Merge pull request 'fix:更新自建仓库' (#13) from damer into main

Reviewed-on: #13

.cursor/commands/myspec.code.md

@@ -2,3 +2,4 @@
 客户端请按照标准的RN架构目录写代码
 后端请按照标准的python FastAPI 架构目录写代码
 现在多语言仅支持 EN / TC
+整个task.md执行完毕后需要在对应的overview.md标记，并且说明变更的文件名

.cursor/commands/myspec.split-spec.md

@@ -9,6 +9,6 @@
    - 输入/输出定义
    - 验收标准（可验证）
 3. 拆分后输出一个 `modules/` 目录结构列表，并为每个模块生成对应 spec 内容。
-4. 保留大 spec.md 的高层背景/总览到 overview 部分。
+4. 保留大 spec.md 的高层背景/总览到 overview 部分，并标明各个模块的实现顺序。
 5. 子模块之间按逻辑关系关联。
 6. 不生成 plan.md 或 tasks.md，仅拆出子模块 spec。

.cursor/commands/myspec.task.md

@@ -2,3 +2,4 @@
 根据对应的plan.md 生成task.md
 任务清单详细可执行 
 执行完要标记
+整个task.md执行完毕后需要在对应的overview.md标记

.cursor/commands/myspec.test.md

@@ -0,0 +1 @@
+使用测试工具完成集成测试，并给我一份简单的测试报告

.cursor/rules/rule.mdc

@@ -28,4 +28,5 @@ modules/ 可嵌套 modules/，每层都独立规范。
 输出时根据这个结构生成内容时，请保持文件职责清晰。
 简短记录项目的该层每个spec的内容 ，每次编码完成后更新overview.md
 可以通过nvm 切换node版本
+在对数据库操作中，禁止执行破坏性操作，如果必须请让我同意，并回复：允许操作数据库
 

.gitea/workflows/README.md

@@ -0,0 +1,5 @@
+docker exec -it gitea-runner bash
+# 然后在容器里安装 Node.js
+curl -fsSL https://deb.nodesource.com/setup_20.x | bash -
+apt-get install -y nodejs
+node -v

.gitea/workflows/server-build.yml

@@ -0,0 +1,116 @@
+name: Build and Push Server Docker Image
+
+# 手动触发 workflow：从哪个分支运行，就打包哪个分支的代码
+on:
+  workflow_dispatch:
+
+jobs:
+  build-and-push:
+    runs-on: ubuntu-latest
+
+    steps:
+      # 1️⃣ Checkout 仓库代码
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          # 需要能 push tag（请在仓库 Secrets 配置 RUNNER_TOKEN）
+          token: ${{ secrets.RUNNER_TOKEN }}
+          persist-credentials: true
+
+      # 2️⃣ 自动递增 tag 并推送回 Gitea 仓库（默认按 vX.Y.Z 的 patch +1）
+      - name: Auto bump tag and push to repository
+        shell: bash
+        run: |
+          set -euo pipefail
+
+          # 配置提交信息（用于创建注释 tag）
+          git config user.name "gitea-actions"
+          git config user.email "actions@local"
+
+          # 确保本地有最新 tags
+          git fetch --tags --force
+
+          # 取最新的 semver tag（vX.Y.Z），按版本号排序
+          LATEST_TAG="$(git tag --list 'v*' --sort=-v:refname | head -n 1 || true)"
+          echo "LATEST_TAG=${LATEST_TAG}"
+
+          if [[ -z "${LATEST_TAG}" ]]; then
+            NEXT_TAG="v1.0.0"
+          else
+            if [[ "${LATEST_TAG}" =~ ^v([0-9]+)\.([0-9]+)\.([0-9]+)$ ]]; then
+              MAJOR="${BASH_REMATCH[1]}"
+              MINOR="${BASH_REMATCH[2]}"
+              PATCH="${BASH_REMATCH[3]}"
+              NEXT_TAG="v${MAJOR}.${MINOR}.$((PATCH + 1))"
+            else
+              # 如果最新 tag 不符合 vX.Y.Z，回退到 v1.0.0，避免误解析
+              NEXT_TAG="v1.0.0"
+            fi
+          fi
+
+          echo "NEXT_TAG=${NEXT_TAG}"
+
+          # 如果 tag 已存在则直接复用（避免重复运行失败）
+          if git rev-parse -q --verify "refs/tags/${NEXT_TAG}" >/dev/null; then
+            echo "Tag ${NEXT_TAG} 已存在，跳过创建。"
+          else
+            git tag -a "${NEXT_TAG}" -m "Release ${NEXT_TAG}"
+            git push origin "${NEXT_TAG}"
+          fi
+
+          # 输出给后续步骤使用
+          if [[ -n "${GITHUB_ENV:-}" ]]; then
+            echo "IMAGE_TAG=${NEXT_TAG}" >> "$GITHUB_ENV"
+          fi
+          # 兼容部分 Gitea Runner 环境变量命名
+          if [[ -n "${GITEA_ENV:-}" ]]; then
+            echo "IMAGE_TAG=${NEXT_TAG}" >> "$GITEA_ENV"
+          fi
+
+      # 3️⃣ 设置镜像仓库与镜像名称（自建 Registry / Docker Hub 都可）
+      - name: Set image variables
+        shell: bash
+        run: |
+          # 自建镜像仓库地址（例如：docker.damer.fun 或 registry.example.com）
+          # 如要推送 Docker Hub：可把 REGISTRY 留空并把 IMAGE_REPO 设置为 yourname/xxx
+          REGISTRY="${DOCKER_REGISTRY:-docker.damer.fun}"
+          # 镜像仓库路径（不含 registry）
+          IMAGE_REPO="${DOCKER_IMAGE_REPO:-damer/mindfulness-server}"
+
+          if [[ -n "$REGISTRY" ]]; then
+            IMAGE_NAME="$REGISTRY/$IMAGE_REPO"
+          else
+            IMAGE_NAME="$IMAGE_REPO"
+          fi
+
+          if [[ -n "${GITHUB_ENV:-}" ]]; then
+            echo "REGISTRY=$REGISTRY" >> "$GITHUB_ENV"
+            echo "IMAGE_REPO=$IMAGE_REPO" >> "$GITHUB_ENV"
+            echo "IMAGE_NAME=$IMAGE_NAME" >> "$GITHUB_ENV"
+          fi
+          if [[ -n "${GITEA_ENV:-}" ]]; then
+            echo "REGISTRY=$REGISTRY" >> "$GITEA_ENV"
+            echo "IMAGE_REPO=$IMAGE_REPO" >> "$GITEA_ENV"
+            echo "IMAGE_NAME=$IMAGE_NAME" >> "$GITEA_ENV"
+          fi
+
+      # 4️⃣ 登录镜像仓库（自建 Registry）
+      - name: Login to Docker Registry
+        uses: docker/login-action@v3
+        with:
+          registry: docker.damer.fun
+          username: ${{ secrets.DOCKER_USERNAME }}
+          password: ${{ secrets.DOCKER_TOKEN }}
+
+      # 5️⃣ 构建 Docker 镜像（使用 server/ 作为构建上下文）
+      - name: Build Docker Image
+        shell: bash
+        run: |
+          docker build -f server/Dockerfile -t "$IMAGE_NAME:$IMAGE_TAG" server
+
+      # 6️⃣ 推送 Docker 镜像到镜像仓库
+      - name: Push Docker Image
+        shell: bash
+        run: |
+          docker push "$IMAGE_NAME:$IMAGE_TAG"

.gitignore

@@ -4,6 +4,10 @@
 .DS_Store
 *.pem
 
+# Python（运行产物）
+__pycache__/
+*.py[cod]
+
 # Node / JS
 node_modules/
 npm-debug.*

client/app.json

@@ -15,7 +15,7 @@
     },
     "ios": {
       "supportsTablet": true,
-      "bundleIdentifier": "com.anonymous.client"
+      "bundleIdentifier": "com.damer.mindfulness"
     },
     "android": {
       "adaptiveIcon": {

client/app/(app)/home.tsx

@@ -14,12 +14,20 @@ import Animated, {
 import { MOCK_CONTENT } from '@/src/constants/mockContent';
 import {
   addFavorite,
+  getRecoFeedCache,
+  getRecoFeedHistory,
   getThemeMode,
   getUserProfile,
+  getUserProfileScoring,
+  recordRecoFeedServed,
+  recordRecoFeedTouched,
+  setRecoFeedCache,
   setReaction,
   setThemeMode,
+  type RecoFeedCacheItem,
   type ThemeMode,
 } from '@/src/storage/appStorage';
+import { fetchRecoFeed } from '@/src/services/recoApi';
 
 import ProfileModal from '@/components/home/ProfileModal';
 import ThemeModal from '@/components/home/ThemeModal';
@@ -75,8 +83,11 @@ export default function HomeScreen() {
   const [profileName, setProfileName] = useState<string | undefined>(undefined);
   const [busy, setBusy] = useState(false);
   const [likeFilled, setLikeFilled] = useState(false);
+  const [feedItems, setFeedItems] = useState<Array<{ content_id: number; text: string }>>([]);
 
-  const item = useMemo(() => MOCK_CONTENT[index % MOCK_CONTENT.length], [index]);
+  const currentList = feedItems.length > 0 ? feedItems : MOCK_CONTENT;
+  const item = useMemo(() => currentList[index % currentList.length], [currentList, index]);
+  const currentContentId = typeof (item as any)?.content_id === 'number' ? Number((item as any).content_id) : null;
 
   // 动画相关 Shared Values
   const translateY = useSharedValue(0);
@@ -100,6 +111,55 @@ export default function HomeScreen() {
     }, [])
   );
 
+  // 首次进入：先读缓存，再拉后端 feed（失败则保持 mock/缓存）
+  useEffect(() => {
+    let cancelled = false;
+    (async () => {
+      const cache = await getRecoFeedCache();
+      if (!cancelled && cache?.items?.length) {
+        setFeedItems(cache.items.map((x: RecoFeedCacheItem) => ({ content_id: x.content_id, text: x.text })));
+      }
+
+      const scoring = await getUserProfileScoring();
+      if (!scoring) return;
+
+      try {
+        const hist = await getRecoFeedHistory();
+        const out = await fetchRecoFeed({
+          k: 30,
+          user_profile: {
+            profile_version: scoring.profile_version,
+            profile_source: scoring.profile_source,
+            profile_generated_at: scoring.profile_generated_at,
+            profile_confidence: scoring.profile_confidence,
+            profile_answered: scoring.profile_answered,
+            stage: scoring.stage,
+            emotion_score: scoring.emotion_score,
+            context: scoring.context,
+            need: scoring.need,
+          },
+          already_recommended_ids: hist.already_recommended_ids,
+          touched_or_viewed_ids: hist.touched_or_viewed_ids,
+        });
+
+        if (!cancelled && out.items?.length) {
+          setFeedItems(out.items.map((x) => ({ content_id: x.content_id, text: x.text })));
+          await setRecoFeedCache({
+            saved_at: new Date().toISOString(),
+            items: out.items.map((x) => ({ content_id: x.content_id, text: x.text })),
+            meta: out.meta as Record<string, unknown>,
+          });
+          await recordRecoFeedServed(out.items.map((x) => x.content_id));
+        }
+      } catch {
+        // 忽略：保持缓存/本地 mock
+      }
+    })();
+    return () => {
+      cancelled = true;
+    };
+  }, []);
+
   const backgroundColor = useMemo(() => {
     if (themeMode === 'color') {
       const colorIndex = Math.floor(index / 10) % THEME_COLORS.length;
@@ -153,6 +213,11 @@ export default function HomeScreen() {
     if (busy) return;
     setBusy(true);
 
+    // 记录“看过/划过”的内容 id（用于下一次向后端请求时去重/频控）
+    if (typeof currentContentId === 'number') {
+      void recordRecoFeedTouched(currentContentId);
+    }
+
     // 1. 当前文案向上移动并消失
     translateY.value = withTiming(-40, { duration: 300, easing: Easing.out(Easing.quad) });
     opacity.value = withTiming(0, { duration: 300 }, (finished) => {
@@ -173,7 +238,7 @@ export default function HomeScreen() {
         });
       }
     });
-  }, [busy, index, translateY, opacity]);
+  }, [busy, currentContentId, index, translateY, opacity]);
 
   const lastTapRef = useRef<number>(0);
   

client/app/(onboarding)/onboarding.tsx

@@ -5,7 +5,16 @@ import { OnboardingLayout } from '@/components/onboarding/OnboardingLayout';
 import { NameInputStep } from '@/components/onboarding/NameInputStep';
 import { SelectionStep } from '@/components/onboarding/SelectionStep';
 import { ReminderStep } from '@/components/onboarding/ReminderStep';
-import { setOnboardingCompleted, setUserProfile, setDailyReminderSettings } from '@/src/storage/appStorage';
+import { buildUserProfileFromQuestionnaire, mapOnboardingSelectionsToQuestionnaireAnswers } from '@/src/features/userProfileScoring';
+import { fetchRecoFeed } from '@/src/services/recoApi';
+import {
+  recordRecoFeedServed,
+  setOnboardingCompleted,
+  setUserProfile,
+  setDailyReminderSettings,
+  setUserProfileScoring,
+  setRecoFeedCache,
+} from '@/src/storage/appStorage';
 
 const STEPS = [
   { id: 'name', type: 'name', title: '我可以怎么称呼你？' },
@@ -71,6 +80,40 @@ export default function OnboardingScreen() {
     const { status } = await Notifications.requestPermissionsAsync();
     const pushEnabled = status === 'granted';
 
+    // 将 Onboarding 选择映射为标准问卷枚举（允许跳过）
+    const answers = mapOnboardingSelectionsToQuestionnaireAnswers(selections);
+
+    // 生成用户画像（供推荐/Push/Widget 复用）
+    const scoringProfile = buildUserProfileFromQuestionnaire(answers);
+    await setUserProfileScoring(scoringProfile);
+
+    // Onboarding 结束后预拉取一次 Feed 文案（失败不阻塞进入首页）
+    try {
+      const { items, meta } = await fetchRecoFeed({
+        k: 30,
+        user_profile: {
+          profile_version: scoringProfile.profile_version,
+          profile_source: scoringProfile.profile_source,
+          profile_generated_at: scoringProfile.profile_generated_at,
+          profile_confidence: scoringProfile.profile_confidence,
+          profile_answered: scoringProfile.profile_answered,
+          stage: scoringProfile.stage,
+          emotion_score: scoringProfile.emotion_score,
+          context: scoringProfile.context,
+          need: scoringProfile.need,
+        },
+      });
+
+      await setRecoFeedCache({
+        saved_at: new Date().toISOString(),
+        items: items.map((x) => ({ content_id: x.content_id, text: x.text })),
+        meta: meta as Record<string, unknown>,
+      });
+      await recordRecoFeedServed(items.map((x) => x.content_id));
+    } catch {
+      // 网络失败时使用首页本地 mock 兜底
+    }
+
     await setUserProfile({ 
       name, 
       intents: Object.values(selections).flat() 
@@ -98,20 +141,29 @@ export default function OnboardingScreen() {
   };
 
   const onSkip = async () => {
+    // 跳过整个 Onboarding：仍生成一个“全跳过”的最小画像，保证下游可用
+    const scoringProfile = buildUserProfileFromQuestionnaire({});
+    await setUserProfileScoring(scoringProfile);
+
+    // 标记已完成，避免下次启动再次进入 Onboarding
     await setOnboardingCompleted(true);
     router.replace('/(app)/home');
   };
 
+  // 题目为单选：再次点击可取消；选择其他选项会替换为唯一选项
   const handleToggleSelection = (id: string) => {
     setSelections(prev => {
       const currentIds = prev[currentStep.id] || [];
-      const nextIds = currentIds.includes(id)
+      const nextIds = currentIds.includes(id) ? [] : [id];
-        ? currentIds.filter(i => i !== id)
-        : [...currentIds, id];
       return { ...prev, [currentStep.id]: nextIds };
     });
   };
 
+  const handleSkipStep = () => {
+    setSelections((prev) => ({ ...prev, [currentStep.id]: [] }));
+    onNext();
+  };
+
   return (
     <OnboardingLayout
       title={currentStep.title}
@@ -135,6 +187,7 @@ export default function OnboardingScreen() {
           selectedIds={selections[currentStep.id] || []}
           onToggle={handleToggleSelection}
           onNext={onNext}
+          onSkip={handleSkipStep}
         />
       )}
 

client/app/index.tsx

@@ -1,9 +1,8 @@
 import { useEffect } from 'react';
 import { ActivityIndicator, StyleSheet, View } from 'react-native';
 import { useRouter } from 'expo-router';
-import AsyncStorage from '@react-native-async-storage/async-storage';
 
-import { getOnboardingCompleted, getConsentAccepted, setOnboardingCompleted, setConsentAccepted } from '@/src/storage/appStorage';
+import { getOnboardingCompleted, getConsentAccepted } from '@/src/storage/appStorage';
 
 /**
  * 启动分发：根据 consent 和 onboarding 状态跳转
@@ -14,10 +13,6 @@ export default function Index() {
   useEffect(() => {
     let cancelled = false;
     (async () => {
-      // 【临时清除数据】：用于测试完整流程
-      await AsyncStorage.clear();
-      console.log('AsyncStorage has been cleared for testing.');
-
       // 1. 检查是否同意协议
       const consentAccepted = await getConsentAccepted();
       if (cancelled) return;
@@ -54,4 +49,3 @@ export default function Index() {
 const styles = StyleSheet.create({
   container: { flex: 1, alignItems: 'center', justifyContent: 'center' },
 });
-

client/components/onboarding/SelectionStep.tsx

@@ -18,9 +18,10 @@ interface SelectionStepProps {
   selectedIds: string[];
   onToggle: (id: string) => void;
   onNext: () => void;
+  onSkip?: () => void;
 }
 
-export function SelectionStep({ options, selectedIds, onToggle, onNext }: SelectionStepProps) {
+export function SelectionStep({ options, selectedIds, onToggle, onNext, onSkip }: SelectionStepProps) {
   const hasSelection = selectedIds.length > 0;
 
   return (
@@ -48,15 +49,19 @@ export function SelectionStep({ options, selectedIds, onToggle, onNext }: Select
 
       {/* 底部按钮：距离底部 12% 高度 */}
       <View style={styles.footer}>
-        <TouchableOpacity 
+        <View style={styles.footerRow}>
-          onPress={onNext} 
+          {onSkip && (
-          disabled={!hasSelection}
+            <TouchableOpacity onPress={onSkip} activeOpacity={0.8} style={styles.skipBtn}>
-          activeOpacity={0.8}
+              <SerifText style={styles.skipText}>跳过</SerifText>
-        >
+            </TouchableOpacity>
+          )}
+
+          <TouchableOpacity onPress={onNext} disabled={!hasSelection} activeOpacity={0.8}>
             {hasSelection ? <BtnClicked width={87} height={57} /> : <BtnNotClicked width={87} height={57} />}
           </TouchableOpacity>
         </View>
       </View>
+    </View>
   );
 }
 
