRoutes(api, db, s3Svc, aiSvc, cfg.S3BucketImages, wechatSvc, cfg.WechatTemplateInspection)
+	handler.RegisterInspectionRoutes(api, db, s3Svc, aiSvc, cfg.S3BucketImages, wechatSvc, cfg.WechatTemplateInspection, cfg.AppEnv)
 	handler.RegisterTrayBatchRoutes(api, db)
 	handler.RegisterWechatRoutes(api, db, wechatSvc)
 	handler.RegisterWeatherRoutes(api, db, weatherSvc)

server-go/internal/database/migrate.go

@@ -8,12 +8,12 @@ import (
 	"github.com/golang-migrate/migrate/v4"
 	"github.com/golang-migrate/migrate/v4/database/postgres"
 	"github.com/golang-migrate/migrate/v4/source/iofs"
-	"silk-server-go/migrations"
 	"gorm.io/gorm"
+	"silk-server-go/migrations"
 )
 
 // CurrentSchemaVersion 是当前后端代码期望的迁移版本。
-const CurrentSchemaVersion = "1"
+const CurrentSchemaVersion = "2"
 
 // RunMigrations 使用嵌入式 SQL 迁移文件将数据库升级到最新版本。
 func RunMigrations(db *gorm.DB) error {

server-go/internal/database/migrate_test.go

@@ -100,4 +100,8 @@ func TestEmbeddedMigrationsIncludeBaseline(t *testing.T) {
 	if version != 1 {
 		t.Fatalf("expected baseline migration version 1, got %d", version)
 	}
+	next, err := driver.Next(version)
+	if err != nil || next != 2 {
+		t.Fatalf("expected risk assessment migration version 2, got %d (err %v)", next, err)
+	}
 }

server-go/internal/handler/health.go

@@ -44,7 +44,7 @@ func roomHealthProfile(db *gorm.DB) gin.HandlerFunc {
 		}
 		db.Table("inspection_records").
 			Select("risk_level, count(*) AS cnt").
-			Where("room_id = ? AND ai_status = 'done' AND risk_level IS NOT NULL AND created_at >= ?", id, since).
+			Where("room_id = ? AND ai_status = 'done' AND risk_level IS NOT NULL AND created_at >= ? AND COALESCE(is_mock, false) = false", id, since).
 			Group("risk_level").
 			Scan(&riskRows)
 		riskCounts := map[string]int64{}

server-go/internal/handler/inspection.go

@@ -5,6 +5,7 @@ import (
 	"context"
 	"encoding/json"
 	"io"
+	"log/slog"
 	"net/http"
 	"regexp"
 	"strconv"
@@ -26,8 +27,8 @@ func isUUID(s string) bool {
 }
 
 // RegisterInspectionRoutes 注册 AI 巡检路由
-func RegisterInspectionRoutes(rg *gin.RouterGroup, db *gorm.DB, s3 *service.S3Service, ai *service.AIClient, imageBucket string, wechat *service.WechatService, inspectionTemplateID string) {
+func RegisterInspectionRoutes(rg *gin.RouterGroup, db *gorm.DB, s3 *service.S3Service, ai *service.AIClient, imageBucket string, wechat *service.WechatService, inspectionTemplateID string, appEnv string) {
-	rg.POST("/inspections", middleware.RequirePermission(db, "inspection:create"), createInspection(db, s3, ai, imageBucket, wechat, inspectionTemplateID))
+	rg.POST("/inspections", middleware.RequirePermission(db, "inspection:create"), createInspection(db, s3, ai, imageBucket, wechat, inspectionTemplateID, appEnv))
 	rg.GET("/inspections", middleware.RequirePermission(db, "inspection:read"), listInspections(db))
 }
 
@@ -43,9 +44,29 @@ func currentUserID(c *gin.Context) *string {
 	return nil
 }
 
+func buildRiskInput(detRes *service.AIDetectResponse) service.RiskInput {
+	status := detRes.Status
+	if status == "" {
+		status = service.AIDetectionStatus(detRes.Detections)
+	}
+	modelVersion := detRes.ModelVersion
+	if modelVersion == "" {
+		modelVersion = "unknown"
+	}
+	in := service.RiskInput{ModelVersion: modelVersion}
+	if status != "unknown" {
+		aiProb := detRes.AbnormalProbability
+		if len(detRes.Detections) > 0 && aiProb == 0 {
+			aiProb = service.AbnormalProbability(detRes.Detections)
+		}
+		in.AI = &aiProb
+	}
+	return in
+}
+
 // createInspection 拍照巡检：图片存 S3 → 调 AI /detect → 写记录。
 // 幂等：客户端传 Idempotency-Key 头时，重复请求返回已有记录。
-func createInspection(db *gorm.DB, s3 *service.S3Service, ai *service.AIClient, bucket string, wechat *service.WechatService, inspectionTemplateID string) gin.HandlerFunc {
+func createInspection(db *gorm.DB, s3 *service.S3Service, ai *service.AIClient, bucket string, wechat *service.WechatService, inspectionTemplateID string, appEnv string) gin.HandlerFunc {
 	return func(c *gin.Context) {
 		idemKey := strings.TrimSpace(c.GetHeader("Idempotency-Key"))
 		roomID := strings.TrimSpace(c.PostForm("roomId"))
@@ -113,26 +134,30 @@ func createInspection(db *gorm.DB, s3 *service.S3Service, ai *service.AIClient,
 		} else {
 			raw, _ := json.Marshal(detRes.Detections)
 			rec.Detections = raw
-
+			isMock := detRes.IsMock
-			// 风险评分（#9）：AI 置信度取检测结果最大值；环境/阶段系数在有 roomId 时按房间数据计算
+			rec.IsMock = &isMock
-			aiConf := 0.0
+			modelVersion := detRes.ModelVersion
-			for _, d := range detRes.Detections {
+			if modelVersion == "" {
-				if d.Confidence > aiConf {
+				modelVersion = "unknown"
-					aiConf = d.Confidence
 			}
-			}
+			rec.ModelVersion = &modelVersion
-			stageCoef, envCoef := loadRoomRisk(db, roomID)
-			score := service.ComputeRiskScore(service.RiskInput{
-				AI:    aiConf,
-				Env:   envCoef,
-				Stage: stageCoef,
-			})
-			rec.RiskScore = &score
-			level := service.RiskLevel(score)
-			rec.RiskLevel = &level
 
-			// 微信订阅消息（#11 骨架）：风险非绿且用户已授权时异步推送
+			if appEnv == "production" && isMock {
-			if key := service.WechatTemplateKey(level); key != "" {
+				slog.Error("生产环境收到 Mock AI 检测结果，按失败记录", "modelVersion", modelVersion, "roomId", roomID)
+				rec.AIStatus = "failed"
+			} else {
+				// 风险评分（#9 V2）：只消费 AI 异常概率；缺失环境/阶段不填 0
+				riskInput := buildRiskInput(detRes)
+				riskInput.Env, riskInput.Stage = loadRoomRisk(db, roomID)
+				assessment := service.ComputeRiskScore(riskInput)
+				rec.RiskScore = &assessment.Score
+				rec.RiskLevel = &assessment.Level
+				rawRisk, _ := json.Marshal(assessment)
+				rec.RiskAssessment = rawRisk
+
+				// 微信订阅消息（#11 骨架）：Mock 结果不进入告警，风险非绿且用户已授权时异步推送
+				if !isMock {
+					if key := service.WechatTemplateKey(assessment.Level); key != "" {
 						go func(uid *string, lv string, sc float64) {
 							if uid == nil || !wechat.Configured() || inspectionTemplateID == "" {
 								return
@@ -155,7 +180,9 @@ func createInspection(db *gorm.DB, s3 *service.S3Service, ai *service.AIClient,
 								service.BuildSubscribeData(lv, sc),
 								"pages/inspection/index",
 							)
-				}(rec.UserID, level, score)
+						}(rec.UserID, assessment.Level, assessment.Score)
+					}
+				}
 			}
 		}
 
@@ -175,17 +202,19 @@ func createInspection(db *gorm.DB, s3 *service.S3Service, ai *service.AIClient,
 	}
 }
 
-// loadRoomRisk 加载房间阶段系数与环境系数（无房间/无数据时返回 0）
+// loadRoomRisk 加载房间阶段系数与环境系数（无房间/无数据时返回 nil）
-func loadRoomRisk(db *gorm.DB, roomID string) (stageCoef, envCoef float64) {
+func loadRoomRisk(db *gorm.DB, roomID string) (*float64, *float64) {
 	if roomID == "" {
-		return 0, 0
+		return nil, nil
 	}
 	var room model.Room
 	if db.Where("id = ?", roomID).First(&room).Error != nil {
-		return 0, 0
+		return nil, nil
 	}
+	var stageCoef *float64
 	if room.Stage != nil {
-		stageCoef = service.StageCoefficient(*room.Stage)
+		value := service.StageCoefficient(*room.Stage)
+		stageCoef = &value
 	}
 
 	var humidity, temperature *float64
@@ -209,7 +238,11 @@ func loadRoomRisk(db *gorm.DB, roomID string) (stageCoef, envCoef float64) {
 		First(&t).Error; err == nil {
 		temperature = &t.Value
 	}
-	envCoef = service.EnvCoefficient(temperature, humidity)
+	var envCoef *float64
+	if humidity != nil || temperature != nil {
+		value := service.EnvCoefficient(temperature, humidity)
+		envCoef = &value
+	}
 	return stageCoef, envCoef
 }
 

server-go/internal/handler/inspection_test.go

@@ -1,6 +1,10 @@
 package handler
 
-import "testing"
+import (
+	"testing"
+
+	"silk-server-go/internal/service"
+)
 