@@ -99,5 +104,20 @@ const styles = StyleSheet.create({
     left: 0,
     right: 0,
     alignItems: 'center',
-  }
+  },
+  footerRow: {
+    flexDirection: 'row',
+    alignItems: 'center',
+    gap: 16,
+  },
+  skipBtn: {
+    paddingVertical: 10,
+    paddingHorizontal: 14,
+    borderRadius: 12,
+    backgroundColor: 'rgba(0,0,0,0.04)',
+  },
+  skipText: {
+    fontSize: 16,
+    color: OnboardingColors.textMuted,
+  },
 });

client/ios/Podfile

@@ -59,5 +59,13 @@ target 'client' do
       :mac_catalyst_enabled => false,
       :ccache_enabled => ccache_enabled?(podfile_properties),
     )
+
+    # 生成并随归档产物携带 dSYM（用于崩溃符号化与 Upload Symbols Failed 修复）
+    installer.pods_project.targets.each do |target|
+      target.build_configurations.each do |build_config|
+        build_config.build_settings['DEBUG_INFORMATION_FORMAT'] = 'dwarf-with-dsym'
+        build_config.build_settings['DWARF_DSYM_FILE_SHOULD_ACCOMPANY_PRODUCT'] = 'YES'
+      end
+    end
   end
 end

client/ios/Podfile.lock

@@ -3,9 +3,6 @@ PODS:
     - ExpoModulesCore
   - EXConstants (18.0.13):
     - ExpoModulesCore
-  - EXJSONUtils (0.15.0)
-  - EXManifests (1.0.10):
-    - ExpoModulesCore
   - EXNotifications (0.32.16):
     - ExpoModulesCore
   - Expo (54.0.32):
@@ -33,177 +30,6 @@ PODS:
     - ReactCommon/turbomodule/core
     - ReactNativeDependencies
     - Yoga
-  - expo-dev-client (6.0.20):
-    - EXManifests
-    - expo-dev-launcher
-    - expo-dev-menu
-    - expo-dev-menu-interface
-    - EXUpdatesInterface
-  - expo-dev-launcher (6.0.20):
-    - EXManifests
-    - expo-dev-launcher/Main (= 6.0.20)
-    - expo-dev-menu
-    - expo-dev-menu-interface
-    - ExpoModulesCore
-    - EXUpdatesInterface
-    - hermes-engine
-    - RCTRequired
-    - RCTTypeSafety
-    - React-Core
-    - React-Core-prebuilt
-    - React-debug
-    - React-Fabric
-    - React-featureflags
-    - React-graphics
-    - React-ImageManager
-    - React-jsi
-    - React-jsinspector
-    - React-NativeModulesApple
-    - React-RCTAppDelegate
-    - React-RCTFabric
-    - React-renderercss
-    - React-rendererdebug
-    - React-utils
-    - ReactAppDependencyProvider
-    - ReactCodegen
-    - ReactCommon/turbomodule/bridging
-    - ReactCommon/turbomodule/core
-    - ReactNativeDependencies
-    - Yoga
-  - expo-dev-launcher/Main (6.0.20):
-    - EXManifests
-    - expo-dev-launcher/Unsafe
-    - expo-dev-menu
-    - expo-dev-menu-interface
-    - ExpoModulesCore
-    - EXUpdatesInterface
-    - hermes-engine
-    - RCTRequired
-    - RCTTypeSafety
-    - React-Core
-    - React-Core-prebuilt
-    - React-debug
-    - React-Fabric
-    - React-featureflags
-    - React-graphics
-    - React-ImageManager
-    - React-jsi
-    - React-jsinspector
-    - React-NativeModulesApple
-    - React-RCTAppDelegate
-    - React-RCTFabric
-    - React-renderercss
-    - React-rendererdebug
-    - React-utils
-    - ReactAppDependencyProvider
-    - ReactCodegen
-    - ReactCommon/turbomodule/bridging
-    - ReactCommon/turbomodule/core
-    - ReactNativeDependencies
-    - Yoga
-  - expo-dev-launcher/Unsafe (6.0.20):
-    - EXManifests
-    - expo-dev-menu
-    - expo-dev-menu-interface
-    - ExpoModulesCore
-    - EXUpdatesInterface
-    - hermes-engine
-    - RCTRequired
-    - RCTTypeSafety
-    - React-Core
-    - React-Core-prebuilt
-    - React-debug
-    - React-Fabric
-    - React-featureflags
-    - React-graphics
-    - React-ImageManager
-    - React-jsi
-    - React-jsinspector
-    - React-NativeModulesApple
-    - React-RCTAppDelegate
-    - React-RCTFabric
-    - React-renderercss
-    - React-rendererdebug
-    - React-utils
-    - ReactAppDependencyProvider
-    - ReactCodegen
-    - ReactCommon/turbomodule/bridging
-    - ReactCommon/turbomodule/core
-    - ReactNativeDependencies
-    - Yoga
-  - expo-dev-menu (7.0.18):
-    - expo-dev-menu/Main (= 7.0.18)
-    - expo-dev-menu/ReactNativeCompatibles (= 7.0.18)
-    - hermes-engine
-    - RCTRequired
-    - RCTTypeSafety
-    - React-Core
-    - React-Core-prebuilt
-    - React-debug
-    - React-Fabric
-    - React-featureflags
-    - React-graphics
-    - React-ImageManager
-    - React-jsi
-    - React-NativeModulesApple
-    - React-RCTFabric
-    - React-renderercss
-    - React-rendererdebug
-    - React-utils
-    - ReactCodegen
-    - ReactCommon/turbomodule/bridging
-    - ReactCommon/turbomodule/core
-    - ReactNativeDependencies
-    - Yoga
-  - expo-dev-menu-interface (2.0.0)
-  - expo-dev-menu/Main (7.0.18):
-    - EXManifests
-    - expo-dev-menu-interface
-    - ExpoModulesCore
-    - hermes-engine
-    - RCTRequired
-    - RCTTypeSafety
-    - React-Core
-    - React-Core-prebuilt
-    - React-debug
-    - React-Fabric
-    - React-featureflags
-    - React-graphics
-    - React-ImageManager
-    - React-jsi
-    - React-jsinspector
-    - React-NativeModulesApple
-    - React-RCTFabric
-    - React-renderercss
-    - React-rendererdebug
-    - React-utils
-    - ReactCodegen
-    - ReactCommon/turbomodule/bridging
-    - ReactCommon/turbomodule/core
-    - ReactNativeDependencies
-    - Yoga
-  - expo-dev-menu/ReactNativeCompatibles (7.0.18):
-    - hermes-engine
-    - RCTRequired
-    - RCTTypeSafety
-    - React-Core
-    - React-Core-prebuilt
-    - React-debug
-    - React-Fabric
-    - React-featureflags
-    - React-graphics
-    - React-ImageManager
-    - React-jsi
-    - React-NativeModulesApple
-    - React-RCTFabric
-    - React-renderercss
-    - React-rendererdebug
-    - React-utils
-    - ReactCodegen
-    - ReactCommon/turbomodule/bridging
-    - ReactCommon/turbomodule/core
-    - ReactNativeDependencies
-    - Yoga
   - ExpoAsset (12.0.12):
     - ExpoModulesCore
   - ExpoFileSystem (19.0.21):
@@ -248,8 +74,6 @@ PODS:
     - ExpoModulesCore
   - ExpoWebBrowser (15.0.10):
     - ExpoModulesCore
-  - EXUpdatesInterface (2.0.0):
-    - ExpoModulesCore
   - FBLazyVector (0.81.5)
   - hermes-engine (0.81.5):
     - hermes-engine/Pre-built (= 0.81.5)
@@ -1974,28 +1798,6 @@ PODS:
     - ReactCommon/turbomodule/core
     - ReactNativeDependencies
     - Yoga
-  - RNGestureHandler (2.30.0):
-    - hermes-engine
-    - RCTRequired
-    - RCTTypeSafety
-    - React-Core
-    - React-Core-prebuilt
-    - React-debug
-    - React-Fabric
-    - React-featureflags
-    - React-graphics
-    - React-ImageManager
-    - React-jsi
-    - React-NativeModulesApple
-    - React-RCTFabric
-    - React-renderercss
-    - React-rendererdebug
-    - React-utils
-    - ReactCodegen
-    - ReactCommon/turbomodule/bridging
-    - ReactCommon/turbomodule/core
-    - ReactNativeDependencies
-    - Yoga
   - RNReanimated (4.1.6):
     - hermes-engine
     - RCTRequired
@@ -2238,26 +2040,19 @@ PODS:
 DEPENDENCIES:
   - "EXApplication (from `../node_modules/.pnpm/expo-application@7.0.8_expo@54.0.32/node_modules/expo-application/ios`)"
   - "EXConstants (from `../node_modules/.pnpm/expo-constants@18.0.13_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0_/node_modules/expo-constants/ios`)"
-  - "EXJSONUtils (from `../node_modules/.pnpm/expo-json-utils@0.15.0/node_modules/expo-json-utils/ios`)"
+  - "EXNotifications (from `../node_modules/.pnpm/expo-notifications@0.32.16_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@1_nvlvke5tn7wk5pigfsu7j4ieeq/node_modules/expo-notifications/ios`)"
-  - "EXManifests (from `../node_modules/.pnpm/expo-manifests@1.0.10_expo@54.0.32/node_modules/expo-manifests/ios`)"
+  - "Expo (from `../node_modules/.pnpm/expo@54.0.32_@babel+core@7.28.6_@expo+metro-runtime@6.1.2_expo-router@6.0.22_react-native@0.8_7rhpxisdkrzvrgzbu7ct455kta/node_modules/expo`)"
-  - "EXNotifications (from `../node_modules/.pnpm/expo-notifications@0.32.16_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+r_758952db70529f49bda448def1c13c49/node_modules/expo-notifications/ios`)"
-  - "Expo (from `../node_modules/.pnpm/expo@54.0.32_@babel+core@7.28.6_@expo+metro-runtime@6.1.2_expo-router@6.0.22_react-nati_18ad48ba284ee86e6eb1cb0f939697b0/node_modules/expo`)"
-  - "expo-dev-client (from `../node_modules/.pnpm/expo-dev-client@6.0.20_expo@54.0.32/node_modules/expo-dev-client/ios`)"
-  - "expo-dev-launcher (from `../node_modules/.pnpm/expo-dev-launcher@6.0.20_expo@54.0.32/node_modules/expo-dev-launcher`)"
-  - "expo-dev-menu (from `../node_modules/.pnpm/expo-dev-menu@7.0.18_expo@54.0.32/node_modules/expo-dev-menu`)"
-  - "expo-dev-menu-interface (from `../node_modules/.pnpm/expo-dev-menu-interface@2.0.0_expo@54.0.32/node_modules/expo-dev-menu-interface/ios`)"
   - "ExpoAsset (from `../node_modules/.pnpm/expo-asset@12.0.12_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0__react@19.1.0/node_modules/expo-asset/ios`)"
   - "ExpoFileSystem (from `../node_modules/.pnpm/expo-file-system@19.0.21_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0_/node_modules/expo-file-system/ios`)"
   - "ExpoFont (from `../node_modules/.pnpm/expo-font@14.0.11_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0__react@19.1.0/node_modules/expo-font/ios`)"
-  - "ExpoHead (from `../node_modules/.pnpm/expo-router@6.0.22_@expo+metro-runtime@6.1.2_@types+react@19.1.17_expo-constants@18.0.1_bd9aa16746ed7110429f931eb008e6d2/node_modules/expo-router/ios`)"
+  - "ExpoHead (from `../node_modules/.pnpm/expo-router@6.0.22_@expo+metro-runtime@6.1.2_@types+react@19.1.17_expo-constants@18.0.13_expo_rjurfbyy5kjn57nkkfxix5iqea/node_modules/expo-router/ios`)"
   - "ExpoKeepAwake (from `../node_modules/.pnpm/expo-keep-awake@15.0.8_expo@54.0.32_react@19.1.0/node_modules/expo-keep-awake/ios`)"
-  - "ExpoLinearGradient (from `../node_modules/.pnpm/expo-linear-gradient@15.0.8_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+_53aef72480df9baa4504f4743d9c64bb/node_modules/expo-linear-gradient/ios`)"
+  - "ExpoLinearGradient (from `../node_modules/.pnpm/expo-linear-gradient@15.0.8_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@_e6k2hjkd5k4lph2ersbp3gfshy/node_modules/expo-linear-gradient/ios`)"
   - "ExpoLinking (from `../node_modules/.pnpm/expo-linking@8.0.11_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0__react@19.1.0/node_modules/expo-linking/ios`)"
   - "ExpoLocalization (from `../node_modules/.pnpm/expo-localization@17.0.8_expo@54.0.32_react@19.1.0/node_modules/expo-localization/ios`)"
   - "ExpoModulesCore (from `../node_modules/.pnpm/expo-modules-core@3.0.29_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0__react@19.1.0/node_modules/expo-modules-core`)"
   - "ExpoSplashScreen (from `../node_modules/.pnpm/expo-splash-screen@31.0.13_expo@54.0.32/node_modules/expo-splash-screen/ios`)"
   - "ExpoWebBrowser (from `../node_modules/.pnpm/expo-web-browser@15.0.10_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0_/node_modules/expo-web-browser/ios`)"
-  - "EXUpdatesInterface (from `../node_modules/.pnpm/expo-updates-interface@2.0.0_expo@54.0.32/node_modules/expo-updates-interface/ios`)"
   - "FBLazyVector (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/Libraries/FBLazyVector`)"
   - "hermes-engine (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/sdks/hermes-engine/hermes-engine.podspec`)"
   - "RCTDeprecation (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactApple/Libraries/RCTFoundation/RCTDeprecation`)"
@@ -2294,7 +2089,7 @@ DEPENDENCIES:
   - "React-logger (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactCommon/logger`)"
   - "React-Mapbuffer (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactCommon`)"
   - "React-microtasksnativemodule (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactCommon/react/nativemodule/microtasks`)"
-  - "react-native-safe-area-context (from `../node_modules/.pnpm/react-native-safe-area-context@5.6.2_react-native@0.81.5_@babel+core@7.28.6_@types+reac_14122a3aa345cfabcc022a0f638ef16d/node_modules/react-native-safe-area-context`)"
+  - "react-native-safe-area-context (from `../node_modules/.pnpm/react-native-safe-area-context@5.6.2_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1_azuxgonsvxb2yngtegtuvyxcpi/node_modules/react-native-safe-area-context`)"
   - "React-NativeModulesApple (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactCommon/react/nativemodule/core/platform/ios`)"
   - "React-oscompat (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactCommon/oscompat`)"
   - "React-perflogger (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactCommon/reactperflogger`)"
@@ -2326,12 +2121,11 @@ DEPENDENCIES:
   - ReactCodegen (from `build/generated/ios`)
   - "ReactCommon/turbomodule/core (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactCommon`)"
   - "ReactNativeDependencies (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/third-party-podspecs/ReactNativeDependencies.podspec`)"
-  - "RNCAsyncStorage (from `../node_modules/.pnpm/@react-native-async-storage+async-storage@2.2.0_react-native@0.81.5_@babel+core@7.28.6__ce3c4972004f3d6791573ec5b64bee38/node_modules/@react-native-async-storage/async-storage`)"
+  - "RNCAsyncStorage (from `../node_modules/.pnpm/@react-native-async-storage+async-storage@2.2.0_react-native@0.81.5_@babel+core@7.28.6_@types_fp4qq3a7mejmut52v6jrlvxlzi/node_modules/@react-native-async-storage/async-storage`)"
-  - "RNGestureHandler (from `../node_modules/.pnpm/react-native-gesture-handler@2.30.0_react-native@0.81.5_@babel+core@7.28.6_@types+react_39cf7da47c9c8531caaa923ee740e293/node_modules/react-native-gesture-handler`)"
+  - "RNReanimated (from `../node_modules/.pnpm/react-native-reanimated@4.1.6_@babel+core@7.28.6_react-native-worklets@0.5.1_@babel+core@7.28_ky3sbxf6i7nkyacc2hzg3xcz4q/node_modules/react-native-reanimated`)"
-  - "RNReanimated (from `../node_modules/.pnpm/react-native-reanimated@4.1.6_@babel+core@7.28.6_react-native-worklets@0.5.1_@babel+cor_c7c888bd389fb93c9cfe2d3c1c8b0777/node_modules/react-native-reanimated`)"
   - "RNScreens (from `../node_modules/.pnpm/react-native-screens@4.16.0_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0__react@19.1.0/node_modules/react-native-screens`)"
   - "RNSVG (from `../node_modules/.pnpm/react-native-svg@15.12.1_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0__react@19.1.0/node_modules/react-native-svg`)"
-  - "RNWorklets (from `../node_modules/.pnpm/react-native-worklets@0.5.1_@babel+core@7.28.6_react-native@0.81.5_@babel+core@7.28.6_@_40f69326ce21d3f9f6d74d3965fd9adf/node_modules/react-native-worklets`)"
+  - "RNWorklets (from `../node_modules/.pnpm/react-native-worklets@0.5.1_@babel+core@7.28.6_react-native@0.81.5_@babel+core@7.28.6_@types+_5atwepuw3zy3crkgvetf35tkve/node_modules/react-native-worklets`)"
   - "Yoga (from `../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactCommon/yoga`)"
 