 func TestIsUUID(t *testing.T) {
 	valid := []string{
@@ -19,3 +23,33 @@ func TestIsUUID(t *testing.T) {
 		}
 	}
 }
+
+func TestBuildRiskInputUsesAbnormalProbability(t *testing.T) {
+	in := buildRiskInput(&service.AIDetectResponse{
+		ModelVersion:        "silk-yolo-2026.08.1",
+		Status:              "abnormal",
+		AbnormalProbability: 0.93,
+	})
+	if in.AI == nil || *in.AI != 0.93 {
+		t.Fatalf("AI 异常概率应进入风险输入，实际 %v", in.AI)
+	}
+	if in.ModelVersion != "silk-yolo-2026.08.1" {
+		t.Errorf("modelVersion = %s", in.ModelVersion)
+	}
+}
+
+func TestBuildRiskInputUnknownKeepsAIMissing(t *testing.T) {
+	in := buildRiskInput(&service.AIDetectResponse{Status: "unknown"})
+	if in.AI != nil {
+		t.Fatalf("unknown 状态不应把 0 当作 AI 组件，实际 %v", in.AI)
+	}
+}
+
+func TestBuildRiskInputFallsBackToDetections(t *testing.T) {
+	in := buildRiskInput(&service.AIDetectResponse{
+		Detections: []service.AIDetection{{ClassName: "sick", Confidence: 0.9}},
+	})
+	if in.AI == nil || *in.AI != 0.9 {
+		t.Fatalf("旧 AI 响应应从检测类别计算异常概率，实际 %v", in.AI)
+	}
+}

server-go/internal/model/inspection.go

@@ -14,6 +14,9 @@ type InspectionRecord struct {
 	Detections     json.RawMessage `gorm:"column:detections;type:jsonb" json:"detections,omitempty"`
 	RiskScore      *float64        `gorm:"column:risk_score;type:float" json:"riskScore,omitempty"`
 	RiskLevel      *string         `gorm:"column:risk_level;size:16" json:"riskLevel,omitempty"`
+	RiskAssessment json.RawMessage `gorm:"column:risk_assessment;type:jsonb" json:"riskAssessment,omitempty"`
+	ModelVersion   *string         `gorm:"column:model_version;size:64" json:"modelVersion,omitempty"`
+	IsMock         *bool           `gorm:"column:is_mock;default:false" json:"isMock,omitempty"`
 	AIStatus       string          `gorm:"column:ai_status;size:16;default:done" json:"aiStatus"`
 	IdempotencyKey *string         `gorm:"column:idempotency_key;size:128;uniqueIndex" json:"idempotencyKey,omitempty"`
 	CreatedAt      time.Time       `gorm:"type:timestamptz" json:"createdAt"`

server-go/internal/service/ai_client.go

@@ -30,9 +30,37 @@ type AIDetection struct {
 // AIDetectResponse /detect 响应
 type AIDetectResponse struct {
 	Model               string        `json:"model"`
+	ModelVersion        string        `json:"modelVersion"`
+	IsMock              bool          `json:"isMock"`
+	Status              string        `json:"status"`
+	AbnormalProbability float64       `json:"abnormalProbability"`
 	Detections          []AIDetection `json:"detections"`
 }
 
+// AIDetectionStatus 从检测结果归纳 AI 状态；空检测或 unknown 不当作 healthy。
+func AIDetectionStatus(detections []AIDetection) string {
+	if len(detections) == 0 {
+		return "unknown"
+	}
+	healthySeen := false
+	unknownSeen := false
+	for _, d := range detections {
+		class := strings.ToLower(strings.TrimSpace(d.ClassName))
+		switch class {
+		case "healthy":
+			healthySeen = true
+		case "", "unknown":
+			unknownSeen = true
+		default:
+			return "abnormal"
+		}
+	}
+	if healthySeen && !unknownSeen {
+		return "healthy"
+	}
+	return "unknown"
+}
+
 // AIClient ai-service HTTP 客户端
 type AIClient struct {
 	baseURL string

server-go/internal/service/ai_client_test.go

@@ -25,6 +25,10 @@ func TestAIClientDetectParsesResult(t *testing.T) {
 		w.Header().Set("Content-Type", "application/json")
 		_ = json.NewEncoder(w).Encode(map[string]any{
 			"model":               "mock",
+			"modelVersion":        "silk-yolo-2026.08.1",
+			"isMock":              true,
+			"status":              "abnormal",
+			"abnormalProbability": 0.93,
 			"detections": []map[string]any{
 				{"bbox": map[string]float64{"x": 1, "y": 2, "w": 3, "h": 4}, "class": "sick", "confidence": 0.93},
 			},
@@ -40,6 +44,15 @@ func TestAIClientDetectParsesResult(t *testing.T) {
 	if res.Model != "mock" {
 		t.Errorf("model = %s, want mock", res.Model)
 	}
+	if res.ModelVersion != "silk-yolo-2026.08.1" {
+		t.Errorf("modelVersion = %s, want silk-yolo-2026.08.1", res.ModelVersion)
+	}
+	if !res.IsMock {
+		t.Error("isMock 应解析为 true")
+	}
+	if res.Status != "abnormal" || res.AbnormalProbability != 0.93 {
+		t.Errorf("AI 语义字段解析不正确: %+v", res)
+	}
 	if len(res.Detections) != 1 {
 		t.Fatalf("detections 数量 = %d, want 1", len(res.Detections))
 	}
@@ -49,6 +62,21 @@ func TestAIClientDetectParsesResult(t *testing.T) {
 	}
 }
 
+func TestAIDetectionStatus(t *testing.T) {
+	if got := AIDetectionStatus(nil); got != "unknown" {
+		t.Errorf("空检测应为 unknown，实际 %s", got)
+	}
+	if got := AIDetectionStatus([]AIDetection{{ClassName: "healthy"}}); got != "healthy" {
+		t.Errorf("全健康应为 healthy，实际 %s", got)
+	}
+	if got := AIDetectionStatus([]AIDetection{{ClassName: "unknown"}}); got != "unknown" {
+		t.Errorf("unknown 应为 unknown，实际 %s", got)
+	}
+	if got := AIDetectionStatus([]AIDetection{{ClassName: "sick"}}); got != "abnormal" {
+		t.Errorf("异常类别应为 abnormal，实际 %s", got)
+	}
+}
+
 func TestAIClientDetectServerError(t *testing.T) {
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		http.Error(w, "boom", http.StatusInternalServerError)

server-go/internal/service/cross_validate.go

@@ -1,5 +1,7 @@
 package service
 
+import "strings"
+
 // CrossValidate 交叉验证：AI 检测结果 vs LAMP 检测结果（规格书：一致→确认诊断；不一致→升级专家会诊）
 func CrossValidate(aiClass, lampResult string, lampDiseases []string) (bool, string) {
 	switch {
@@ -7,6 +9,8 @@ func CrossValidate(aiClass, lampResult string, lampDiseases []string) (bool, str
 		return false, "LAMP 判读无效，建议复检或专家会诊"
 	case aiClass == "":
 		return false, "未找到关联巡检记录，暂无法交叉验证"
+	case aiClass == "unknown":
+		return false, "AI 结果未知，建议复检或专家会诊"
 	case aiClass == "sick" && lampResult == "positive":
 		return true, "AI 检出异常与 LAMP 阳性一致，确认诊断"
 	case aiClass == "healthy" && lampResult == "negative":
@@ -20,15 +24,26 @@ func CrossValidate(aiClass, lampResult string, lampDiseases []string) (bool, str
 	}
 }
 
-// AIClassFromDetections 从巡检检测结果归纳 AI 结论（任一非 healthy 视为 sick）
+// AIClassFromDetections 从巡检检测结果归纳 AI 结论；空检测或 unknown 不当作 healthy。
 func AIClassFromDetections(detections []AIDetection) string {
 	if len(detections) == 0 {
-		return ""
+		return "unknown"
 	}
+	healthySeen := false
+	unknownSeen := false
 	for _, d := range detections {
-		if d.ClassName != "healthy" {
+		class := strings.ToLower(strings.TrimSpace(d.ClassName))
+		switch class {
+		case "healthy":
+			healthySeen = true
+		case "", "unknown":
+			unknownSeen = true
+		default:
 			return "sick"
 		}
 	}
+	if healthySeen && !unknownSeen {
 		return "healthy"
 	}
+	return "unknown"
+}

server-go/internal/service/cross_validate_test.go

@@ -13,6 +13,7 @@ func TestCrossValidate(t *testing.T) {
 		{"AI健康+LAMP阴性 一致", "healthy", "negative", true},
 		{"AI异常+LAMP阴性 不一致", "sick", "negative", false},
 		{"AI健康+LAMP阳性 不一致", "healthy", "positive", false},
+		{"AI未知+LAMP阳性 不一致", "unknown", "positive", false},
 		{"LAMP无效 不一致", "sick", "invalid", false},
 		{"无AI记录 不一致", "", "positive", false},
 	}
@@ -28,8 +29,8 @@ func TestCrossValidate(t *testing.T) {
 }
 
 func TestAIClassFromDetections(t *testing.T) {
-	if got := AIClassFromDetections(nil); got != "" {
+	if got := AIClassFromDetections(nil); got != "unknown" {
-		t.Errorf("空检测应为空，实际 %s", got)
+		t.Errorf("空检测应为 unknown，实际 %s", got)
 	}
 	if got := AIClassFromDetections([]AIDetection{{ClassName: "healthy", Confidence: 0.9}}); got != "healthy" {
 		t.Errorf("全健康应为 healthy，实际 %s", got)
@@ -37,4 +38,7 @@ func TestAIClassFromDetections(t *testing.T) {
 	if got := AIClassFromDetections([]AIDetection{{ClassName: "healthy"}, {ClassName: "sick", Confidence: 0.6}}); got != "sick" {
 		t.Errorf("含 sick 应为 sick，实际 %s", got)
 	}
+	if got := AIClassFromDetections([]AIDetection{{ClassName: "unknown", Confidence: 0.6}}); got != "unknown" {
+		t.Errorf("unknown 应为 unknown，实际 %s", got)
+	}
 }

server-go/internal/service/risk.go

@@ -1,21 +1,117 @@
 package service
 
-import "math"
+import (
+	"math"
+	"strings"
+)
 
-// RiskInput 风险评分输入（各系数取值 0~1）
+// RiskRuleVersion 当前风险规则版本；权重仍沿用试点公式，待数据校准后升版。
+const RiskRuleVersion = "risk-v2-2026.08.14"
+
+// RiskInput 风险评分输入。nil 表示未采集，不能按 0 参与归一化。
 type RiskInput struct {
-	AI         float64 // AI 识别置信度（权重 0.5）
+	AI           *float64 // AI 异常概率（权重 0.5）
-	Env        float64 // 环境风险系数（权重 0.2）
+	Env          *float64 // 环境风险系数（权重 0.2）
-	Stage      float64 // 饲养阶段风险系数（权重 0.15）
+	Stage        *float64 // 饲养阶段风险系数（权重 0.15）
-	Uniformity float64 // 群体整齐度偏离度（权重 0.15）
+	Uniformity   *float64 // 群体整齐度偏离度（权重 0.15）
+	ModelVersion string
 }
 