 EXTERNAL SOURCES:
@@ -2339,22 +2133,10 @@ EXTERNAL SOURCES:
     :path: "../node_modules/.pnpm/expo-application@7.0.8_expo@54.0.32/node_modules/expo-application/ios"
   EXConstants:
     :path: "../node_modules/.pnpm/expo-constants@18.0.13_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0_/node_modules/expo-constants/ios"
-  EXJSONUtils:
-    :path: "../node_modules/.pnpm/expo-json-utils@0.15.0/node_modules/expo-json-utils/ios"
-  EXManifests:
-    :path: "../node_modules/.pnpm/expo-manifests@1.0.10_expo@54.0.32/node_modules/expo-manifests/ios"
   EXNotifications:
-    :path: "../node_modules/.pnpm/expo-notifications@0.32.16_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+r_758952db70529f49bda448def1c13c49/node_modules/expo-notifications/ios"
+    :path: "../node_modules/.pnpm/expo-notifications@0.32.16_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@1_nvlvke5tn7wk5pigfsu7j4ieeq/node_modules/expo-notifications/ios"
   Expo:
-    :path: "../node_modules/.pnpm/expo@54.0.32_@babel+core@7.28.6_@expo+metro-runtime@6.1.2_expo-router@6.0.22_react-nati_18ad48ba284ee86e6eb1cb0f939697b0/node_modules/expo"
+    :path: "../node_modules/.pnpm/expo@54.0.32_@babel+core@7.28.6_@expo+metro-runtime@6.1.2_expo-router@6.0.22_react-native@0.8_7rhpxisdkrzvrgzbu7ct455kta/node_modules/expo"
-  expo-dev-client:
-    :path: "../node_modules/.pnpm/expo-dev-client@6.0.20_expo@54.0.32/node_modules/expo-dev-client/ios"
-  expo-dev-launcher:
-    :path: "../node_modules/.pnpm/expo-dev-launcher@6.0.20_expo@54.0.32/node_modules/expo-dev-launcher"
-  expo-dev-menu:
-    :path: "../node_modules/.pnpm/expo-dev-menu@7.0.18_expo@54.0.32/node_modules/expo-dev-menu"
-  expo-dev-menu-interface:
-    :path: "../node_modules/.pnpm/expo-dev-menu-interface@2.0.0_expo@54.0.32/node_modules/expo-dev-menu-interface/ios"
   ExpoAsset:
     :path: "../node_modules/.pnpm/expo-asset@12.0.12_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0__react@19.1.0/node_modules/expo-asset/ios"
   ExpoFileSystem:
@@ -2362,11 +2144,11 @@ EXTERNAL SOURCES:
   ExpoFont:
     :path: "../node_modules/.pnpm/expo-font@14.0.11_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0__react@19.1.0/node_modules/expo-font/ios"
   ExpoHead:
-    :path: "../node_modules/.pnpm/expo-router@6.0.22_@expo+metro-runtime@6.1.2_@types+react@19.1.17_expo-constants@18.0.1_bd9aa16746ed7110429f931eb008e6d2/node_modules/expo-router/ios"
+    :path: "../node_modules/.pnpm/expo-router@6.0.22_@expo+metro-runtime@6.1.2_@types+react@19.1.17_expo-constants@18.0.13_expo_rjurfbyy5kjn57nkkfxix5iqea/node_modules/expo-router/ios"
   ExpoKeepAwake:
     :path: "../node_modules/.pnpm/expo-keep-awake@15.0.8_expo@54.0.32_react@19.1.0/node_modules/expo-keep-awake/ios"
   ExpoLinearGradient:
-    :path: "../node_modules/.pnpm/expo-linear-gradient@15.0.8_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+_53aef72480df9baa4504f4743d9c64bb/node_modules/expo-linear-gradient/ios"
+    :path: "../node_modules/.pnpm/expo-linear-gradient@15.0.8_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@_e6k2hjkd5k4lph2ersbp3gfshy/node_modules/expo-linear-gradient/ios"
   ExpoLinking:
     :path: "../node_modules/.pnpm/expo-linking@8.0.11_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0__react@19.1.0/node_modules/expo-linking/ios"
   ExpoLocalization:
@@ -2377,8 +2159,6 @@ EXTERNAL SOURCES:
     :path: "../node_modules/.pnpm/expo-splash-screen@31.0.13_expo@54.0.32/node_modules/expo-splash-screen/ios"
   ExpoWebBrowser:
     :path: "../node_modules/.pnpm/expo-web-browser@15.0.10_expo@54.0.32_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0_/node_modules/expo-web-browser/ios"
-  EXUpdatesInterface:
-    :path: "../node_modules/.pnpm/expo-updates-interface@2.0.0_expo@54.0.32/node_modules/expo-updates-interface/ios"
   FBLazyVector:
     :path: "../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/Libraries/FBLazyVector"
   hermes-engine:
@@ -2451,7 +2231,7 @@ EXTERNAL SOURCES:
   React-microtasksnativemodule:
     :path: "../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactCommon/react/nativemodule/microtasks"
   react-native-safe-area-context:
-    :path: "../node_modules/.pnpm/react-native-safe-area-context@5.6.2_react-native@0.81.5_@babel+core@7.28.6_@types+reac_14122a3aa345cfabcc022a0f638ef16d/node_modules/react-native-safe-area-context"
+    :path: "../node_modules/.pnpm/react-native-safe-area-context@5.6.2_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1_azuxgonsvxb2yngtegtuvyxcpi/node_modules/react-native-safe-area-context"
   React-NativeModulesApple:
     :path: "../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactCommon/react/nativemodule/core/platform/ios"
   React-oscompat:
@@ -2515,43 +2295,34 @@ EXTERNAL SOURCES:
   ReactNativeDependencies:
     :podspec: "../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/third-party-podspecs/ReactNativeDependencies.podspec"
   RNCAsyncStorage:
-    :path: "../node_modules/.pnpm/@react-native-async-storage+async-storage@2.2.0_react-native@0.81.5_@babel+core@7.28.6__ce3c4972004f3d6791573ec5b64bee38/node_modules/@react-native-async-storage/async-storage"
+    :path: "../node_modules/.pnpm/@react-native-async-storage+async-storage@2.2.0_react-native@0.81.5_@babel+core@7.28.6_@types_fp4qq3a7mejmut52v6jrlvxlzi/node_modules/@react-native-async-storage/async-storage"
-  RNGestureHandler:
-    :path: "../node_modules/.pnpm/react-native-gesture-handler@2.30.0_react-native@0.81.5_@babel+core@7.28.6_@types+react_39cf7da47c9c8531caaa923ee740e293/node_modules/react-native-gesture-handler"
   RNReanimated:
-    :path: "../node_modules/.pnpm/react-native-reanimated@4.1.6_@babel+core@7.28.6_react-native-worklets@0.5.1_@babel+cor_c7c888bd389fb93c9cfe2d3c1c8b0777/node_modules/react-native-reanimated"
+    :path: "../node_modules/.pnpm/react-native-reanimated@4.1.6_@babel+core@7.28.6_react-native-worklets@0.5.1_@babel+core@7.28_ky3sbxf6i7nkyacc2hzg3xcz4q/node_modules/react-native-reanimated"
   RNScreens:
     :path: "../node_modules/.pnpm/react-native-screens@4.16.0_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0__react@19.1.0/node_modules/react-native-screens"
   RNSVG:
     :path: "../node_modules/.pnpm/react-native-svg@15.12.1_react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0__react@19.1.0/node_modules/react-native-svg"
   RNWorklets:
-    :path: "../node_modules/.pnpm/react-native-worklets@0.5.1_@babel+core@7.28.6_react-native@0.81.5_@babel+core@7.28.6_@_40f69326ce21d3f9f6d74d3965fd9adf/node_modules/react-native-worklets"
+    :path: "../node_modules/.pnpm/react-native-worklets@0.5.1_@babel+core@7.28.6_react-native@0.81.5_@babel+core@7.28.6_@types+_5atwepuw3zy3crkgvetf35tkve/node_modules/react-native-worklets"
   Yoga:
     :path: "../node_modules/.pnpm/react-native@0.81.5_@babel+core@7.28.6_@types+react@19.1.17_react@19.1.0/node_modules/react-native/ReactCommon/yoga"
 
 SPEC CHECKSUMS:
-  EXApplication: 1e98d4b1dccdf30627f92917f4b2c5a53c330e5f
+  EXApplication: 13420f8139864183f8a04fd6099077bdf8cfb186
-  EXConstants: fce59a631a06c4151602843667f7cfe35f81e271
+  EXConstants: 3feb66fd1d94202fc1f0946d74e029d8b224b60e
-  EXJSONUtils: 1d3e4590438c3ee593684186007028a14b3686cd
+  EXNotifications: 2a3feb7af6194828d9aafda72f63a9a03866230a
-  EXManifests: a8d97683e5c7a3b026ffbd58559c64dc655b747b
+  Expo: b8d64eb9a496ebe8c71e3dae7eeb7f394b146b80
-  EXNotifications: 9eec98712cc814ceff916d876cb53859003b0597
+  ExpoAsset: d999f3bbd998a750f3b74cb913229848901b926b
-  Expo: 4e503a041c59c4e34c8be262a135848ad5cd3710
+  ExpoFileSystem: aefcd337b94b874f88752ebefc52813b84992fad
-  expo-dev-client: 425ee077d6754a98cfe3a2e2410d29b440b24c9d
+  ExpoFont: c625dbd97ed57e9089b172b2a7bb99003d074664
-  expo-dev-launcher: a4f4cdef064ab1fb8621e5b8c7c457cd6e9568c3
+  ExpoHead: b691a2ed7ab02ed820b6c6468941832d34969c29
-  expo-dev-menu: 05b18812110c175814c6af0d09dd658abcc5e00d
+  ExpoKeepAwake: 44bf6715bc1d2ddb17afe19d927cd039cda123f0
-  expo-dev-menu-interface: 600df12ea01efecdd822daaf13cc0ac091775533
+  ExpoLinearGradient: 814a21fc4056c3cf606e4f19e31e47074c5b5a86
-  ExpoAsset: f867e55ceb428aab99e1e8c082b5aee7c159ea18
+  ExpoLinking: ebf543fd411d56375cb4eee07f6ab4e31c7ad959
-  ExpoFileSystem: 858a44267a3e6e9057e0888ad7c7cfbf55d52063
+  ExpoLocalization: 6ac6f326210f0a3141ef6f58ab8f8f4ed003b485
-  ExpoFont: f543ce20a228dd702813668b1a07b46f51878d47
+  ExpoModulesCore: 77496909fd3c800f97f7f2007dd26aeac4bb3798
-  ExpoHead: 4425246bc93411f0fe7f6945f95f698e91db8780
+  ExpoSplashScreen: 72fbc6dd9d6404dd9d0725a56c9ac1383bc0b14f
-  ExpoKeepAwake: 55f75eca6499bb9e4231ebad6f3e9cb8f99c0296
+  ExpoWebBrowser: 88b116cd378d9609c776c0903fe4070fca461588
-  ExpoLinearGradient: 809102bdb979f590083af49f7fa4805cd931bd58
-  ExpoLinking: 8f0aaf69aa56f832913030503b6263dc6f647f37
-  ExpoLocalization: d9168d5300a5b03e5e78b986124d11fb6ec3ebbd
-  ExpoModulesCore: f3da4f1ab5a8375d0beafab763739dbee8446583
-  ExpoSplashScreen: bc3cffefca2716e5f22350ca109badd7e50ec14d
-  ExpoWebBrowser: 17b064c621789e41d4816c95c93f429b84971f52
-  EXUpdatesInterface: 5adf50cb41e079c861da6d9b4b954c3db9a50734
   FBLazyVector: e95a291ad2dadb88e42b06e0c5fb8262de53ec12
   hermes-engine: 9f4dfe93326146a1c99eb535b1cb0b857a3cd172
   RCTDeprecation: 943572d4be82d480a48f4884f670135ae30bf990
@@ -2559,74 +2330,73 @@ SPEC CHECKSUMS:
   RCTTypeSafety: 16a4144ca3f959583ab019b57d5633df10b5e97c
   React: 914f8695f9bf38e6418228c2ffb70021e559f92f
   React-callinvoker: 1c0808402aee0c6d4a0d8e7220ce6547af9fba71
-  React-Core: c61410ef0ca6055e204a963992e363227e0fd1c5
+  React-Core: 4ae98f9e8135b8ddbd7c98730afb6fdae883db90
-  React-Core-prebuilt: 02f0ad625ddd47463c009c2d0c5dd35c0d982599
+  React-Core-prebuilt: 8f4cca589c14e8cf8fc6db4587ef1c2056b5c151
-  React-CoreModules: 1f6d1744b5f9f2ec684a4bb5ced25370f87e5382
+  React-CoreModules: e878a90bb19b8f3851818af997dbae3b3b0a27ac
-  React-cxxreact: 3af79478e8187b63ffc22b794cd42d3fc1f1f2da
+  React-cxxreact: 28af9844f6dc87be1385ab521fbfb3746f19563c
   React-debug: 6328c2228e268846161f10082e80dc69eac2e90a
-  React-defaultsnativemodule: d635ef36d755321e5d6fc065bd166b2c5a0e9833
+  React-defaultsnativemodule: afc9d809ec75780f39464a6949c07987fbea488c
-  React-domnativemodule: dd28f6d96cd21236e020be2eff6fe0b7d4ec3b66
+  React-domnativemodule: 91a233260411d41f27f67aa1358b7f9f0bfd101d
-  React-Fabric: 2e32c3fdbb1fbcf5fde54607e3abe453c6652ce2
+  React-Fabric: 21f349b5e93f305a3c38c885902683a9c79cf983
-  React-FabricComponents: 5ed0cdb81f6b91656cb4d3be432feaa28a58071a
+  React-FabricComponents: 47ac634cc9ecc64b30a9997192f510eebe4177e4
-  React-FabricImage: 2bc714f818cb24e454f5d3961864373271b2faf8
+  React-FabricImage: 21873acd6d4a51a0b97c133141051c7acb11cc86
-  React-featureflags: 847642f41fa71ad4eec5e0351badebcad4fe6171
+  React-featureflags: 653f469f0c3c9dc271d610373e3b6e66a9fd847d
-  React-featureflagsnativemodule: c868a544b2c626fa337bcbd364b1befe749f0d3f
+  React-featureflagsnativemodule: c91a8a3880e0f4838286402241ead47db43aed28
-  React-graphics: 192ec701def5b3f2a07db2814dfba5a44986cff6
+  React-graphics: b4bdb0f635b8048c652a5d2b73eb8b1ddd950f24
-  React-hermes: e875778b496c86d07ab2ccaa36a9505d248a254b
+  React-hermes: fcfad3b917400f49026f3232561e039c9d1c34bf
-  React-idlecallbacksnativemodule: 4d57965cdf82c14ee3b337189836cd8491632b76
+  React-idlecallbacksnativemodule: 8cb83207e39f8179ac1d344b6177c6ab3ccebcdc
-  React-ImageManager: bd0b99e370b13de82c9cd15f0f08144ff3de079e
+  React-ImageManager: 396128004783fc510e629124dce682d38d1088e7
-  React-jserrorhandler: a2fdef4cbcfdcdf3fa9f5d1f7190f7fd4535248d
+  React-jserrorhandler: b58b788d788cdbf8bda7db74a88ebfcffc8a0795
-  React-jsi: 89d43d1e7d4d0663f8ba67e0b39eb4e4672c27de
+  React-jsi: d2c3f8555175371c02da6dfe7ed1b64b55a9d6c0
-  React-jsiexecutor: abe4874aaab90dfee5dec480680220b2f8af07e3
+  React-jsiexecutor: ba537434eb45ee018b590ed7d29ee233fddb8669
-  React-jsinspector: a0b3e051aef842b0b2be2353790ae2b2a5a65a8f
+  React-jsinspector: f21b6654baf96cb9f71748844a32468a5f73ad51
-  React-jsinspectorcdp: 6346013b2247c6263fbf5199adf4a8751e53bd89
+  React-jsinspectorcdp: 3f8be4830694c3c1c39442e50f8db877966d43f0
-  React-jsinspectornetwork: 26281aa50d49fc1ec93abf981d934698fa95714f
+  React-jsinspectornetwork: 70e41469565712ad60e11d9c8b8f999b9f7f61eb
-  React-jsinspectortracing: 55eedf6d57540507570259a778663b90060bbd6e
+  React-jsinspectortracing: eccf9bfa4ec7f130d514f215cfb2222dc3c0e270
-  React-jsitooling: 0e001113fa56d8498aa8ac28437ac0d36348e51a
+  React-jsitooling: b376a695f5a507627f7934748533b24eed1751ca
-  React-jsitracing: b713793eb8a5bbc4d86a84e9d9e5023c0f58cbaf
+  React-jsitracing: 5c8c3273dda2d95191cc0612fb5e71c4d9018d2a
-  React-logger: 50fdb9a8236da90c0b1072da5c32ee03aeb5bf28
+  React-logger: c3e2f8a2e284341205f61eef3d4677ab5a309dfd
-  React-Mapbuffer: 9050ee10c19f4f7fca8963d0211b2854d624973e
+  React-Mapbuffer: 603c18db65844bb81dbe62fee8fcc976eaeb7108
-  React-microtasksnativemodule: f775db9e991c6f3b8ccbc02bfcde22770f96e23b
+  React-microtasksnativemodule: d77e0c426fce34c23227394c96ca1033b30c813c
-  react-native-safe-area-context: 37e680fc4cace3c0030ee46e8987d24f5d3bdab2
+  react-native-safe-area-context: 53f796cb6c814661bbe99fbdfd0585d07b996cdd
-  React-NativeModulesApple: 8969913947d5b576de4ed371a939455a8daf28aa
+  React-NativeModulesApple: 1664340b8750d64e0ef3907c5e53d9481f74bcbd
   React-oscompat: ce47230ed20185e91de62d8c6d139ae61763d09c
-  React-perflogger: 02b010e665772c7dcb859d85d44c1bfc5ac7c0e4
+  React-perflogger: b1af3cfb3f095f819b2814910000392a8e17ba9f
-  React-performancetimeline: 130db956b5a83aa4fb41ddf5ae68da89f3fb1526
+  React-performancetimeline: f9ec65b77bcadbc7bd8b47a6f4b4b697da7b1490
   React-RCTActionSheet: 0b14875b3963e9124a5a29a45bd1b22df8803916
-  React-RCTAnimation: a7b90fd2af7bb9c084428867445a1481a8cb112e
+  React-RCTAnimation: 60f6eca214a62b9673f64db6df3830cee902b5af
-  React-RCTAppDelegate: 3262bedd01263f140ec62b7989f4355f57cec016
+  React-RCTAppDelegate: 37734b39bac108af30a0fd9d3e1149ec68b82c28
-  React-RCTBlob: c17531368702f1ebed5d0ada75a7cf5915072a53
+  React-RCTBlob: 83fbcbd57755caf021787324aac2fe9b028cc264
-  React-RCTFabric: 6409edd8cfdc3133b6cc75636d3b858fdb1d11ea
+  React-RCTFabric: a05cb1df484008db3753c8b4a71e4c6d9f1e43a6
-  React-RCTFBReactNativeSpec: c004b27b4fa3bd85878ad2cf53de3bbec85da797
+  React-RCTFBReactNativeSpec: d58d7ae9447020bbbac651e3b0674422aba18266
-  React-RCTImage: c68078a120d0123f4f07a5ac77bea3bb10242f32
+  React-RCTImage: 47aba3be7c6c64f956b7918ab933769602406aac
-  React-RCTLinking: cf8f9391fe7fe471f96da3a5f0435235eca18c5b
+  React-RCTLinking: 2dbaa4df2e4523f68baa07936bd8efdfa34d5f31
-  React-RCTNetwork: ca31f7c879355760c2d9832a06ee35f517938a20
+  React-RCTNetwork: 1fca7455f9dedf7de2b95bec438da06680f3b000
-  React-RCTRuntime: a6cf4a1e42754fc87f493e538f2ac6b820e45418
+  React-RCTRuntime: 17819dd1dfc8613efaf4cbb9d8686baae4a83e5b
-  React-RCTSettings: e0e140b2ff4bf86d34e9637f6316848fc00be035
+  React-RCTSettings: 01bf91c856862354d3d2f642ccb82f3697a4284a
-  React-RCTText: 75915bace6f7877c03a840cc7b6c622fb62bfa6b
+  React-RCTText: cb576a3797dcb64933613c522296a07eaafc0461
-  React-RCTVibration: 25f26b85e5e432bb3c256f8b384f9269e9529f25
+  React-RCTVibration: 560af8c086741f3525b8456a482cdbe27f9d098e
   React-rendererconsistency: 2dac03f448ff337235fd5820b10f81633328870d
-  React-renderercss: 477da167bb96b5ac86d30c5d295412fb853f5453
+  React-renderercss: c5c6b7a15948dd28facca39a18ac269073718490
-  React-rendererdebug: 2a1798c6f3ef5f22d466df24c33653edbabb5b89
+  React-rendererdebug: 3c9d5e1634273f5a24d84cc5669f290ce0bdc812
-  React-RuntimeApple: 28cf4d8eb18432f6a21abbed7d801ab7f6b6f0b4
+  React-RuntimeApple: 887637d1e12ea8262df7d32bc100467df2302613
-  React-RuntimeCore: 41bf0fd56a00de5660f222415af49879fa49c4f0
+  React-RuntimeCore: 91f779835dc4f8f84777fe5dd24f1a22f96454e4
-  React-runtimeexecutor: 1afb774dde3011348e8334be69d2f57a359ea43e
+  React-runtimeexecutor: 8bb6b738f37b0ada4a6269e6f8ab1133dea0285c
-  React-RuntimeHermes: f3b158ea40e8212b1a723a68b4315e7a495c5fc6
+  React-RuntimeHermes: 4cb93de9fa8b1cc753d200dbe61a01b9ec5f5562
-  React-runtimescheduler: 3e1e2bec7300bae512533107d8e54c6e5c63fe0f
+  React-runtimescheduler: 83dc28f530bfbd2fce84ed13aa7feebdc24e5af7
-  React-timing: 6fa9883de2e41791e5dc4ec404e5e37f3f50e801
+  React-timing: 03c7217455d2bff459b27a3811be25796b600f47
-  React-utils: 6e2035b53d087927768649a11a26c4e092448e34
+  React-utils: 6d46795ae0444ec8a5d9a5f201157b286bf5250a
-  ReactAppDependencyProvider: 1bcd3527ac0390a1c898c114f81ff954be35ed79
+  ReactAppDependencyProvider: c277c5b231881ad4f00cd59e3aa0671b99d7ebee
-  ReactCodegen: fffa79906f5866f6a5ab5d98480b375191190271
+  ReactCodegen: 88a1f4643f15841573f833b895bfa2a0c6cb4e7f
-  ReactCommon: 08810150b1206cc44aecf5f6ae19af32f29151a8
+  ReactCommon: e6e232202a447d353e5531f2be82f50f47cbaa9a
   ReactNativeDependencies: 71ce9c28beb282aa720ea7b46980fff9669f428a
-  RNCAsyncStorage: 3a4f5e2777dae1688b781a487923a08569e27fe4
+  RNCAsyncStorage: e85a99325df9eb0191a6ee2b2a842644c7eb29f4
-  RNGestureHandler: e0d0bce5599f6120b7adf90c38d2805e2935795f
+  RNReanimated: 10415bc8396eaeac0d7b2c9a1538eae7e607ec9c
-  RNReanimated: 9c6a550b41de91cf374e60afd79db93a362f1126
+  RNScreens: dd61bc3a3e6f6901ad833efa411917d44827cf51
-  RNScreens: d8d6f1792f6e7ac12b0190d33d8d390efc0c1845
+  RNSVG: 2825ee146e0f6a16221e852299943e4cceef4528
-  RNSVG: 31d6639663c249b7d5abc9728dde2041eb2a3c34
+  RNWorklets: 9ccdc8112b17af6eee2c85a233891cb80db150ad
-  RNWorklets: 1b50cb7595142f95e70518196ba247ad7f46a52e
   Yoga: 5934998fbeaef7845dbf698f698518695ab4cd1a
 