-// ComputeRiskScore 按规格书 3.1.4 公式计算 0~100 风险分：
+// RiskAssessment 可解释风险输出。
-// 风险分 = 0.5×AI置信度 + 0.2×环境系数 + 0.15×阶段系数 + 0.15×整齐度偏离度
+type RiskAssessment struct {
-func ComputeRiskScore(in RiskInput) float64 {
+	Score        float64             `json:"score"`
-	score := 0.5*in.AI + 0.2*in.Env + 0.15*in.Stage + 0.15*in.Uniformity
+	Level        string              `json:"level"`
-	score = math.Max(0, math.Min(1, score))
+	Confidence   string              `json:"confidence"`
-	return score * 100
+	ModelVersion string              `json:"modelVersion"`
+	RuleVersion  string              `json:"ruleVersion"`
+	Components   map[string]*float64 `json:"components"`
+	Missing      []string            `json:"missing"`
+}
+
+// ComputeRiskScore 按可用权重归一化计算 0~100 风险分，缺失组件返回 null。
+// 当前 0.5/0.2/0.15/0.15 是待试点校准规则，不是科学结论。
+func ComputeRiskScore(in RiskInput) RiskAssessment {
+	defs := []struct {
+		key    string
+		weight float64
+		value  *float64
+	}{
+		{"aiAbnormalProbability", 0.5, in.AI},
+		{"environment", 0.2, in.Env},
+		{"stage", 0.15, in.Stage},
+		{"uniformity", 0.15, in.Uniformity},
+	}
+	components := make(map[string]*float64, len(defs))
+	missing := make([]string, 0, len(defs))
+	weighted := 0.0
+	totalWeight := 0.0
+	for _, def := range defs {
+		if def.value == nil {
+			components[def.key] = nil
+			missing = append(missing, def.key)
+			continue
+		}
+		value := clamp01(*def.value)
+		components[def.key] = &value
+		weighted += def.weight * value
+		totalWeight += def.weight
+	}
+
+	score := 0.0
+	if totalWeight > 0 {
+		score = math.Round(clamp01(weighted/totalWeight)*10000) / 100
+	}
+	return RiskAssessment{
+		Score:        score,
+		Level:        RiskLevel(score),
+		Confidence:   RiskConfidence(in),
+		ModelVersion: in.ModelVersion,
+		RuleVersion:  RiskRuleVersion,
+		Components:   components,
+		Missing:      missing,
+	}
+}
+
+// RiskConfidence 当前仅给出 low/medium/unknown，避免把未校准规则描述为 high。
+func RiskConfidence(in RiskInput) string {
+	if in.AI == nil {
+		return "unknown"
+	}
+	if *in.AI >= 0.25 {
+		return "medium"
+	}
+	return "low"
+}
+
+// AbnormalProbability 从 AI 检测结果计算异常概率：healthy/unknown 不贡献风险。
+// abnormalClasses 可选；缺省时任一非 healthy/unknown 类别都视为异常类别。
+func AbnormalProbability(detections []AIDetection, abnormalClasses ...string) float64 {
+	configured := make(map[string]struct{}, len(abnormalClasses))
+	for _, class := range abnormalClasses {
+		if class = strings.ToLower(strings.TrimSpace(class)); class != "" {
+			configured[class] = struct{}{}
+		}
+	}
+	best := 0.0
+	for _, d := range detections {
+		class := strings.ToLower(strings.TrimSpace(d.ClassName))
+		if class == "" || class == "healthy" || class == "unknown" {
+			continue
+		}
+		if len(configured) > 0 {
+			if _, ok := configured[class]; !ok {
+				continue
+			}
+		}
+		if d.Confidence > best {
+			best = d.Confidence
+		}
+	}
+	return clamp01(best)
+}
+
+func clamp01(v float64) float64 {
+	return math.Max(0, math.Min(1, v))
 }
 
 // RiskLevel 按规格书 3.1.4 分级：绿 0-30 / 黄 31-60 / 橙 61-80 / 红 81-100

server-go/internal/service/risk_test.go

@@ -1,30 +1,107 @@
 package service
 
-import "testing"
+import (
+	"math"
+	"testing"
+)
 