-PODFILE CHECKSUM: dfe3cc75dee014a0abd367bc9e1bdbab0ba64ee3
+PODFILE CHECKSUM: 4d5c52f9fa870c1d398cf59e37c149f66700c061
 
 COCOAPODS: 1.16.2

client/ios/client.xcodeproj/project.pbxproj

@@ -3,7 +3,7 @@
 	archiveVersion = 1;
 	classes = {
 	};
-	objectVersion = 56;
+	objectVersion = 77;
 	objects = {
 
 /* Begin PBXBuildFile section */
@@ -17,6 +17,8 @@
 		EB3DAF812F2A4B8E00450593 /* WidgetKit.framework in Frameworks */ = {isa = PBXBuildFile; fileRef = EB3DAF802F2A4B8D00450593 /* WidgetKit.framework */; };
 		EB3DAF832F2A4B8E00450593 /* SwiftUI.framework in Frameworks */ = {isa = PBXBuildFile; fileRef = EB3DAF822F2A4B8E00450593 /* SwiftUI.framework */; };
 		EB3DAF942F2A4B8F00450593 /* 情绪小组件Extension.appex in Embed Foundation Extensions */ = {isa = PBXBuildFile; fileRef = EB3DAF7F2F2A4B8D00450593 /* 情绪小组件Extension.appex */; settings = {ATTRIBUTES = (RemoveHeadersOnCopy, ); }; };
+		EB630B302F30BE2700DC761A /* Assets.xcassets in Resources */ = {isa = PBXBuildFile; fileRef = EB630B2F2F30BE2700DC761A /* Assets.xcassets */; };
+		EB630B312F30BE2700DC761A /* Assets.xcassets in Resources */ = {isa = PBXBuildFile; fileRef = EB630B2F2F30BE2700DC761A /* Assets.xcassets */; };
 		F11748422D0307B40044C1D9 /* AppDelegate.swift in Sources */ = {isa = PBXBuildFile; fileRef = F11748412D0307B40044C1D9 /* AppDelegate.swift */; };
 /* End PBXBuildFile section */
 
@@ -59,6 +61,7 @@
 		EB3DAF802F2A4B8D00450593 /* WidgetKit.framework */ = {isa = PBXFileReference; lastKnownFileType = wrapper.framework; name = WidgetKit.framework; path = System/Library/Frameworks/WidgetKit.framework; sourceTree = SDKROOT; };
 		EB3DAF822F2A4B8E00450593 /* SwiftUI.framework */ = {isa = PBXFileReference; lastKnownFileType = wrapper.framework; name = SwiftUI.framework; path = System/Library/Frameworks/SwiftUI.framework; sourceTree = SDKROOT; };
 		EB3DAF9A2F2A4D0900450593 /* MindfulnessWidget.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = MindfulnessWidget.swift; sourceTree = "<group>"; };
+		EB630B2F2F30BE2700DC761A /* Assets.xcassets */ = {isa = PBXFileReference; lastKnownFileType = folder.assetcatalog; path = Assets.xcassets; sourceTree = "<group>"; };
 		ED297162215061F000B7C4FE /* JavaScriptCore.framework */ = {isa = PBXFileReference; lastKnownFileType = wrapper.framework; name = JavaScriptCore.framework; path = System/Library/Frameworks/JavaScriptCore.framework; sourceTree = SDKROOT; };
 		F11748412D0307B40044C1D9 /* AppDelegate.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; name = AppDelegate.swift; path = client/AppDelegate.swift; sourceTree = "<group>"; };
 		F11748442D0722820044C1D9 /* client-Bridging-Header.h */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.c.h; name = "client-Bridging-Header.h"; path = "client/client-Bridging-Header.h"; sourceTree = "<group>"; };
@@ -156,6 +159,7 @@
 		83CBB9F61A601CBA00E9B192 = {
 			isa = PBXGroup;
 			children = (
+				EB630B2F2F30BE2700DC761A /* Assets.xcassets */,
 				13B07FAE1A68108700A75B9A /* client */,
 				832341AE1AAA6A7D00B99B32 /* Libraries */,
 				EB3DAF842F2A4B8E00450593 /* 情绪小组件 */,
@@ -305,6 +309,7 @@
 				13B07FBF1A68108700A75B9A /* Images.xcassets in Resources */,
 				3E461D99554A48A4959DE609 /* SplashScreen.storyboard in Resources */,
 				0BE245B56A79D95AB0A7B4BA /* PrivacyInfo.xcprivacy in Resources */,
+				EB630B312F30BE2700DC761A /* Assets.xcassets in Resources */,
 			);
 			runOnlyForDeploymentPostprocessing = 0;
 		};
@@ -312,6 +317,7 @@
 			isa = PBXResourcesBuildPhase;
 			buildActionMask = 2147483647;
 			files = (
+				EB630B302F30BE2700DC761A /* Assets.xcassets in Resources */,
 			);
 			runOnlyForDeploymentPostprocessing = 0;
 		};
@@ -374,8 +380,6 @@
 				"${PODS_CONFIGURATION_BUILD_DIR}/RNSVG/RNSVGFilters.bundle",
 				"${PODS_CONFIGURATION_BUILD_DIR}/React-Core/React-Core_privacy.bundle",
 				"${PODS_CONFIGURATION_BUILD_DIR}/React-cxxreact/React-cxxreact_privacy.bundle",
-				"${PODS_CONFIGURATION_BUILD_DIR}/expo-dev-launcher/EXDevLauncher.bundle",
-				"${PODS_CONFIGURATION_BUILD_DIR}/expo-dev-menu/EXDevMenu.bundle",
 			);
 			name = "[CP] Copy Pods Resources";
 			outputPaths = (
@@ -389,8 +393,6 @@
 				"${TARGET_BUILD_DIR}/${UNLOCALIZED_RESOURCES_FOLDER_PATH}/RNSVGFilters.bundle",
 				"${TARGET_BUILD_DIR}/${UNLOCALIZED_RESOURCES_FOLDER_PATH}/React-Core_privacy.bundle",
 				"${TARGET_BUILD_DIR}/${UNLOCALIZED_RESOURCES_FOLDER_PATH}/React-cxxreact_privacy.bundle",
-				"${TARGET_BUILD_DIR}/${UNLOCALIZED_RESOURCES_FOLDER_PATH}/EXDevLauncher.bundle",
-				"${TARGET_BUILD_DIR}/${UNLOCALIZED_RESOURCES_FOLDER_PATH}/EXDevMenu.bundle",
 			);
 			runOnlyForDeploymentPostprocessing = 0;
 			shellPath = /bin/sh;
@@ -479,9 +481,11 @@
 			baseConfigurationReference = FFF632A94C7A551AAA096858 /* Pods-client.debug.xcconfig */;
 			buildSettings = {
 				ASSETCATALOG_COMPILER_APPICON_NAME = AppIcon;
+				ASSETCATALOG_COMPILER_INCLUDE_ALL_APPICON_ASSETS = YES;
 				CLANG_ENABLE_MODULES = YES;
 				CODE_SIGN_ENTITLEMENTS = client/client.entitlements;
-				CURRENT_PROJECT_VERSION = 1;
+				CURRENT_PROJECT_VERSION = 2;
+				DEVELOPMENT_TEAM = WS92GPX9H2;
 				ENABLE_BITCODE = NO;
 				GCC_PREPROCESSOR_DEFINITIONS = (
 					"$(inherited)",
@@ -493,14 +497,14 @@
 					"$(inherited)",
 					"@executable_path/Frameworks",
 				);
-				MARKETING_VERSION = 1.0;
+				MARKETING_VERSION = 1.0.0;
 				OTHER_LDFLAGS = (
 					"$(inherited)",
 					"-ObjC",
 					"-lc++",
 				);
 				OTHER_SWIFT_FLAGS = "$(inherited) -D EXPO_CONFIGURATION_DEBUG";
-				PRODUCT_BUNDLE_IDENTIFIER = com.anonymous.client;
+				PRODUCT_BUNDLE_IDENTIFIER = com.damer.mindfulness;
 				PRODUCT_NAME = client;
 				SUPPORTED_PLATFORMS = "iphoneos iphonesimulator";
 				SUPPORTS_MACCATALYST = NO;
@@ -509,7 +513,7 @@
 				SWIFT_OBJC_BRIDGING_HEADER = "client/client-Bridging-Header.h";
 				SWIFT_OPTIMIZATION_LEVEL = "-Onone";
 				SWIFT_VERSION = 5.0;
-				TARGETED_DEVICE_FAMILY = "1,2";
+				TARGETED_DEVICE_FAMILY = 1;
 				VERSIONING_SYSTEM = "apple-generic";
 			};
 			name = Debug;
@@ -519,23 +523,27 @@
 			baseConfigurationReference = 3C76CA16D0801CBF0D731C7C /* Pods-client.release.xcconfig */;
 			buildSettings = {
 				ASSETCATALOG_COMPILER_APPICON_NAME = AppIcon;
+				ASSETCATALOG_COMPILER_INCLUDE_ALL_APPICON_ASSETS = YES;
 				CLANG_ENABLE_MODULES = YES;
 				CODE_SIGN_ENTITLEMENTS = client/client.entitlements;
-				CURRENT_PROJECT_VERSION = 1;
+				CURRENT_PROJECT_VERSION = 2;
+				DEBUG_INFORMATION_FORMAT = "dwarf-with-dsym";
+				DEVELOPMENT_TEAM = WS92GPX9H2;
+				DWARF_DSYM_FILE_SHOULD_ACCOMPANY_PRODUCT = YES;
 				INFOPLIST_FILE = client/Info.plist;
 				IPHONEOS_DEPLOYMENT_TARGET = 15.1;
 				LD_RUNPATH_SEARCH_PATHS = (
 					"$(inherited)",
 					"@executable_path/Frameworks",
 				);
-				MARKETING_VERSION = 1.0;
+				MARKETING_VERSION = 1.0.0;
 				OTHER_LDFLAGS = (
 					"$(inherited)",
 					"-ObjC",
 					"-lc++",
 				);
 				OTHER_SWIFT_FLAGS = "$(inherited) -D EXPO_CONFIGURATION_RELEASE";
-				PRODUCT_BUNDLE_IDENTIFIER = com.anonymous.client;
+				PRODUCT_BUNDLE_IDENTIFIER = com.damer.mindfulness;
 				PRODUCT_NAME = client;
 				SUPPORTED_PLATFORMS = "iphoneos iphonesimulator";
 				SUPPORTS_MACCATALYST = NO;
@@ -543,7 +551,7 @@
 				SUPPORTS_XR_DESIGNED_FOR_IPHONE_IPAD = NO;
 				SWIFT_OBJC_BRIDGING_HEADER = "client/client-Bridging-Header.h";
 				SWIFT_VERSION = 5.0;
-				TARGETED_DEVICE_FAMILY = "1,2";
+				TARGETED_DEVICE_FAMILY = 1;
 				VERSIONING_SYSTEM = "apple-generic";
 			};
 			name = Release;
@@ -680,8 +688,9 @@
 				CLANG_WARN_QUOTED_INCLUDE_IN_FRAMEWORK_HEADER = YES;
 				CLANG_WARN_UNGUARDED_AVAILABILITY = YES_AGGRESSIVE;
 				CODE_SIGN_STYLE = Automatic;
-				CURRENT_PROJECT_VERSION = 1;
+				CURRENT_PROJECT_VERSION = 2;
 				DEBUG_INFORMATION_FORMAT = dwarf;
+				DEVELOPMENT_TEAM = WS92GPX9H2;
 				ENABLE_USER_SCRIPT_SANDBOXING = YES;
 				GCC_C_LANGUAGE_STANDARD = gnu17;
 				GENERATE_INFOPLIST_FILE = YES;
@@ -699,7 +708,7 @@
 				MTL_ENABLE_DEBUG_INFO = INCLUDE_SOURCE;
 				MTL_FAST_MATH = YES;
 				OTHER_SWIFT_FLAGS = "$(inherited) -D EXPO_CONFIGURATION_DEBUG";
-				PRODUCT_BUNDLE_IDENTIFIER = com.anonymous.client.emotionwidget;
+				PRODUCT_BUNDLE_IDENTIFIER = com.damer.mindfulness.emotionwidget;
 				PRODUCT_NAME = "$(TARGET_NAME)";
 				SKIP_INSTALL = YES;
 				STRING_CATALOG_GENERATE_SYMBOLS = YES;
@@ -713,7 +722,7 @@
 				SWIFT_OPTIMIZATION_LEVEL = "-Onone";
 				SWIFT_UPCOMING_FEATURE_MEMBER_IMPORT_VISIBILITY = YES;
 				SWIFT_VERSION = 5.0;
-				TARGETED_DEVICE_FAMILY = "1,2";
+				TARGETED_DEVICE_FAMILY = 1;
 			};
 			name = Debug;
 		};
@@ -732,8 +741,9 @@
 				CLANG_WARN_UNGUARDED_AVAILABILITY = YES_AGGRESSIVE;
 				CODE_SIGN_STYLE = Automatic;
 				COPY_PHASE_STRIP = NO;
-				CURRENT_PROJECT_VERSION = 1;
+				CURRENT_PROJECT_VERSION = 2;
 				DEBUG_INFORMATION_FORMAT = "dwarf-with-dsym";
+				DEVELOPMENT_TEAM = WS92GPX9H2;
 				ENABLE_USER_SCRIPT_SANDBOXING = YES;
 				GCC_C_LANGUAGE_STANDARD = gnu17;
 				GENERATE_INFOPLIST_FILE = YES;
@@ -750,7 +760,7 @@
 				MARKETING_VERSION = 1.0.0;
 				MTL_FAST_MATH = YES;
 				OTHER_SWIFT_FLAGS = "$(inherited) -D EXPO_CONFIGURATION_RELEASE";
-				PRODUCT_BUNDLE_IDENTIFIER = com.anonymous.client.emotionwidget;
+				PRODUCT_BUNDLE_IDENTIFIER = com.damer.mindfulness.emotionwidget;
 				PRODUCT_NAME = "$(TARGET_NAME)";
 				SKIP_INSTALL = YES;
 				STRING_CATALOG_GENERATE_SYMBOLS = YES;
@@ -763,7 +773,7 @@
 				SWIFT_EMIT_LOC_STRINGS = YES;
 				SWIFT_UPCOMING_FEATURE_MEMBER_IMPORT_VISIBILITY = YES;
 				SWIFT_VERSION = 5.0;
-				TARGETED_DEVICE_FAMILY = "1,2";
+				TARGETED_DEVICE_FAMILY = 1;
 			};
 			name = Release;
 		};

client/ios/client/Info.plist

@@ -19,7 +19,7 @@
 	<key>CFBundlePackageType</key>
 	<string>$(PRODUCT_BUNDLE_PACKAGE_TYPE)</string>
 	<key>CFBundleShortVersionString</key>
-    <string>1.0.0</string>
+	<string>$(MARKETING_VERSION)</string>
 	<key>CFBundleSignature</key>
 	<string>????</string>
 	<key>CFBundleURLTypes</key>
@@ -28,12 +28,12 @@
 			<key>CFBundleURLSchemes</key>
 			<array>
 				<string>client</string>
-          <string>com.anonymous.client</string>
+				<string>$(PRODUCT_BUNDLE_IDENTIFIER)</string>
 			</array>
 		</dict>
 	</array>
 	<key>CFBundleVersion</key>
-    <string>1</string>
+	<string>$(CURRENT_PROJECT_VERSION)</string>
 	<key>LSMinimumSystemVersion</key>
 	<string>12.0</string>
 	<key>LSRequiresIPhoneOS</key>
@@ -52,7 +52,7 @@
 	<key>RCTNewArchEnabled</key>
 	<true/>
 	<key>UILaunchStoryboardName</key>
-    <string>SplashScreen</string>
+	<string></string>
 	<key>UIRequiredDeviceCapabilities</key>
 	<array>
 		<string>arm64</string>

client/ios/client/client.entitlements

@@ -3,6 +3,6 @@
 <plist version="1.0">
   <dict>
     <key>aps-environment</key>
-    <string>development</string>
+    <string>production</string>
   </dict>
 </plist>

client/package.json

@@ -6,7 +6,8 @@
     "start": "expo start",
     "android": "expo run:android",
     "ios": "expo run:ios",
-    "web": "expo start --web"
+    "web": "expo start --web",
+    "test": "vitest run"
   },
   "dependencies": {
     "@expo/vector-icons": "^15.0.3",
@@ -41,7 +42,8 @@
   "devDependencies": {
     "@types/react": "~19.1.0",
     "react-test-renderer": "19.1.0",
-    "typescript": "~5.9.2"
+    "typescript": "~5.9.2",
+    "vitest": "^4.0.18"
   },
   "private": true
 }

client/src/constants/env.ts

@@ -20,14 +20,31 @@ function getOptionalEnv(name: string, fallback: string): string {
   return process.env[name] ?? fallback;
 }
 