-func f(v float64) *float64 { return &v }
+func ptr(v float64) *float64 { return &v }
 
 func TestComputeRiskScoreWeights(t *testing.T) {
-	// 全 1：0.5*1 + 0.2*1 + 0.15*1 + 0.15*1 = 1 → 100
+	got := ComputeRiskScore(RiskInput{AI: ptr(1), Env: ptr(1), Stage: ptr(1), Uniformity: ptr(1)})
-	if s := ComputeRiskScore(RiskInput{AI: 1, Env: 1, Stage: 1, Uniformity: 1}); s != 100 {
+	if got.Score != 100 {
-		t.Errorf("全 1 应得 100，实际 %.2f", s)
+		t.Errorf("全 1 应得 100，实际 %.2f", got.Score)
 	}
-	// 仅 AI 置信度 1：0.5*1 = 0.5 → 50
+	if len(got.Missing) != 0 {
-	if s := ComputeRiskScore(RiskInput{AI: 1}); s != 50 {
+		t.Errorf("全组件可用时不应有 missing，实际 %v", got.Missing)
-		t.Errorf("仅 AI=1 应得 50，实际 %.2f", s)
 	}
-	// 0.5*0.8 + 0.2*0.5 = 0.5
+	if got.Components["aiAbnormalProbability"] == nil || got.Components["uniformity"] == nil {
-	if s := ComputeRiskScore(RiskInput{AI: 0.8, Env: 0.5}); s != 50 {
+		t.Errorf("全组件可用时 components 不应为 null: %v", got.Components)
-		t.Errorf("0.8/0.5 应得 50，实际 %.2f", s)
+	}
+}
+
+func TestComputeRiskScoreNormalizesMissingComponents(t *testing.T) {
+	got := ComputeRiskScore(RiskInput{AI: ptr(1)})
+	if got.Score != 100 {
+		t.Errorf("仅 AI=1 归一化后应得 100，实际 %.2f", got.Score)
+	}
+	if len(got.Missing) != 3 {
+		t.Errorf("missing 应为 environment/stage/uniformity，实际 %v", got.Missing)
+	}
+	if got.Components["environment"] != nil {
+		t.Error("缺失 environment 应序列化为 null")
+	}
+}
+
+func TestHealthyHighConfidenceDoesNotIncreaseRisk(t *testing.T) {
+	got := AbnormalProbability([]AIDetection{{ClassName: "healthy", Confidence: .95}})
+	if got != 0 {
+		t.Fatalf("healthy 高置信度不应产生异常概率，实际 %v", got)
+	}
+	assessment := ComputeRiskScore(RiskInput{AI: ptr(got)})
+	if assessment.Score != 0 {
+		t.Fatalf("healthy 高置信度风险分应为 0，实际 %.2f", assessment.Score)
+	}
+}
+
+func TestSickHighConfidenceIncreasesRisk(t *testing.T) {
+	got := AbnormalProbability([]AIDetection{{ClassName: "sick", Confidence: .9}})
+	if got != .9 {
+		t.Fatalf("sick 高置信度异常概率应为 0.9，实际 %v", got)
+	}
+	assessment := ComputeRiskScore(RiskInput{AI: ptr(got)})
+	if assessment.Score != 90 {
+		t.Fatalf("仅 AI=0.9 归一化后风险分应为 90，实际 %.2f", assessment.Score)
+	}
+}
+
+func TestAbnormalProbabilityIgnoresUnknown(t *testing.T) {
+	if got := AbnormalProbability([]AIDetection{{ClassName: "unknown", Confidence: .9}}); got != 0 {
+		t.Errorf("unknown 不应贡献异常概率，实际 %v", got)
+	}
+}
+
+func TestComputeRiskScoreMissingAll(t *testing.T) {
+	got := ComputeRiskScore(RiskInput{})
+	if got.Score != 0 {
+		t.Errorf("无任何组件时分数应为 0，实际 %.2f", got.Score)
+	}
+	if got.Confidence != "unknown" {
+		t.Errorf("无 AI 组件时 confidence 应为 unknown，实际 %s", got.Confidence)
+	}
+	if len(got.Missing) != 4 {
+		t.Errorf("missing 应为 4 项，实际 %v", got.Missing)
 	}
 }
 
 func TestComputeRiskScoreClamps(t *testing.T) {
-	if s := ComputeRiskScore(RiskInput{AI: 2, Env: 2, Stage: 2, Uniformity: 2}); s > 100 {
+	got := ComputeRiskScore(RiskInput{AI: ptr(2), Env: ptr(2), Stage: ptr(2), Uniformity: ptr(2)})
-		t.Errorf("应钳制到 100，实际 %.2f", s)
+	if got.Score > 100 {
+		t.Errorf("应钳制到 100，实际 %.2f", got.Score)
 	}
-	if s := ComputeRiskScore(RiskInput{AI: -1}); s < 0 {
+	if got := ComputeRiskScore(RiskInput{AI: ptr(-1)}).Score; got < 0 {
-		t.Errorf("应钳制到 0，实际 %.2f", s)
+		t.Errorf("应钳制到 0，实际 %.2f", got)
+	}
+}
+
+func TestRiskConfidenceLowMediumUnknown(t *testing.T) {
+	if got := RiskConfidence(RiskInput{}); got != "unknown" {
+		t.Errorf("无 AI 应为 unknown，实际 %s", got)
+	}
+	if got := RiskConfidence(RiskInput{AI: ptr(0.2)}); got != "low" {
+		t.Errorf("低异常概率应为 low，实际 %s", got)
+	}
+	if got := RiskConfidence(RiskInput{AI: ptr(0.9)}); got != "medium" {
+		t.Errorf("高异常概率当前最多应为 medium，实际 %s", got)
+	}
+}
+
+func TestComputeRiskScoreBoundaryUsesWeightNormalization(t *testing.T) {
+	// 0.5*0.8 + 0.2*0.5 的可用权重为 0.7：0.5/0.7 = 71.43
+	got := ComputeRiskScore(RiskInput{AI: ptr(0.8), Env: ptr(0.5)})
+	want := 50.0 / 0.7
+	if math.Abs(got.Score-want) > 0.01 {
+		t.Errorf("归一化应得 %.2f，实际 %.2f", want, got.Score)
 	}
 }
 