-export const APP_ENV = (getOptionalEnv('EXPO_PUBLIC_ENV', 'dev') as AppEnv) ?? 'dev';
+export type AppRuntimeEnv = 'local' | 'dev' | 'prod';
 
-export const API_BASE_URL = getRequiredEnv('EXPO_PUBLIC_API_BASE_URL');
+export const APP_ENV = (getOptionalEnv('EXPO_PUBLIC_ENV', 'local') as AppRuntimeEnv) ?? 'local';
+
+function getApiBaseUrl(env: AppRuntimeEnv): string {
+  // 向后兼容：若直接提供了 EXPO_PUBLIC_API_BASE_URL，则优先使用（不再强制要求 *_DEV/_PROD）
+  const direct = process.env.EXPO_PUBLIC_API_BASE_URL;
+  if (direct && String(direct).trim()) return String(direct).trim();
+
+  // 约定：local/dev/prod 三套域名分别配置， 便于后续直接切环境而不改代码
+  if (env === 'local') {
+    return getOptionalEnv('EXPO_PUBLIC_API_BASE_URL_LOCAL', 'http://localhost:8000');
+  }
+  if (env === 'dev') {
+    return getOptionalEnv('EXPO_PUBLIC_API_BASE_URL_DEV', getOptionalEnv('EXPO_PUBLIC_API_BASE_URL_LOCAL', 'http://localhost:8000'));
+  }
+  return getOptionalEnv('EXPO_PUBLIC_API_BASE_URL_PROD', getOptionalEnv('EXPO_PUBLIC_API_BASE_URL_LOCAL', 'http://localhost:8000'));
+}
+
+export const API_BASE_URL = getApiBaseUrl(APP_ENV);
 
 /**
  * 默认语言策略：
- * - auto：优先设备语言（支持列表内时），否则回退 zh-CN
+ * - auto：优先设备语言（支持列表内时），否则回退 en
- * - zh-CN/en/es/pt/zh-TW：固定默认语言（仍允许用户在设置中手动切换并持久化）
+ * - en/zh-TW：固定默认语言（仍允许用户在设置中手动切换并持久化）
  */
 export const DEFAULT_LANGUAGE = getOptionalEnv('EXPO_PUBLIC_DEFAULT_LANGUAGE', 'auto');
 

client/src/features/userProfileScoring/__tests__/onboarding-integration.test.ts

@@ -0,0 +1,51 @@
+import { describe, expect, it } from 'vitest';
+
+import { buildUserProfileFromQuestionnaire } from '../index';
+import { mapOnboardingSelectionsToQuestionnaireAnswers } from '../onboardingMapping';
+
+describe('Onboarding → UserProfileScoring 集成', () => {
+  it('完整作答：Onboarding 选择能正确映射并生成画像', () => {
+    const selections = {
+      status: ['pregnant'],
+      emotion: ['calm'],
+      influence: ['work'],
+      support: ['balance'],
+    };
+
+    const answers = mapOnboardingSelectionsToQuestionnaireAnswers(selections);
+    expect(answers).toEqual({
+      mom_stage: 'expecting',
+      emotion: 'calm',
+      context: 'work',
+      need: 'rest_balance',
+    });
+
+    const p = buildUserProfileFromQuestionnaire(answers, {
+      generatedAt: '2026-01-30T00:00:00Z',
+      now: '2026-01-30T00:00:00Z',
+    });
+
+    expect(p.stage).toEqual({ expecting: 1, parenting: 0, unknown: 0 });
+    expect(p.emotion_score).toBe(0.8);
+    expect(p.context).toEqual({ work: 1 });
+    expect(p.need).toEqual({ rest_balance: 1 });
+    expect(p.profile_answered).toEqual({ stage: true, emotion: true, context: true, need: true });
+  });
+
+  it('全部跳过：仍能生成最小可计算画像（unknown=1）', () => {
+    const answers = mapOnboardingSelectionsToQuestionnaireAnswers({});
+    expect(answers).toEqual({ mom_stage: null, emotion: null, context: null, need: null });
+
+    const p = buildUserProfileFromQuestionnaire(answers, {
+      generatedAt: '2026-01-30T00:00:00Z',
+      now: '2026-01-30T00:00:00Z',
+    });
+
+    expect(p.stage).toEqual({ unknown: 1 });
+    expect(p.emotion_score).toBeNull();
+    expect(p.context).toEqual({});
+    expect(p.need).toEqual({});
+    expect(p.profile_answered).toEqual({ stage: false, emotion: false, context: false, need: false });
+  });
+});
+

client/src/features/userProfileScoring/__tests__/scoring.test.ts

@@ -0,0 +1,70 @@
+import { describe, expect, it } from 'vitest';
+
+import {
+  buildUserProfileFromQuestionnaire,
+  computeProfileConfidence,
+  computeTimeConfidence,
+  normalizeAnswers,
+} from '../index';
+
+describe('userProfileScoring V1.2', () => {
+  it('normalizeAnswers: 非法值按跳过处理', () => {
+    // @ts-expect-error: 模拟非法输入
+    const out = normalizeAnswers({ mom_stage: 'xxx', emotion: 'yyy', context: 'zzz', need: 'ooo' });
+    expect(out).toEqual({ mom_stage: undefined, emotion: undefined, context: undefined, need: undefined });
+  });
+
+  it('computeTimeConfidence: 分段衰减', () => {
+    const gen = new Date('2026-01-01T00:00:00Z');
+
+    // 0–7 天：1.0
+    expect(computeTimeConfidence(gen, new Date('2026-01-05T00:00:00Z'))).toBe(1.0);
+
+    // 30 天以上：0.5
+    expect(computeTimeConfidence(gen, new Date('2026-02-15T00:00:00Z'))).toBe(0.5);
+  });
+
+  it('computeProfileConfidence: 完整度因子 + clamp', () => {
+    const confTime = 1.0;
+
+    // 全部跳过：completion=0 → completionFactor=0.5 → 0.5
+    expect(
+      computeProfileConfidence(confTime, { stage: false, emotion: false, context: false, need: false })
+    ).toBe(0.5);
+
+    // 全部作答：completion=1 → completionFactor=1 → 1
+    expect(computeProfileConfidence(confTime, { stage: true, emotion: true, context: true, need: true })).toBe(1.0);
+  });
+
+  it('buildUserProfileFromQuestionnaire: 全部跳过输出最小可计算画像', () => {
+    const p = buildUserProfileFromQuestionnaire({}, { generatedAt: '2026-01-30T00:00:00Z', now: '2026-01-30T00:00:00Z' });
+
+    expect(p.profile_version).toBe('v1.2');
+    expect(p.profile_source).toBe('questionnaire');
+
+    expect(p.profile_answered).toEqual({ stage: false, emotion: false, context: false, need: false });
+    expect(p.stage).toEqual({ unknown: 1 });
+    expect(p.emotion_score).toBeNull();
+    expect(p.context).toEqual({});
+    expect(p.need).toEqual({});
+
+    // conf_time=1，completionFactor=0.5
+    expect(p.profile_confidence).toBe(0.5);
+
+    // unknown 会命中 unsafe_for_stage_unknown，并带跨维度谓词
+    expect(p.hard_rules.forbidden_risk_flags).toContain('unsafe_for_stage_unknown');
+    expect(p.hard_rules.forbidden_content_predicates.some((x) => x.id === 'unknown_block_parenting_pressure_personalized')).toBe(
+      true
+    );
+  });
+
+  it('buildUserProfileFromQuestionnaire: emotion<=0.2 命中 unsafe_for_emotion_low', () => {
+    const p = buildUserProfileFromQuestionnaire(
+      { mom_stage: 'expecting', emotion: 'overwhelmed', context: 'health', need: 'anxiety_relief' },
+      { generatedAt: '2026-01-30T00:00:00Z', now: '2026-01-30T00:00:00Z' }
+    );
+    expect(p.emotion_score).toBe(0.2);
+    expect(p.hard_rules.forbidden_risk_flags).toContain('unsafe_for_emotion_low');
+  });
+});
+

client/src/features/userProfileScoring/index.ts

@@ -0,0 +1,19 @@
+export type {
+  BuildUserProfileOptions,
+  QuestionnaireAnswersV1_2,
+  UserProfileV1_2,
+  UserProfileV1_2_Extended,
+} from './types';
+
+export type { OnboardingSelections } from './onboardingMapping';
+
+export {
+  buildUserProfileFromQuestionnaire,
+  computeProfileAnswered,
+  computeProfileConfidence,
+  computeTimeConfidence,
+  normalizeAnswers,
+} from './scoring';
+
+export { mapOnboardingSelectionsToQuestionnaireAnswers } from './onboardingMapping';
+

client/src/features/userProfileScoring/onboardingMapping.ts

@@ -0,0 +1,61 @@
+import type { QuestionnaireAnswersV1_2 } from './types';
+
+/**
+ * Onboarding UI 的选项 ID → 标准问卷枚举（可跳过）
+ *
+ * 说明：
+ * - UI 侧每题目前是单选，但数据结构是 string[]；这里取第 1 个作为答案
+ * - 不存在错误处理：未知/非法值统一按“跳过”处理（返回 null）
+ */
+export type OnboardingSelections = Record<string, string[] | undefined>;
+
+export function mapOnboardingSelectionsToQuestionnaireAnswers(
+  selections: OnboardingSelections
+): QuestionnaireAnswersV1_2 {
+  return {
+    mom_stage: mapMomStage(selections.status?.[0]),
+    emotion: mapEmotion(selections.emotion?.[0]),
+    context: mapContext(selections.influence?.[0]),
+    need: mapNeed(selections.support?.[0]),
+  };
+}
+
+function mapMomStage(raw: string | undefined): QuestionnaireAnswersV1_2['mom_stage'] {
+  // 跳过：null（显式跳过）
+  if (!raw) return null;
+  // UI id → 标准枚举
+  if (raw === 'pregnant') return 'expecting';
+  if (raw === 'has_kids') return 'parenting';
+  if (raw === 'no_fill') return 'unknown';
+  // 其他非法值：按跳过处理
+  return null;
+}
+
+function mapEmotion(raw: string | undefined): QuestionnaireAnswersV1_2['emotion'] {
+  if (!raw) return null;
+  // UI 当前选项：happy/calm/stressed/low
+  if (raw === 'happy') return 'joyful';
+  if (raw === 'calm') return 'calm';
+  if (raw === 'stressed') return 'overwhelmed';
+  if (raw === 'low') return 'low';
+  return null;
+}
+
+function mapContext(raw: string | undefined): QuestionnaireAnswersV1_2['context'] {
+  if (!raw) return null;
+  // UI id 已与标准枚举一致：family/work/relationship/friends/health
+  if (raw === 'family' || raw === 'work' || raw === 'relationship' || raw === 'friends' || raw === 'health') return raw;
+  return null;
+}
+
+function mapNeed(raw: string | undefined): QuestionnaireAnswersV1_2['need'] {
+  if (!raw) return null;
+  // UI id → 标准枚举
+  if (raw === 'emotional') return 'emotional_support';
+  if (raw === 'parenting') return 'parenting_pressure';
+  if (raw === 'self_worth') return 'self_worth';
+  if (raw === 'anxiety') return 'anxiety_relief';
+  if (raw === 'balance') return 'rest_balance';
+  return null;
+}
+

client/src/features/userProfileScoring/scoring.ts

@@ -0,0 +1,233 @@
+/**
+ * 用户画像打分（User Profile Scoring）V1.2
+ *
+ * 规则来源：
+ * - `spec_kit/User Profile Scoring/spec.md`
+ * - `设计说明文档/客戶端問卷打分規則.md`（V1.2）
+ */
+
+import type {
+  BuildUserProfileOptions,
+  ContextAnswer,
+  EmotionAnswer,
+  HardRules,
+  MomStageAnswer,
+  NeedAnswer,
+  ProfileAnswered,
+  QuestionnaireAnswersV1_2,
+  SparseOneHot,
+  UserProfileV1_2_Extended,
+  UserStageOneHot,
+} from './types';
+
+const MS_PER_DAY = 24 * 60 * 60 * 1000;
+
+function clamp(value: number, min: number, max: number): number {
+  if (!Number.isFinite(value)) return min;
+  return Math.min(max, Math.max(min, value));
+}
+
+function toDate(value: Date | string | undefined): Date | null {
+  if (!value) return null;
+  if (value instanceof Date) return Number.isFinite(value.getTime()) ? value : null;
+  const d = new Date(value);
+  return Number.isFinite(d.getTime()) ? d : null;
+}
+
+function isMomStageAnswer(v: unknown): v is MomStageAnswer {
+  return v === 'expecting' || v === 'parenting' || v === 'unknown';
+}
+
+function isEmotionAnswer(v: unknown): v is EmotionAnswer {
+  return (
+    v === 'low' ||
+    v === 'overwhelmed' ||
+    v === 'tired' ||
+    v === 'neutral' ||
+    v === 'calm' ||
+    v === 'joyful'
+  );
+}
+
+function isContextAnswer(v: unknown): v is ContextAnswer {
+  return v === 'family' || v === 'work' || v === 'relationship' || v === 'friends' || v === 'health';
+}
+
+function isNeedAnswer(v: unknown): v is NeedAnswer {
+  return (
+    v === 'emotional_support' ||
+    v === 'parenting_pressure' ||
+    v === 'self_worth' ||
+    v === 'anxiety_relief' ||
+    v === 'rest_balance'
+  );
+}
+
+/**
+ * 归一化答案：非法值按“跳过”处理（归一化为 undefined）
+ * - `null` 保留，表示显式跳过/无值
+ */
+export function normalizeAnswers(raw: QuestionnaireAnswersV1_2): QuestionnaireAnswersV1_2 {
+  const mom_stage =
+    raw.mom_stage === null ? null : isMomStageAnswer(raw.mom_stage) ? raw.mom_stage : undefined;
+  const emotion = raw.emotion === null ? null : isEmotionAnswer(raw.emotion) ? raw.emotion : undefined;
+  const context = raw.context === null ? null : isContextAnswer(raw.context) ? raw.context : undefined;
+  const need = raw.need === null ? null : isNeedAnswer(raw.need) ? raw.need : undefined;
+
+  return { mom_stage, emotion, context, need };
+}
+
+export function computeProfileAnswered(normalized: QuestionnaireAnswersV1_2): ProfileAnswered {
+  return {
+    stage: normalized.mom_stage !== undefined && normalized.mom_stage !== null,
+    emotion: normalized.emotion !== undefined && normalized.emotion !== null,
+    context: normalized.context !== undefined && normalized.context !== null,
+    need: normalized.need !== undefined && normalized.need !== null,
+  };
+}
+
+/**
+ * 时间衰减置信度（conf_time）
+ * - 0–7 天：1.0
+ * - 7–30 天：线性衰减到 0.7（含第 30 天）
+ * - 30 天以上：0.5
+ */
+export function computeTimeConfidence(generatedAt: Date, now: Date): number {
+  const deltaMs = now.getTime() - generatedAt.getTime();
+  if (!Number.isFinite(deltaMs) || deltaMs <= 0) return 1.0;
+
+  const days = deltaMs / MS_PER_DAY;
+  if (days <= 7) return 1.0;
+  if (days <= 30) {
+    const t = (days - 7) / (30 - 7); // 0..1
+    return 1.0 - 0.3 * t; // 1 -> 0.7
+  }
+  return 0.5;
+}
+
+/**
+ * V1.2：profile_confidence（conf_U）
+ * conf = clamp(conf_time * (0.5 + 0.5 * completion), 0.2, 1.0)
+ */
+export function computeProfileConfidence(confTime: number, answered: ProfileAnswered): number {
+  const answeredCount =
+    (answered.stage ? 1 : 0) + (answered.emotion ? 1 : 0) + (answered.context ? 1 : 0) + (answered.need ? 1 : 0);
+  const completion = answeredCount / 4;
+  const completionFactor = 0.5 + 0.5 * completion;
+  return clamp(confTime * completionFactor, 0.2, 1.0);
+}
+
+function buildStageOneHot(momStage: MomStageAnswer | null | undefined): UserStageOneHot {
+  // V1.2：mom_stage 跳过按安全策略输出 unknown=1
+  if (momStage === null || momStage === undefined) {
+    return { unknown: 1 };
+  }
+
+  return {
+    expecting: momStage === 'expecting' ? 1 : 0,
+    parenting: momStage === 'parenting' ? 1 : 0,
+    unknown: momStage === 'unknown' ? 1 : 0,
+  };
+}
+
+function mapEmotionScore(emotion: EmotionAnswer | null | undefined): number | null {
+  if (emotion === null || emotion === undefined) return null;
+  switch (emotion) {
+    case 'low':
+      return 0.0;
+    case 'overwhelmed':
+      return 0.2;
+    case 'tired':
+      return 0.4;
+    case 'neutral':
+      return 0.6;
+    case 'calm':
+      return 0.8;
+    case 'joyful':
+      return 1.0;
+  }
+}
+
+function buildSparseOneHot(value: string | null | undefined): SparseOneHot {
+  if (value === null || value === undefined) return {};
+  return { [value]: 1 };
+}
+
+function computeRuleHitsAndHardRules(profile: {
+  stage: UserStageOneHot;
+  emotion_score: number | null;
+}): { rule_hits: string[]; hard_rules: HardRules } {
+  const rule_hits: string[] = [];
+  const forbidden_risk_flags: string[] = [];
+
+  const stageUnknown = profile.stage.unknown === 1;
+  const stageParenting = profile.stage.parenting === 1;
+
+  if (stageUnknown) {
+    rule_hits.push('unsafe_for_stage_unknown');
+    forbidden_risk_flags.push('unsafe_for_stage_unknown');
+  }
+
+  if (stageParenting) {
+    rule_hits.push('unsafe_for_stage_parenting');
+    forbidden_risk_flags.push('unsafe_for_stage_parenting');
+  }
+
+  if (profile.emotion_score !== null && profile.emotion_score <= 0.2) {
+    rule_hits.push('unsafe_for_emotion_low');
+    forbidden_risk_flags.push('unsafe_for_emotion_low');
+  }
+
+  const forbidden_content_predicates = [];
+  if (stageUnknown) {
+    forbidden_content_predicates.push({
+      id: 'unknown_block_parenting_pressure_personalized',
+      when_user: { stage_unknown: true },
+      forbid_content: { need: 'parenting_pressure', personalization_power: 1 },
+    });
+  }
+
+  return {
+    rule_hits,
+    hard_rules: {
+      forbidden_risk_flags,
+      forbidden_content_predicates,
+    },
+  };
+}
+
+export function buildUserProfileFromQuestionnaire(
+  rawAnswers: QuestionnaireAnswersV1_2,
+  options: BuildUserProfileOptions = {}
+): UserProfileV1_2_Extended {
+  const normalized = normalizeAnswers(rawAnswers);
+  const profile_answered = computeProfileAnswered(normalized);
+
+  const now = toDate(options.now) ?? new Date();
+  const generatedAt = toDate(options.generatedAt) ?? now;
+
+  const confTime = computeTimeConfidence(generatedAt, now);
+  const profile_confidence = computeProfileConfidence(confTime, profile_answered);
+
+  const stage = buildStageOneHot(normalized.mom_stage);
+  const emotion_score = mapEmotionScore(normalized.emotion);
+  const context = buildSparseOneHot(normalized.context);
+  const need = buildSparseOneHot(normalized.need);
+
+  const { rule_hits, hard_rules } = computeRuleHitsAndHardRules({ stage, emotion_score });
+
+  return {
+    profile_version: 'v1.2',
+    profile_source: 'questionnaire',
+    profile_generated_at: generatedAt.toISOString(),
+    profile_confidence,
+    profile_answered,
+    stage,
+    emotion_score,
+    context,
+    need,
+    rule_hits,
+    hard_rules,
+  };
+}
+

client/src/features/userProfileScoring/types.ts

@@ -0,0 +1,95 @@
+/**
+ * 用户画像打分（User Profile Scoring）V1.2 类型定义
+ *
+ * 说明：
+ * - 本模块用于：问卷答案（可跳过）→ 用户画像（可计算、可观测、可版本化）
+ * - 字段与规则以 `spec_kit/User Profile Scoring/spec.md`（V1.2）为准
+ */
+
+export type MomStageAnswer = 'expecting' | 'parenting' | 'unknown';
+export type EmotionAnswer = 'low' | 'overwhelmed' | 'tired' | 'neutral' | 'calm' | 'joyful';
+export type ContextAnswer = 'family' | 'work' | 'relationship' | 'friends' | 'health';
+export type NeedAnswer =
+  | 'emotional_support'
+  | 'parenting_pressure'
+  | 'self_worth'
+  | 'anxiety_relief'
+  | 'rest_balance';
+
+/**
+ * V1.2：每题可跳过
+ * - `undefined`：字段缺失（可能是“没传”）
+ * - `null`：显式跳过/无值（例如 UI 明确传 null）
+ */
+export type QuestionnaireAnswersV1_2 = {
+  mom_stage?: MomStageAnswer | null;
+  emotion?: EmotionAnswer | null;
+  context?: ContextAnswer | null;
+  need?: NeedAnswer | null;
+};
+
+export type ProfileAnswered = {
+  stage: boolean;
+  emotion: boolean;
+  context: boolean;
+  need: boolean;
+};
+
+export type UserStageOneHot = {
+  expecting?: 0 | 1;
+  parenting?: 0 | 1;
+  unknown: 0 | 1;
+};
+
+export type SparseOneHot = Record<string, 1>;
+
+export type UserProfileV1_2 = {
+  profile_version: 'v1.2';
+  profile_source: 'questionnaire';
+  profile_generated_at: string; // ISO8601
+  profile_confidence: number; // 0–1
+  profile_answered: ProfileAnswered;
+  stage: UserStageOneHot;
+  emotion_score: number | null;
+  context: SparseOneHot;
+  need: SparseOneHot;
+};
+
+export type ForbiddenContentPredicate = {
+  /**
+   * 谓词 ID：用于可观测与回归测试
+   */
+  id: string;
+  /**
+   * 触发条件（用户侧）
+   * 说明：这里刻意保持为 object，便于未来接入规则引擎时做 schema 对齐。
+   */
+  when_user: Record<string, unknown>;
+  /**
+   * 禁推条件（内容侧）
+   * 说明：本模块不判断内容的 `personalization_power`，只输出可执行条件。
+   */
+  forbid_content: Record<string, unknown>;
+};
+
+export type HardRules = {
+  forbidden_risk_flags: string[];
+  forbidden_content_predicates: ForbiddenContentPredicate[];
+};
+
+export type UserProfileV1_2_Extended = UserProfileV1_2 & {
+  rule_hits: string[];
+  hard_rules: HardRules;
+};
+
+export type BuildUserProfileOptions = {
+  /**
+   * 画像生成时间；不传则使用当前时间
+   */
+  generatedAt?: Date | string;
+  /**
+   * 当前时间（用于计算 time decay）；不传则使用当前时间
+   */
+  now?: Date | string;
+};
+

client/src/i18n/index.ts

@@ -4,30 +4,21 @@ import i18n from 'i18next';
 import { initReactI18next } from 'react-i18next';
 
 import en from './locales/en.json';
-import es from './locales/es.json';
-import pt from './locales/pt.json';
-import zhCN from './locales/zh-CN.json';
 import zhTW from './locales/zh-TW.json';
 
 /**
  * 语言码约定：
- * - 简体中文：zh-CN
  * - 繁体中文：zh-TW
  * - 英语：en
- * - 西班牙语：es
- * - 葡萄牙语：pt
  */
-export type AppLanguage = 'zh-CN' | 'zh-TW' | 'en' | 'es' | 'pt';
+export type AppLanguage = 'zh-TW' | 'en';
 
 export const SUPPORTED_LANGUAGES: readonly AppLanguage[] = [
-  'zh-CN',
   'zh-TW',
   'en',
-  'es',
-  'pt',
 ] as const;
 
-const DEFAULT_FALLBACK_LANGUAGE: AppLanguage = 'zh-CN';
+const DEFAULT_FALLBACK_LANGUAGE: AppLanguage = 'en';
 const STORAGE_KEY_LANGUAGE = 'settings.language';
 
 function isSupportedLanguage(lang: string): lang is AppLanguage {
@@ -37,19 +28,13 @@ function isSupportedLanguage(lang: string): lang is AppLanguage {
 function normalizeDeviceLanguageTagToAppLanguage(languageTag: string): AppLanguage {
   const tag = languageTag.toLowerCase();
 
-  // 中文：优先区分繁简
+  // 中文：当前仅支持繁体中文（zh-TW）
   if (tag.startsWith('zh')) {
-    // 常见繁体标记：zh-TW / zh-HK / zh-Hant
-    if (tag.includes('tw') || tag.includes('hk') || tag.includes('hant')) {
     return 'zh-TW';
   }
-    return 'zh-CN';
-  }
 
-  // 其他语言：按前缀匹配
+  // 其他语言：按前缀匹配（当前仅支持英文）
   if (tag.startsWith('en')) return 'en';
-  if (tag.startsWith('es')) return 'es';
-  if (tag.startsWith('pt')) return 'pt';
 
   return DEFAULT_FALLBACK_LANGUAGE;
 }
@@ -87,7 +72,7 @@ export async function clearLanguagePreference(): Promise<void> {
  * 语言选择优先级：
  * 1) 用户设置（若存在）
  * 2) 设备语言（在支持列表内时生效；否则会被 normalize 到默认回退）
- * 3) 默认回退（zh-CN）
+  * 3) 默认回退（en）
  */
 export async function initI18n(): Promise<void> {
   if (i18n.isInitialized) return;
@@ -98,11 +83,8 @@ export async function initI18n(): Promise<void> {
 
   await i18n.use(initReactI18next).init({
     resources: {
-      'zh-CN': { translation: zhCN },
       'zh-TW': { translation: zhTW },
       en: { translation: en },
-      es: { translation: es },
-      pt: { translation: pt },
     },
     lng: initialLang,
     fallbackLng: DEFAULT_FALLBACK_LANGUAGE,

client/src/services/recoApi.ts

@@ -0,0 +1,63 @@
+import i18n from 'i18next';
+
+import { API_BASE_URL } from '@/src/constants/env';
+import type { UserProfileV1_2 } from '@/src/features/userProfileScoring';
+
+export type RecommendedItem = {
+  content_id: number;
+  text: string;
+  final_score: number;
+  fallback_level_final: number;
+  explanations?: Record<string, unknown> | null;
+};
+
+export type RecoMeta = Record<string, unknown>;
+
+export type RecoEngineResult = {
+  items: RecommendedItem[];
+  meta: RecoMeta;
+};
+
+export type RecoRequest = {
+  k?: number;
+  user_profile: UserProfileV1_2;
+  already_recommended_ids?: Array<string | number>;
+  touched_or_viewed_ids?: Array<string | number>;
+  now?: string; // ISO8601（可选）
+};
+
+function withTimeout(ms: number): AbortController {
+  const controller = new AbortController();
+  setTimeout(() => controller.abort(), ms);
+  return controller;
+}
+
+export async function fetchRecoFeed(req: RecoRequest): Promise<RecoEngineResult> {
+  const controller = withTimeout(12_000);
+  const url = `${API_BASE_URL}/v1/reco/feed`;
+
+  const res = await fetch(url, {
+    method: 'POST',
+    headers: {
+      'Content-Type': 'application/json',
+      // 让后端做 locale 选择（目前后端只区分 en/tc）
+      'Accept-Language': i18n.language || 'en',
+    },
+    body: JSON.stringify({
+      k: req.k,
+      user_profile: req.user_profile,
+      already_recommended_ids: req.already_recommended_ids ?? [],
+      touched_or_viewed_ids: req.touched_or_viewed_ids ?? [],
+      now: req.now,
+    }),
+    signal: controller.signal,
+  });
+
+  if (!res.ok) {
+    const text = await res.text().catch(() => '');
+    throw new Error(`推荐接口请求失败：${res.status} ${res.statusText} ${text}`.trim());
+  }
+
+  return (await res.json()) as RecoEngineResult;
+}
+

client/src/storage/appStorage.ts

@@ -1,4 +1,5 @@
 import AsyncStorage from '@react-native-async-storage/async-storage';
+import type { UserProfileV1_2_Extended } from '@/src/features/userProfileScoring';
 
 /**
  * 本地存储 key 统一管理，避免 UI 里散落硬编码
@@ -9,6 +10,9 @@ const KEY_CONTENT_REACTIONS = 'content.reactions';
 const KEY_FAVORITES_ITEMS = 'favorites.items';
 const KEY_CONSENT_ACCEPTED = 'consent.accepted';
 const KEY_USER_PROFILE = 'user.profile';
+const KEY_USER_PROFILE_SCORING = 'user.profileScoring';
+const KEY_RECO_FEED_CACHE = 'reco.feedCache';
+const KEY_RECO_FEED_HISTORY = 'reco.feedHistory';
 const KEY_UI_THEME_MODE = 'ui.theme.mode';
 const KEY_DAILY_REMINDER_SETTINGS = 'dailyReminder.settings';
 
@@ -20,11 +24,42 @@ export type UserProfile = {
   name?: string;
   intents?: string[];
 };
+
+/**
+ * 用户画像（问卷打分输出）
+ * 说明：用于推荐/Push/Widget 统一复用；结构以 `src/features/userProfileScoring` 输出为准。
+ */
+export type UserProfileScoring = UserProfileV1_2_Extended;
 export type DailyReminderSettings = {
   timesPerDay: number;
   pushEnabled: boolean;
 };
 
+export type RecoFeedCacheItem = {
+  content_id: number;
+  text: string;
+};
+
+export type RecoFeedCache = {
+  saved_at: string; // ISO8601
+  items: RecoFeedCacheItem[];
+  meta?: Record<string, unknown>;
+};
+
+/**
+ * Feed 链路可观测输入（用于下一次请求携带给后端）
+ *
+ * - already_recommended_ids：本设备已下发过的内容（避免重复下发）
+ * - touched_or_viewed_ids：本设备用户已看过/划过的内容（用于频控/去重/降重复）
+ *
+ * 说明：后端不需要“实时知道”，只要在下一次拉取时带上即可。
+ */
+export type RecoFeedHistory = {
+  updated_at: string; // ISO8601
+  already_recommended_ids: number[];
+  touched_or_viewed_ids: number[];
+};
+
 
 async function getJson<T>(key: string, fallback: T): Promise<T> {
   const raw = await AsyncStorage.getItem(key);
@@ -123,6 +158,103 @@ export async function setUserProfile(profile: UserProfile): Promise<void> {
   await setJson(KEY_USER_PROFILE, { ...current, ...profile });
 }
 
+export async function getUserProfileScoring(): Promise<UserProfileScoring | null> {
+  const raw = await AsyncStorage.getItem(KEY_USER_PROFILE_SCORING);
+  if (!raw) return null;
+  try {
+    return JSON.parse(raw) as UserProfileScoring;
+  } catch {
+    return null;
+  }
+}
+
+export async function setUserProfileScoring(profile: UserProfileScoring): Promise<void> {
+  await setJson(KEY_USER_PROFILE_SCORING, profile);
+}
+
+export async function getRecoFeedCache(): Promise<RecoFeedCache | null> {
+  const raw = await AsyncStorage.getItem(KEY_RECO_FEED_CACHE);
+  if (!raw) return null;
+  try {
+    return JSON.parse(raw) as RecoFeedCache;
+  } catch {
+    return null;
+  }
+}
+
+export async function setRecoFeedCache(cache: RecoFeedCache): Promise<void> {
+  await setJson(KEY_RECO_FEED_CACHE, cache);
+}
+
+export async function getRecoFeedHistory(): Promise<RecoFeedHistory> {
+  const raw = await AsyncStorage.getItem(KEY_RECO_FEED_HISTORY);
+  if (!raw) {
+    return {
+      updated_at: new Date().toISOString(),
+      already_recommended_ids: [],
+      touched_or_viewed_ids: [],
+    };
+  }
+  try {
+    const parsed = JSON.parse(raw) as Partial<RecoFeedHistory>;
+    return {
+      updated_at: typeof parsed.updated_at === 'string' ? parsed.updated_at : new Date().toISOString(),
+      already_recommended_ids: Array.isArray(parsed.already_recommended_ids)
+        ? parsed.already_recommended_ids.filter((x) => Number.isFinite(x)).map((x) => Number(x))
+        : [],
+      touched_or_viewed_ids: Array.isArray(parsed.touched_or_viewed_ids)
+        ? parsed.touched_or_viewed_ids.filter((x) => Number.isFinite(x)).map((x) => Number(x))
+        : [],
+    };
+  } catch {
+    return {
+      updated_at: new Date().toISOString(),
+      already_recommended_ids: [],
+      touched_or_viewed_ids: [],
+    };
+  }
+}
+
+export async function setRecoFeedHistory(history: RecoFeedHistory): Promise<void> {
+  await setJson(KEY_RECO_FEED_HISTORY, history);
+}
+
+function uniqKeepLatest(list: number[], max: number): number[] {
+  const seen = new Set<number>();
+  const out: number[] = [];
+  for (let i = list.length - 1; i >= 0; i -= 1) {
+    const v = list[i];
+    if (!Number.isFinite(v)) continue;
+    if (seen.has(v)) continue;
+    seen.add(v);
+    out.push(v);
+    if (out.length >= max) break;
+  }
+  return out.reverse();
+}
+
+export async function recordRecoFeedServed(contentIds: number[]): Promise<void> {
+  if (!contentIds?.length) return;
+  const h = await getRecoFeedHistory();
+  const next = {
+    ...h,
+    updated_at: new Date().toISOString(),
+    already_recommended_ids: uniqKeepLatest([...h.already_recommended_ids, ...contentIds], 500),
+  };
+  await setRecoFeedHistory(next);
+}
+
+export async function recordRecoFeedTouched(contentId: number): Promise<void> {
+  if (!Number.isFinite(contentId)) return;
+  const h = await getRecoFeedHistory();
+  const next = {
+    ...h,
+    updated_at: new Date().toISOString(),
+    touched_or_viewed_ids: uniqKeepLatest([...h.touched_or_viewed_ids, contentId], 500),
+  };
+  await setRecoFeedHistory(next);
+}
+
 export async function getDailyReminderSettings(): Promise<DailyReminderSettings> {
   const s = await getJson<DailyReminderSettings>(KEY_DAILY_REMINDER_SETTINGS, {
     timesPerDay: 3,

server/.dockerignore

@@ -0,0 +1,20 @@
+.venv
+__pycache__/
+*.pyc
+*.pyo
+*.pyd
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+
+# 本地/测试数据
+.test.db
+*.db
+
+# 测试与开发脚本（按需移除）
+tests/
+.env*
+
+# Git 元数据
+.git/
+.gitignore

server/.env.dev

@@ -7,13 +7,13 @@ APP_HOST=0.0.0.0
 APP_PORT=8000
 
 # 数据库（dev 指向 mindfulness_dev；prod 指向 mindfulness）
-DATABASE_URL=mysql+aiomysql://<用户名>:<密码>@<MYSQL_HOST>:3306/mindfulness_dev?charset=utf8mb4
+DATABASE_URL=mysql+aiomysql://damer:damer@43.163.242.87:3306/mindfulness_dev?charset=utf8mb4
 
 # Redis（使用 ACL 用户；并确保应用侧 key 带 dev:/pro: 前缀）
-REDIS_URL=redis://<REDIS_USER>:<REDIS_PASSWORD>@<REDIS_HOST>:6379/0
+REDIS_URL=redis://dev_damer:damer@43.163.242.87:6379/0
 
 # Celery（默认不启用结果存储，避免 Redis 内存压力）
-CELERY_BROKER_URL=redis://<REDIS_USER>:<REDIS_PASSWORD>@<REDIS_HOST>:6379/0
+CELERY_BROKER_URL=redis://dev_damer:damer@43.163.242.87:6379/0
 # CELERY_RESULT_BACKEND=redis://<REDIS_USER>:<REDIS_PASSWORD>@<REDIS_HOST>:6379/0
 