@@ -63,19 +140,19 @@ func TestStageCoefficientMapping(t *testing.T) {
 
 func TestEnvCoefficientRules(t *testing.T) {
 	// 湿度 >=80 → 高（真菌病）
-	if c := EnvCoefficient(f(25), f(85)); c < 0.7 {
+	if c := EnvCoefficient(ptr(25), ptr(85)); c < 0.7 {
 		t.Errorf("湿度 85 应 ≥0.7，实际 %.2f", c)
 	}
 	// 湿度 75-80 → 中
-	if c := EnvCoefficient(f(25), f(78)); c < 0.3 {
+	if c := EnvCoefficient(ptr(25), ptr(78)); c < 0.3 {
 		t.Errorf("湿度 78 应 ≥0.3，实际 %.2f", c)
 	}
 	// 温度突变 >30 → 中（核型多角体病诱发）
-	if c := EnvCoefficient(f(32), f(60)); c < 0.3 {
+	if c := EnvCoefficient(ptr(32), ptr(60)); c < 0.3 {
 		t.Errorf("温度 32 应 ≥0.3，实际 %.2f", c)
 	}
 	// 舒适环境 → 0
-	if c := EnvCoefficient(f(25), f(60)); c != 0 {
+	if c := EnvCoefficient(ptr(25), ptr(60)); c != 0 {
 		t.Errorf("舒适环境应为 0，实际 %.2f", c)
 	}
 	// 缺数据 → 0

server-go/migrations/000002_risk_assessment.down.sql

@@ -0,0 +1,6 @@
+DROP INDEX IF EXISTS idx_inspection_records_mock_created_at;
+
+ALTER TABLE inspection_records
+    DROP COLUMN IF EXISTS is_mock,
+    DROP COLUMN IF EXISTS model_version,
+    DROP COLUMN IF EXISTS risk_assessment;

server-go/migrations/000002_risk_assessment.up.sql

@@ -0,0 +1,7 @@
+ALTER TABLE inspection_records
+    ADD COLUMN IF NOT EXISTS risk_assessment jsonb,
+    ADD COLUMN IF NOT EXISTS model_version varchar(64),
+    ADD COLUMN IF NOT EXISTS is_mock boolean NOT NULL DEFAULT false;
+
+CREATE INDEX IF NOT EXISTS idx_inspection_records_mock_created_at
+    ON inspection_records (is_mock, created_at);

web/src/dal/inspections.ts

@@ -15,6 +15,17 @@ export interface InspectionRecord {
   detections?: AIDetection[];
   riskScore?: number;
   riskLevel?: string;
+  riskAssessment?: {
+    score: number;
+    level: string;
+    confidence: string;
+    modelVersion: string;
+    ruleVersion: string;
+    components: Record<string, number | null>;
+    missing: string[];
+  };
+  modelVersion?: string;
+  isMock?: boolean;
   aiStatus: string;
   createdAt?: string;
 }

web/src/pages/Inspections.tsx

@@ -60,7 +60,17 @@ export default function InspectionsPage() {
       width: 80,
       render: (_, r) => (r.imageUrl ? <Image src={r.imageUrl} width={48} height={48} style={{ objectFit: 'cover', borderRadius: 4 }} /> : '-'),
     },
-    { title: '状态', dataIndex: 'aiStatus', search: false, render: (_, r) => (r.aiStatus === 'done' ? <Tag color="green">成功</Tag> : <Tag color="red">失败</Tag>) },
+    {
+      title: '状态',
+      dataIndex: 'aiStatus',
+      search: false,
+      render: (_, r) => (
+        <>
+          {r.aiStatus === 'done' ? <Tag color="green">成功</Tag> : <Tag color="red">失败</Tag>}
+          {r.isMock ? <Tag color="orange">Mock</Tag> : null}
+        </>
+      ),
+    },
     {
       title: '操作',
       valueType: 'option',
@@ -101,6 +111,8 @@ export default function InspectionsPage() {
               <br />
               状态：{detail.aiStatus === 'done' ? '检测成功' : '检测失败'}
               <br />
+              数据：{detail.isMock ? '联调 Mock，不计入生产统计' : detail.modelVersion || '未知模型'}
+              <br />
               风险分：{detail.riskScore !== undefined ? Math.round(detail.riskScore) : '-'}（
               {detail.riskLevel ? riskLevelLabel(detail.riskLevel) : '-'}）
             </Typography.Paragraph>

后续工作计划.md

@@ -1,6 +1,6 @@
 # 后续工作计划
 
-> **完成状态（2026-08-13 更新）**：#5-#24、#27 已完成（详见 `开发交接记录.md`）；#1-4 因物理机问题挂起；#23/#26 骨架完成；微信/天气真实数据待凭证。
+> **完成状态（2026-08-14 更新）**：#5-#24、#27 已完成，Task 0/1/2/4/5/6 整改代码完成（详见 `开发交接记录.md`）；#1-4 因物理机问题挂起；#23/#26 骨架完成；微信/天气真实数据待凭证。
 