 # 推送（Expo）

server/.env.prod

@@ -0,0 +1,20 @@
+# 运行环境：dev 或 prod
+APP_ENV=prod
+
+# Web 服务
+APP_NAME=mindfulness-server
+APP_HOST=0.0.0.0
+APP_PORT=8000
+
+# 数据库（dev 指向 mindfulness_dev；prod 指向 mindfulness）
+DATABASE_URL=mysql+aiomysql://damer:damer@43.163.242.87:3306/mindfulness?charset=utf8mb4
+
+# Redis（使用 ACL 用户；并确保应用侧 key 带 dev:/pro: 前缀）
+REDIS_URL=redis://prod_damer:damer@43.163.242.87:6379/0
+
+# Celery（默认不启用结果存储，避免 Redis 内存压力）
+CELERY_BROKER_URL=redis://prod_damer:damer@43.163.242.87:6379/0
+# CELERY_RESULT_BACKEND=redis://<REDIS_USER>:<REDIS_PASSWORD>@<REDIS_HOST>:6379/0
+
+# 推送（Expo）
+# EXPO_ACCESS_TOKEN=

server/Dockerfile

@@ -0,0 +1,27 @@
+FROM python:3.11-slim
+
+# 运行时基础环境
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1
+
+WORKDIR /app
+
+# 系统依赖（按需扩展；多数依赖为纯 Python）
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends build-essential curl \
+    && rm -rf /var/lib/apt/lists/*
+
+# 先复制依赖清单以利用 Docker layer cache
+COPY requirements.txt /app/requirements.txt
+RUN python -m pip install -U pip \
+    && pip install --no-cache-dir -r /app/requirements.txt
+
+# 复制后端代码与迁移配置
+COPY app /app/app
+COPY alembic /app/alembic
+COPY alembic.ini /app/alembic.ini
+
+EXPOSE 8000
+
+# 生产镜像默认不开启 reload
+CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "8000"]

server/alembic.ini

@@ -0,0 +1,39 @@
+[alembic]
+script_location = alembic
+
+# 注意：实际连接串由 alembic/env.py 从环境变量 DATABASE_URL 注入
+sqlalchemy.url = driver://user:pass@localhost/dbname
+
+[loggers]
+keys = root,sqlalchemy,alembic
+
+[handlers]
+keys = console
+
+[formatters]
+keys = generic
+
+[logger_root]
+level = WARN
+handlers = console
+qualname =
+
+[logger_sqlalchemy]
+level = WARN
+handlers =
+qualname = sqlalchemy.engine
+
+[logger_alembic]
+level = INFO
+handlers =
+qualname = alembic
+
+[handler_console]
+class = StreamHandler
+args = (sys.stderr,)
+level = NOTSET
+formatter = generic
+
+[formatter_generic]
+format = %(levelname)-5.5s [%(name)s] %(message)s
+

server/alembic/README.md

@@ -0,0 +1,40 @@
+# Alembic（数据库迁移）
+
+## 1. 前置
+
+- 在 `server/` 下准备 `.env.dev`（或系统环境变量），至少包含：
+  - `DATABASE_URL=mysql+aiomysql://...`
+
+> 注意：本仓库推荐使用 Python 虚拟环境（venv）。示例以 `server/.venv` 为准。
+
+---
+
+## 2. 安装依赖（一次性）
+
+在仓库根目录：
+
+```bash
+python3 -m venv server/.venv
+source server/.venv/bin/activate
+pip install -r server/requirements.txt
+```
+
+---
+
+## 3. 常用命令
+
+在 `server/` 目录运行：
+
+```bash
+source .venv/bin/activate
+alembic -c alembic.ini history
+alembic -c alembic.ini upgrade head
+```
+
+---
+
+## 4. 说明
+
+- 连接串由 `alembic/env.py` 从环境变量 `DATABASE_URL`（或 `app/core/config.py`）读取。
+- 初始迁移版本为：`0001_init_content_tables`（创建推荐系统最小内容表与画像表）。
+

server/alembic/env.py

@@ -0,0 +1,137 @@
+from __future__ import annotations
+
+import asyncio
+import os
+import sys
+from logging.config import fileConfig
+from pathlib import Path
+
+from alembic import context
+from sqlalchemy import pool
+from sqlalchemy.engine import Connection
+from sqlalchemy.ext.asyncio import async_engine_from_config
+
+# 让 alembic 在 `server/` 下运行时也能 import app.*
+SERVER_DIR = Path(__file__).resolve().parents[1]  # .../server/alembic -> .../server
+sys.path.append(str(SERVER_DIR))
+
+from app.db.base import Base  # noqa: E402
+import app.db.models  # noqa: F401,E402  # 确保模型被导入，metadata 完整
+
+# Alembic Config 对象
+config = context.config
+
+# 配置日志
+if config.config_file_name is not None:
+    fileConfig(config.config_file_name)
+
+# 目标 metadata（autogenerate 依赖）
+target_metadata = Base.metadata
+
+
+def _read_env_kv(env_path: Path) -> dict[str, str]:
+    """
+    读取 .env 文件中的 KEY=VALUE。
+
+    说明：
+    - 迁移阶段只需要 DATABASE_URL，不应因为 Redis/Celery 等配置缺失而失败
+    - 这里不依赖 pydantic-settings 的 Settings 校验，避免“缺字段导致迁移不可用”
+    """
+
+    data: dict[str, str] = {}
+    if not env_path.exists():
+        return data
+    for raw in env_path.read_text(encoding="utf-8").splitlines():
+        line = raw.strip()
+        if not line or line.startswith("#"):
+            continue
+        if "=" not in line:
+            continue
+        k, v = line.split("=", 1)
+        k = k.strip()
+        v = v.strip().strip('"').strip("'")
+        if k:
+            data[k] = v
+    return data
+
+
+def _get_database_url() -> str:
+    """
+    获取数据库连接串。
+
+    约定：
+    - 优先读取环境变量 `DATABASE_URL`
+    - 若未设置，则按 `APP_ENV`（默认 dev）读取 `server/.env.dev` 或 `server/.env.prod`
+
+    注意：迁移阶段仅依赖 DATABASE_URL；不应强制要求 REDIS_URL / CELERY_BROKER_URL 等配置存在。
+    """
+
+    # 允许在 alembic 命令时临时覆盖
+    env_url = os.getenv("DATABASE_URL")
+    if env_url:
+        return env_url
+
+    app_env = (os.getenv("APP_ENV") or "dev").strip() or "dev"
+    env_file = SERVER_DIR / (".env.prod" if app_env == "prod" else ".env.dev")
+    kv = _read_env_kv(env_file)
+    url = kv.get("DATABASE_URL")
+    if url:
+        return url
+
+    raise RuntimeError(
+        "缺少 DATABASE_URL：请设置环境变量 DATABASE_URL，或在 server/.env.dev（或 .env.prod）中配置 DATABASE_URL。"
+    )
+
+
+def run_migrations_offline() -> None:
+    """离线模式：生成 SQL 脚本，不连接数据库。"""
+
+    url = _get_database_url()
+    context.configure(
+        url=url,
+        target_metadata=target_metadata,
+        literal_binds=True,
+        dialect_opts={"paramstyle": "named"},
+        compare_type=True,
+    )
+
+    with context.begin_transaction():
+        context.run_migrations()
+
+
+def do_run_migrations(connection: Connection) -> None:
+    """在线模式：在已有连接上执行迁移。"""
+
+    context.configure(
+        connection=connection,
+        target_metadata=target_metadata,
+        compare_type=True,
+    )
+
+    with context.begin_transaction():
+        context.run_migrations()
+
+
+async def run_migrations_online() -> None:
+    """在线模式：使用异步引擎执行迁移。"""
+
+    url = _get_database_url()
+    config.set_main_option("sqlalchemy.url", url)
+
+    connectable = async_engine_from_config(
+        config.get_section(config.config_ini_section) or {},
+        prefix="sqlalchemy.",
+        poolclass=pool.NullPool,
+    )
+
+    async with connectable.connect() as connection:
+        await connection.run_sync(do_run_migrations)
+
+    await connectable.dispose()
+
+
+if context.is_offline_mode():
+    run_migrations_offline()
+else:
+    asyncio.run(run_migrations_online())
+

server/alembic/script.py.mako

@@ -0,0 +1,27 @@
+"""${message}
+
+Revision ID: ${up_revision}
+Revises: ${down_revision | comma,n}
+Create Date: ${create_date}
+
+"""
+from __future__ import annotations
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = ${repr(up_revision)}
+down_revision = ${repr(down_revision)}
+branch_labels = ${repr(branch_labels)}
+depends_on = ${repr(depends_on)}
+
+
+def upgrade() -> None:
+    ${upgrades if upgrades else "pass"}
+
+
+def downgrade() -> None:
+    ${downgrades if downgrades else "pass"}
+

server/alembic/versions/0001_init_content_tables.py

@@ -0,0 +1,147 @@
+"""init content tables
+
+Revision ID: 0001_init_content_tables
+Revises:
+Create Date: 2026-02-01
+
+"""
+
+from __future__ import annotations
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import mysql
+
+
+# revision identifiers, used by Alembic.
+revision = "0001_init_content_tables"
+down_revision = None
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "contents",
+        sa.Column(
+            "content_id",
+            mysql.BIGINT(unsigned=True),
+            primary_key=True,
+            autoincrement=True,
+            comment="文案唯一 ID（自增；文案微调时保持不变）",
+        ),
+        sa.Column("text_en", sa.Text(), nullable=True, comment="英文文案（可空；若为空则必须提供 text_tc）"),
+        sa.Column("text_tc", sa.Text(), nullable=True, comment="繁体中文文案（可空；若为空则必须提供 text_en）"),
+        sa.Column("author_id", sa.String(length=255), nullable=True, comment="作者/来源 ID（可空；用于多样性与频控）"),
+        sa.Column("template_id", sa.String(length=255), nullable=True, comment="模板 ID（可空；用于多样性与频控）"),
+        sa.Column("created_at", sa.DateTime(), server_default=sa.func.now(), nullable=False, comment="创建时间"),
+        sa.Column("updated_at", sa.DateTime(), server_default=sa.func.now(), nullable=False, comment="更新时间"),
+        sa.CheckConstraint(
+            "(text_en IS NOT NULL) OR (text_tc IS NOT NULL)",
+            name="chk_contents_text_present",
+        ),
+        mysql_charset="utf8mb4",
+    )
+    op.create_index("idx_contents_author_id", "contents", ["author_id"], unique=False)
+    op.create_index("idx_contents_template_id", "contents", ["template_id"], unique=False)
+
+    op.create_table(
+        "content_profiles",
+        sa.Column(
+            "content_id",
+            mysql.BIGINT(unsigned=True),
+            sa.ForeignKey("contents.content_id", ondelete="CASCADE"),
+            primary_key=True,
+            comment="FK -> contents.content_id",
+        ),
+        sa.Column(
+            "stage",
+            sa.Enum("general", "expecting", "parenting", "unknown", name="content_stage"),
+            server_default="general",
+            nullable=False,
+            comment="母职阶段定位（general/expecting/parenting/unknown）",
+        ),
+        sa.Column("emotion_score", sa.Numeric(3, 2), nullable=True, comment="情绪调性 0~1；NULL 表示 general"),
+        sa.Column(
+            "context_suitability_json",
+            sa.JSON(),
+            nullable=False,
+            comment="各 context 的适配度（JSON：0/0.5/1；必须包含 5 个 key）",
+        ),
+        sa.Column(
+            "need_suitability_json",
+            sa.JSON(),
+            nullable=False,
+            comment="各 need 的适配度（JSON：0/0.5/1；必须包含 5 个 key）",
+        ),
+        sa.Column(
+            "personalization_power",
+            sa.SmallInteger(),
+            server_default="0",
+            nullable=False,
+            comment="个性化力度（约定只允许 0/5/10，分别映射 0/0.5/1）",
+        ),
+        sa.Column(
+            "review_confidence",
+            sa.Numeric(3, 2),
+            nullable=True,
+            comment="标注置信度 0~1；NULL 表示由推荐侧按 0.7 兜底",
+        ),
+        sa.Column(
+            "is_safe_pool",
+            sa.Boolean(),
+            server_default=sa.text("0"),
+            nullable=False,
+            comment="是否属于通用安全池（L3 兜底）",
+        ),
+        sa.Column("updated_at", sa.DateTime(), server_default=sa.func.now(), nullable=False, comment="画像更新时间"),
+        mysql_charset="utf8mb4",
+    )
+    op.create_index("idx_profiles_is_safe_pool", "content_profiles", ["is_safe_pool"], unique=False)
+    op.create_index("idx_profiles_personalization_power", "content_profiles", ["personalization_power"], unique=False)
+    op.create_index("idx_profiles_stage", "content_profiles", ["stage"], unique=False)
+
+    op.create_table(
+        "content_risk_flags",
+        sa.Column(
+            "id",
+            mysql.BIGINT(unsigned=True),
+            primary_key=True,
+            autoincrement=True,
+            comment="主键",
+        ),
+        sa.Column(
+            "content_id",
+            mysql.BIGINT(unsigned=True),
+            sa.ForeignKey("contents.content_id", ondelete="CASCADE"),
+            nullable=False,
+            comment="FK -> contents.content_id",
+        ),
+        sa.Column(
+            "flag",
+            sa.String(length=64),
+            nullable=False,
+            comment="风险标记（unsafe_for_* / block_* / soft_*）",
+        ),
+        sa.Column("created_at", sa.DateTime(), server_default=sa.func.now(), nullable=False, comment="创建时间"),
+        sa.UniqueConstraint("content_id", "flag", name="uniq_content_flag"),
+        mysql_charset="utf8mb4",
+    )
+    op.create_index("idx_content_id", "content_risk_flags", ["content_id"], unique=False)
+    op.create_index("idx_flag", "content_risk_flags", ["flag"], unique=False)
+
+
+def downgrade() -> None:
+    op.drop_index("idx_flag", table_name="content_risk_flags")
+    op.drop_index("idx_content_id", table_name="content_risk_flags")
+    op.drop_table("content_risk_flags")
+
+    op.drop_index("idx_profiles_stage", table_name="content_profiles")
+    op.drop_index("idx_profiles_personalization_power", table_name="content_profiles")
+    op.drop_index("idx_profiles_is_safe_pool", table_name="content_profiles")
+    op.drop_table("content_profiles")
+
+    op.drop_index("idx_contents_template_id", table_name="contents")
+    op.drop_index("idx_contents_author_id", table_name="contents")
+    op.drop_table("contents")
+

server/app/api/__init__.py

@@ -0,0 +1,6 @@
+"""
+API 路由入口
+
+说明：按 FastAPI 常见工程结构拆分 api/v1/* 路由模块。
+"""
+

server/app/api/limits.py

@@ -0,0 +1,62 @@
+from __future__ import annotations
+
+import time
+from dataclasses import dataclass, field
+from typing import Dict, Tuple
+
+from fastapi import HTTPException, Request
+
+
+@dataclass
+class FixedWindowRateLimiter:
+    """
+    固定窗口限流（内存版）。
+
+    约束：
+    - 适用于单进程/单实例；多进程/多实例下不共享计数（V1 可接受）
+    - 窗口粒度：按分钟 bucket（window_seconds 建议为 60）
+    """
+
+    limit: int
+    window_seconds: int
+    _counters: Dict[Tuple[str, int], int] = field(default_factory=dict)
+    _last_gc_bucket: int = 0
+
+    def _bucket(self, now_ts: float) -> int:
+        return int(now_ts // float(self.window_seconds))
+
+    def _gc(self, current_bucket: int) -> None:
+        # 每隔一段时间清理一次，避免 dict 无限增长（保留最近 3 个 bucket）
+        if self._last_gc_bucket == current_bucket:
+            return
+        self._last_gc_bucket = current_bucket
+        keep_from = current_bucket - 2
+        to_delete = [k for k in self._counters.keys() if k[1] < keep_from]
+        for k in to_delete:
+            self._counters.pop(k, None)
+
+    def allow(self, *, key: str, now_ts: float) -> None:
+        bucket = self._bucket(now_ts)
+        self._gc(bucket)
+
+        k = (str(key), int(bucket))
+        n = int(self._counters.get(k, 0)) + 1
+        self._counters[k] = n
+        if n > int(self.limit):
+            raise HTTPException(status_code=429, detail="rate_limited")
+
+
+_reco_rate_limiter = FixedWindowRateLimiter(limit=10, window_seconds=60)
+
+
+async def rate_limit_reco_by_ip(request: Request) -> None:
+    """
+    推荐接口限流：按 IP，1 分钟 10 次。
+    """
+
+    ip = "unknown"
+    if request.client and request.client.host:
+        ip = str(request.client.host)
+
+    _reco_rate_limiter.allow(key=ip, now_ts=time.time())
+

server/app/api/v1/__init__.py

@@ -0,0 +1,4 @@
+"""
+V1 API 路由集合
+"""
+

server/app/api/v1/reco.py

@@ -0,0 +1,156 @@
+from __future__ import annotations
+
+from datetime import datetime, timezone
+from typing import Any, Optional
+
+from fastapi import APIRouter, Depends, Header
+from pydantic import BaseModel, Field
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.api.limits import rate_limit_reco_by_ip
+from app.db.session import get_db
+from app.features.personalized_reco.content_repository.interface import ContentRepository
+from app.features.personalized_reco.content_repository.sqlalchemy_repo import SqlAlchemyContentRepository
+from app.features.personalized_reco.reco_engine import recommend
+from app.features.personalized_reco.reco_engine.types import RecoConstraints, RecoEngineResult
+from app.features.user_profile_scoring.types import UserProfileV1_2
+
+router = APIRouter(
+    prefix="/v1/reco",
+    tags=["reco"],
+    dependencies=[Depends(rate_limit_reco_by_ip)],
+)
+
+
+class RecoRequest(BaseModel):
+    k: Optional[int] = None
+    user_profile: UserProfileV1_2
+    already_recommended_ids: list[Any] = Field(default_factory=list)
+    touched_or_viewed_ids: list[Any] = Field(default_factory=list)
+    now: Optional[datetime] = None
+
+
+def _parse_now_from_header(x_now: Optional[str]) -> Optional[datetime]:
+    if not x_now:
+        return None
+    raw = str(x_now).strip()
+    if not raw:
+        return None
+    # 支持 Z
+    if raw.endswith("Z"):
+        raw = raw[:-1] + "+00:00"
+    try:
+        dt = datetime.fromisoformat(raw)
+    except Exception:
+        return None
+    if dt.tzinfo is None:
+        dt = dt.replace(tzinfo=timezone.utc)
+    return dt
+
+
+def _pick_now(*, header_now: Optional[str], body_now: Optional[datetime]) -> datetime:
+    dt = _parse_now_from_header(header_now)
+    if dt is not None:
+        return dt
+    if body_now is not None:
+        if body_now.tzinfo is None:
+            return body_now.replace(tzinfo=timezone.utc)
+        return body_now
+    return datetime.now(timezone.utc)
+
+
+def _pick_locale_from_accept_language(accept_language: Optional[str]) -> str:
+    """
+    从 Accept-Language 映射 locale：
+    - 缺失/空 -> en
+    - 含 zh-TW/zh-HK/tc -> tc
+    - 其他 -> en
+    """
+
+    raw = (accept_language or "").strip().lower()
+    if not raw:
+        return "en"
+    if "zh-tw" in raw or "zh-hk" in raw or "tc" in raw:
+        return "tc"
+    return "en"
+
+
+async def get_reco_repo(db: AsyncSession = Depends(get_db)) -> ContentRepository:
+    """
+    构造推荐 repo（可在测试中 override，避免依赖真实 DB）。
+    """
+
+    return SqlAlchemyContentRepository(db)
+
+
+@router.post("/feed", response_model=RecoEngineResult)
+async def reco_feed(
+    req: RecoRequest,
+    repo: ContentRepository = Depends(get_reco_repo),
+    x_now: Optional[str] = Header(default=None, alias="X-Now"),
+    accept_language: Optional[str] = Header(default=None, alias="Accept-Language"),
+) -> RecoEngineResult:
+    k_i = 30 if req.k is None else int(req.k)
+    now = _pick_now(header_now=x_now, body_now=req.now)
+    locale = _pick_locale_from_accept_language(accept_language)
+
+    return await recommend(
+        repo=repo,
+        scene="feed",
+        user_profile=req.user_profile,
+        already_recommended_ids=list(req.already_recommended_ids or []),
+        touched_or_viewed_ids=list(req.touched_or_viewed_ids or []),
+        k=k_i,
+        now=now,
+        locale=locale,
+        constraints=RecoConstraints(),
+    )
+
+
+@router.post("/push", response_model=RecoEngineResult)
+async def reco_push(
+    req: RecoRequest,
+    repo: ContentRepository = Depends(get_reco_repo),
+    x_now: Optional[str] = Header(default=None, alias="X-Now"),
+    accept_language: Optional[str] = Header(default=None, alias="Accept-Language"),
+) -> RecoEngineResult:
+    k_i = 1 if req.k is None else int(req.k)
+    now = _pick_now(header_now=x_now, body_now=req.now)
+    locale = _pick_locale_from_accept_language(accept_language)
+
+    return await recommend(
+        repo=repo,
+        scene="push",
+        user_profile=req.user_profile,
+        already_recommended_ids=list(req.already_recommended_ids or []),
+        touched_or_viewed_ids=list(req.touched_or_viewed_ids or []),
+        k=k_i,
+        now=now,
+        locale=locale,
+        constraints=RecoConstraints(),
+    )
+
+
+@router.post("/widget", response_model=RecoEngineResult)
+async def reco_widget(
+    req: RecoRequest,
+    repo: ContentRepository = Depends(get_reco_repo),
+    x_now: Optional[str] = Header(default=None, alias="X-Now"),
+    accept_language: Optional[str] = Header(default=None, alias="Accept-Language"),
+) -> RecoEngineResult:
+    k_i = 1 if req.k is None else int(req.k)
+    now = _pick_now(header_now=x_now, body_now=req.now)
+    locale = _pick_locale_from_accept_language(accept_language)
+
+    return await recommend(
+        repo=repo,
+        scene="widget",
+        user_profile=req.user_profile,
+        already_recommended_ids=list(req.already_recommended_ids or []),
+        touched_or_viewed_ids=list(req.touched_or_viewed_ids or []),
+        k=k_i,
+        now=now,
+        locale=locale,
+        constraints=RecoConstraints(),
+    )
+