 ## 整改实施计划 Wave 0-4（2026-08-13 启动）
 
@@ -14,7 +14,7 @@
 | Wave 1 | P0 安全与正确性 | Task 3 移除默认密钥与默认管理员密码 | 延后到最后（跳过） | 用户 2026-08-13 明确要求跳过并留到最后 |
 | Wave 1 | P0 安全与正确性 | Task 4 收口视频访问与摄像头密钥输出 | 部分可用 | 待开发服务器部署联调 |
 | Wave 1 | P0 安全与正确性 | Task 5 修复 WebSocket 越权与 AI 流 SSRF | 部分可用 | 待开发服务器部署与真实 WS/AI 联调 |
-| Wave 1 | P0 安全与正确性 | Task 6 修复 AI 风险语义并隔离 Mock 数据 | 未开始 | 无 |
+| Wave 1 | P0 安全与正确性 | Task 6 修复 AI 风险语义并隔离 Mock 数据 | 部分可用 | 待开发服务器迁移部署与真实模型接入；历史数据待人工审阅 |
 | Wave 1 | P0 安全与正确性 | Task 7 修订 qPCR 判读与检测质控 | 未开始 | 需领域专家确认 |
 | Wave 2 | 工程可靠性 | Task 8 建立可靠通知、吊销与跨实例状态 | 未开始 | 无 |
 | Wave 2 | 工程可靠性 | Task 9 建立统一检测任务、样本链与发病事件 | 未开始 | 无 |

开发交接记录.md

@@ -921,3 +921,33 @@ MVP 沿用 IoTDB（现状）；TDengine 作为生产规模化候选（先基准
 
 - 本任务前分支提交为 `1a29def`；回滚可还原 Task 5 提交；
 - WS/AI 改动无数据库 schema 变更；若已部署，恢复旧二进制并重启即可，但需同步回滚客户端 WS 连接方式。
+
+## 2026-08-14 整改 Task 6：修复 AI 风险语义并隔离 Mock 数据
+
+### 做了什么
+
+- AI 服务 `/detect` 响应新增 `modelVersion`、`isMock`、`status`、`abnormalProbability`；`healthy` 高置信度不再贡献异常概率，空检测或 `unknown` 返回 `unknown`，不自动当健康；
+- Go 风险评分改为只消费 AI 异常概率：`RiskInput` 四个组件全部可空，缺失项不填 0；可用权重归一化后输出 `RiskAssessment`（score/level/confidence/modelVersion/ruleVersion/components/missing）；
+- 巡检记录新增 `risk_assessment`、`model_version`、`is_mock` 字段和 `000002_risk_assessment` 迁移；健康画像统计默认排除 mock，Mock 结果不触发微信告警；
+- `APP_ENV=production` 时收到 `isMock=true` 按失败记录并输出 error 日志，避免把联调数据当作生产检测结论；
+- Web/小程序巡检页显示 Mock 标识；交叉验证中空检测/`unknown` 不再被当作 healthy；
+- 新增只读历史评估 SQL `scripts/risk_historical_review.sql`，统计历史 healthy 高置信度记录与风险分布，未执行批量重算。
+
+### 设计思路与决策依据
+
+- 规格书 AI-INS-002 要求只用 `abnormalProbability` 计算 AI 风险，因此修复了原先取最大 confidence 导致 healthy=0.95 也能得 47.5 分的问题；
+- RISK-002 要求缺失数据不能以 0 冒充正常，所以环境/阶段/整齐度改为指针输入，缺失项进入 `missing` 并在 JSON 中返回 null；
+- 权重沿用 0.5/0.2/0.15/0.15 并标记为 `risk-v2-2026.08.14` 待试点校准规则；`confidence` 只输出 low/medium/unknown，不声称未校准结论为 high；
+- `000007_inspection_idempotency` 未另建迁移，因为 `000001_baseline` 已包含 `idempotency_key` 唯一索引；本次只补风险语义相关字段。
+- 历史旧记录没有 `is_mock` 标识，不能自动判别是否来自 mock；只读 SQL 报告用于人工审阅，未批量重算或改写旧数据。
+
+### 验证结果
+
+- `scripts/verify.ps1` exit 0：Go test/vet/build、Web test/lint/build、小程序 typecheck/build、APP typecheck/lint、AI pytest 15/15 均通过；
+- 新增测试覆盖 healthy/sick/unknown 异常概率、缺失组件归一化、风险分级边界、AI 响应解析、空检测 unknown、handler 风险输入和迁移版本；
+- 未部署开发服务器，未对现有库执行 `000002` 迁移；历史只读 SQL 报告未执行。
+
+### 回滚点
+
+- 本任务前分支提交为 `839ba91`；回滚可还原 Task 6 提交；
+- 数据库回滚执行 `migrate -path ... -database ... down 1` 或手工执行 `000002_risk_assessment.down.sql`，可移除新增三列和索引；`risk_score/risk_level` 仍保留，历史数据不回写。