server/app/api/v1/user_profile_scoring.py

@@ -0,0 +1,26 @@
+from __future__ import annotations
+
+from fastapi import APIRouter
+
+from app.features.user_profile_scoring.scoring import build_user_profile_from_questionnaire
+from app.features.user_profile_scoring.types import BuildUserProfileRequest, UserProfileV1_2_Extended
+
+router = APIRouter(prefix="/v1/user-profile", tags=["user-profile"])
+
+
+@router.post("/score", response_model=UserProfileV1_2_Extended)
+async def score_user_profile(req: BuildUserProfileRequest) -> UserProfileV1_2_Extended:
+    """
+    根据问卷答案生成用户画像（V1.2）
+
+    说明：
+    - 问卷题目允许跳过
+    - 允许注入 generated_at/now，用于回归测试或离线批处理
+    """
+
+    return build_user_profile_from_questionnaire(
+        req.answers,
+        generated_at=req.generated_at,
+        now=req.now,
+    )
+

server/app/db/models/__init__.py

@@ -0,0 +1,14 @@
+"""
+数据库 ORM 模型集合。
+
+说明：
+- 该包用于集中定义 SQLAlchemy ORM models，供 Alembic autogenerate 扫描。
+- 需要在此处导入所有模型，确保 `Base.metadata` 完整。
+"""
+
+from app.db.models.content import Content
+from app.db.models.content_profile import ContentProfile
+from app.db.models.content_risk_flag import ContentRiskFlag
+
+__all__ = ["Content", "ContentProfile", "ContentRiskFlag"]
+

server/app/db/models/content.py

@@ -0,0 +1,70 @@
+from __future__ import annotations
+
+from datetime import datetime
+
+from sqlalchemy import CheckConstraint, DateTime, Index, Text, func
+from sqlalchemy.orm import Mapped, mapped_column
+
+from app.db.base import Base
+
+
+class Content(Base):
+    """
+    文案主体表。
+
+    多语言约束：
+    - 当前仅支持 EN / TC（繁体中文）
+    - 至少需要提供 `text_en` 或 `text_tc` 之一
+    """
+
+    __tablename__ = "contents"
+
+    __table_args__ = (
+        CheckConstraint(
+            "(text_en IS NOT NULL) OR (text_tc IS NOT NULL)",
+            name="chk_contents_text_present",
+        ),
+        Index("idx_contents_author_id", "author_id"),
+        Index("idx_contents_template_id", "template_id"),
+    )
+
+    content_id: Mapped[int] = mapped_column(
+        primary_key=True,
+        autoincrement=True,
+        comment="文案唯一 ID（自增；文案微调时保持不变）",
+    )
+
+    text_en: Mapped[str | None] = mapped_column(
+        Text,
+        nullable=True,
+        comment="英文文案（可空；若为空则必须提供 text_tc）",
+    )
+    text_tc: Mapped[str | None] = mapped_column(
+        Text,
+        nullable=True,
+        comment="繁体中文文案（可空；若为空则必须提供 text_en）",
+    )
+
+    author_id: Mapped[str | None] = mapped_column(
+        nullable=True,
+        comment="作者/来源 ID（可空；用于多样性与频控）",
+    )
+    template_id: Mapped[str | None] = mapped_column(
+        nullable=True,
+        comment="模板 ID（可空；用于多样性与频控）",
+    )
+
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime,
+        nullable=False,
+        server_default=func.now(),
+        comment="创建时间",
+    )
+    updated_at: Mapped[datetime] = mapped_column(
+        DateTime,
+        nullable=False,
+        server_default=func.now(),
+        server_onupdate=func.now(),
+        comment="更新时间",
+    )
+

server/app/db/models/content_profile.py

@@ -0,0 +1,95 @@
+from __future__ import annotations
+
+from datetime import datetime
+from typing import Literal, Optional
+
+from sqlalchemy import (
+    JSON,
+    Boolean,
+    DateTime,
+    Enum,
+    ForeignKey,
+    Index,
+    Numeric,
+    func,
+)
+from sqlalchemy.orm import Mapped, mapped_column
+
+from app.db.base import Base
+
+ContentStage = Literal["general", "expecting", "parenting", "unknown"]
+
+
+class ContentProfile(Base):
+    """
+    内容画像表（Content Profile / Cᵢ）。
+
+    字段语义必须严格对齐：
+    - `设计说明文档/句子文案打分規則.md`
+    """
+
+    __tablename__ = "content_profiles"
+
+    __table_args__ = (
+        Index("idx_profiles_stage", "stage"),
+        Index("idx_profiles_personalization_power", "personalization_power"),
+        Index("idx_profiles_is_safe_pool", "is_safe_pool"),
+    )
+
+    content_id: Mapped[int] = mapped_column(
+        ForeignKey("contents.content_id", ondelete="CASCADE"),
+        primary_key=True,
+        comment="FK -> contents.content_id",
+    )
+
+    stage: Mapped[ContentStage] = mapped_column(
+        Enum("general", "expecting", "parenting", "unknown", name="content_stage"),
+        nullable=False,
+        server_default="general",
+        comment="母职阶段定位（general/expecting/parenting/unknown）",
+    )
+
+    emotion_score: Mapped[Optional[float]] = mapped_column(
+        Numeric(3, 2),
+        nullable=True,
+        comment="情绪调性 0~1；NULL 表示 general",
+    )
+
+    context_suitability_json: Mapped[dict] = mapped_column(
+        JSON,
+        nullable=False,
+        comment="各 context 的适配度（JSON：0/0.5/1；必须包含 5 个 key）",
+    )
+    need_suitability_json: Mapped[dict] = mapped_column(
+        JSON,
+        nullable=False,
+        comment="各 need 的适配度（JSON：0/0.5/1；必须包含 5 个 key）",
+    )
+
+    personalization_power: Mapped[int] = mapped_column(
+        nullable=False,
+        server_default="0",
+        comment="个性化力度（约定只允许 0/5/10，分别映射 0/0.5/1）",
+    )
+
+    review_confidence: Mapped[Optional[float]] = mapped_column(
+        Numeric(3, 2),
+        nullable=True,
+        comment="标注置信度 0~1；NULL 表示由推荐侧按 0.7 兜底",
+    )
+
+    is_safe_pool: Mapped[bool] = mapped_column(
+        Boolean,
+        nullable=False,
+        server_default="0",
+        comment="是否属于通用安全池（L3 兜底）",
+    )
+
+    updated_at: Mapped[datetime] = mapped_column(
+        DateTime,
+        nullable=False,
+        server_default=func.now(),
+        server_onupdate=func.now(),
+        comment="画像更新时间",
+    )
+

server/app/db/models/content_risk_flag.py

@@ -0,0 +1,51 @@
+from __future__ import annotations
+
+from datetime import datetime
+
+from sqlalchemy import DateTime, ForeignKey, Index, UniqueConstraint, func
+from sqlalchemy.orm import Mapped, mapped_column
+
+from app.db.base import Base
+
+
+class ContentRiskFlag(Base):
+    """
+    内容风险标记（risk_flags）关联表。
+
+    命名约束（语义来源：句子文案打分规则）：
+    - 仅允许 `unsafe_for_*` / `block_*` / `soft_*` 前缀
+    - 旧 flag（如 `block_stage_unknown`）需在写入/读取层做映射
+    """
+
+    __tablename__ = "content_risk_flags"
+
+    __table_args__ = (
+        UniqueConstraint("content_id", "flag", name="uniq_content_flag"),
+        Index("idx_flag", "flag"),
+        Index("idx_content_id", "content_id"),
+    )
+
+    id: Mapped[int] = mapped_column(
+        primary_key=True,
+        autoincrement=True,
+        comment="主键",
+    )
+
+    content_id: Mapped[int] = mapped_column(
+        ForeignKey("contents.content_id", ondelete="CASCADE"),
+        nullable=False,
+        comment="FK -> contents.content_id",
+    )
+
+    flag: Mapped[str] = mapped_column(
+        nullable=False,
+        comment="风险标记（unsafe_for_* / block_* / soft_*）",
+    )
+
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime,
+        nullable=False,
+        server_default=func.now(),
+        comment="创建时间",
+    )
+

server/app/features/personalized_reco/__init__.py

@@ -0,0 +1,6 @@
+"""
+Personalized Reco（个性化推荐）功能模块集合。
+
+该目录用于承载推荐引擎与其子模块（数据访问、打分、重排、可观测等）。
+"""
+

server/app/features/personalized_reco/content_repository/__init__.py

@@ -0,0 +1,8 @@
+"""
+Content Repository（候选查询与数据访问层）。
+
+说明：
+- 本模块为推荐引擎提供可注入的数据访问接口（与 ORM/SQL 解耦）。
+- 负责将 DB 存储形态规范化为上层稳定的 ContentProfile 结构。
+"""
+

server/app/features/personalized_reco/content_repository/interface.py

@@ -0,0 +1,36 @@
+from __future__ import annotations
+
+from typing import Protocol
+
+from app.features.personalized_reco.content_repository.types import ContentProfileDTO
+
+
+class ContentRepository(Protocol):
+    """
+    推荐引擎依赖的内容数据访问抽象接口（用于解耦 ORM/SQL）。
+    """
+
+    async def fetch_candidates(
+        self,
+        *,
+        scene: str,
+        user_profile: object,
+        fallback_level: int,
+        limit: int,
+        locale: str,
+        exclude_content_ids: list[int] | None = None,
+    ) -> list[ContentProfileDTO]:
+        """
+        按场景与用户画像拉取候选内容画像（用于候选池）。
+        """
+
+    async def fetch_contents_by_ids(
+        self,
+        *,
+        content_ids: list[int],
+        locale: str,
+    ) -> list[ContentProfileDTO]:
+        """
+        按 content_id 批量获取内容画像（去重、按输入顺序返回；缺语言/缺记录的 id 跳过）。
+        """
+

server/app/features/personalized_reco/content_repository/normalization.py

@@ -0,0 +1,130 @@
+from __future__ import annotations
+
+from typing import Any
+
+from app.features.personalized_reco.content_repository.types import Locale, normalize_locale
+
+
+CONTEXT_KEYS: tuple[str, ...] = ("family", "work", "relationship", "friends", "health")
+NEED_KEYS: tuple[str, ...] = (
+    "emotional_support",
+    "parenting_pressure",
+    "self_worth",
+    "anxiety_relief",
+    "rest_balance",
+)
+
+
+def _normalize_discrete_score(v: Any, *, default: float = 0.5) -> float:
+    """
+    将 suitability 的离散值规范化为 0/0.5/1。
+
+    非法值一律兜底 default（默认 0.5）。
+    """
+
+    try:
+        if v in (0, 0.0):
+            return 0.0
+        if v in (0.5,):
+            return 0.5
+        if v in (1, 1.0):
+            return 1.0
+        # 允许字符串形式的 "0"/"0.5"/"1"
+        if isinstance(v, str):
+            s = v.strip()
+            if s == "0":
+                return 0.0
+            if s == "0.5":
+                return 0.5
+            if s == "1":
+                return 1.0
+    except Exception:
+        return default
+    return default
+
+
+def normalize_suitability(raw: Any, *, keys: tuple[str, ...]) -> dict[str, float]:
+    """
+    解析 suitability JSON，缺失时补齐全 0.5。
+
+    raw 期望为 dict；否则视为缺失。
+    """
+
+    data: dict[str, Any] = raw if isinstance(raw, dict) else {}
+    return {k: _normalize_discrete_score(data.get(k), default=0.5) for k in keys}
+
+
+def normalize_review_confidence(raw: Any) -> float:
+    """
+    review_confidence 缺失/NULL 时兜底 0.7。
+    """
+
+    try:
+        if raw is None:
+            return 0.7
+        v = float(raw)
+        if 0.0 <= v <= 1.0:
+            return v
+    except Exception:
+        pass
+    return 0.7
+
+
+def normalize_personalization_power(raw: Any) -> float:
+    """
+    DB 约定存 0/5/10，读取层输出 0/0.5/1。
+    """
+
+    try:
+        if raw is None:
+            return 0.0
+        v = int(raw)
+        if v == 0:
+            return 0.0
+        if v == 5:
+            return 0.5
+        if v == 10:
+            return 1.0
+    except Exception:
+        pass
+    return 0.0
+
+
+_RISK_FLAG_MAP: dict[str, str] = {
+    "block_stage_unknown": "unsafe_for_stage_unknown",
+    "block_stage_parenting": "unsafe_for_stage_parenting",
+    "block_emotion_low": "unsafe_for_emotion_low",
+    "block_health_sensitive": "block_health_medical",
+}
+
+
+def normalize_risk_flags(raw_flags: list[str] | None) -> list[str]:
+    """
+    risk_flags 旧→新映射、去重、稳定排序（字典序）。
+    """
+
+    flags = raw_flags or []
+    mapped: set[str] = set()
+    for f in flags:
+        if not f:
+            continue
+        name = _RISK_FLAG_MAP.get(f, f)
+        mapped.add(name)
+    return sorted(mapped)
+
+
+def pick_text(*, text_en: str | None, text_tc: str | None, locale: str) -> str | None:
+    """
+    按 locale 选择输出文案文本。
+
+    当前仅支持 EN/TC，且不允许语言回退：
+    - locale=en*：必须使用 text_en
+    - locale=tc/zh-TW/zh-HK：必须使用 text_tc
+    """
+
+    loc: Locale = normalize_locale(locale)
+    if loc == "en":
+        return text_en if text_en else None
+    # loc == "tc"
+    return text_tc if text_tc else None
+

server/app/features/personalized_reco/content_repository/sqlalchemy_repo.py

@@ -0,0 +1,265 @@
+from __future__ import annotations
+
+from collections import defaultdict
+from dataclasses import dataclass
+from typing import Any, Iterable
+
+from sqlalchemy import Select, and_, desc, not_, select
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.db.models.content import Content
+from app.db.models.content_profile import ContentProfile
+from app.db.models.content_risk_flag import ContentRiskFlag
+from app.features.personalized_reco.content_repository.interface import ContentRepository
+from app.features.personalized_reco.content_repository.normalization import (
+    CONTEXT_KEYS,
+    NEED_KEYS,
+    normalize_personalization_power,
+    normalize_review_confidence,
+    normalize_risk_flags,
+    normalize_suitability,
+    pick_text,
+)
+from app.features.personalized_reco.content_repository.types import ContentProfileDTO
+
+
+@dataclass(frozen=True)
+class _UserSignals:
+    """
+    从 user_profile 中提取 repository 级别需要的最小信号。
+
+    注意：更复杂的规则（Hard Filter/Scoring/Rerank）不在本层处理。
+    """
+
+    missing_need: bool
+    missing_context: bool
+    missing_emotion: bool
+    stage: str | None  # expecting/parenting/unknown/general/None
+
+
+def _bool(v: Any) -> bool:
+    return bool(v)
+
+
+def _extract_user_signals(user_profile: object) -> _UserSignals:
+    """
+    兼容 pydantic model / dict / 其他对象的最小字段读取。
+    """
+
+    def _get(obj: Any, key: str, default: Any = None) -> Any:
+        if obj is None:
+            return default
+        if isinstance(obj, dict):
+            return obj.get(key, default)
+        return getattr(obj, key, default)
+
+    need = _get(user_profile, "need", {}) or {}
+    context = _get(user_profile, "context", {}) or {}
+    emotion_score = _get(user_profile, "emotion_score", None)
+
+    missing_need = len(need) == 0
+    missing_context = len(context) == 0
+    missing_emotion = emotion_score is None
+
+    # stage: from user_profile.stage (one-hot)
+    stage_obj = _get(user_profile, "stage", None)
+    stage: str | None = None
+    if stage_obj is not None:
+        expecting = _get(stage_obj, "expecting", None)
+        parenting = _get(stage_obj, "parenting", None)
+        unknown = _get(stage_obj, "unknown", None)
+        if _bool(expecting):
+            stage = "expecting"
+        elif _bool(parenting):
+            stage = "parenting"
+        elif _bool(unknown):
+            stage = "unknown"
+
+    return _UserSignals(
+        missing_need=missing_need,
+        missing_context=missing_context,
+        missing_emotion=missing_emotion,
+        stage=stage,
+    )
+
+
+def _dedupe_preserve_order(ids: Iterable[int]) -> list[int]:
+    seen: set[int] = set()
+    out: list[int] = []
+    for i in ids:
+        if i in seen:
+            continue
+        seen.add(i)
+        out.append(i)
+    return out
+
+
+class SqlAlchemyContentRepository(ContentRepository):
+    """
+    基于 SQLAlchemy AsyncSession 的 ContentRepository 实现。
+    """
+
+    def __init__(self, session: AsyncSession):
+        self._session = session
+
+    async def fetch_contents_by_ids(self, *, content_ids: list[int], locale: str) -> list[ContentProfileDTO]:
+        """
+        - 输入去重
+        - 输出顺序与输入一致（按首次出现顺序）
+        - 缺记录或缺目标语言文本：跳过
+        - 不产生 N+1（主体+画像一次，flags 一次）
+        """
+
+        unique_ids = _dedupe_preserve_order(content_ids)
+        if not unique_ids:
+            return []
+
+        # locale 文本存在性过滤（不允许语言回退）
+        # en -> 必须 text_en；tc -> 必须 text_tc
+        # 过滤在 DB 层做，避免后续组装无意义
+        from app.features.personalized_reco.content_repository.types import normalize_locale
+
+        loc = normalize_locale(locale)
+        text_filter = Content.text_en.is_not(None) if loc == "en" else Content.text_tc.is_not(None)
+
+        stmt: Select = (
+            select(Content, ContentProfile)
+            .join(ContentProfile, Content.content_id == ContentProfile.content_id)
+            .where(and_(Content.content_id.in_(unique_ids), text_filter))
+        )
+
+        rows = (await self._session.execute(stmt)).all()
+        if not rows:
+            return []
+
+        # 先组装主体+画像，后续再补 risk_flags
+        by_id: dict[int, dict[str, Any]] = {}
+        valid_ids: list[int] = []
+        for content, profile in rows:
+            cid = int(content.content_id)
+            text = pick_text(text_en=content.text_en, text_tc=content.text_tc, locale=locale)
+            if not text:
+                continue
+            by_id[cid] = {
+                "content": content,
+                "profile": profile,
+                "text": text,
+            }
+            valid_ids.append(cid)
+
+        if not by_id:
+            return []
+
+        # 批量取 flags（避免 join 行膨胀）
+        flags_stmt = select(ContentRiskFlag.content_id, ContentRiskFlag.flag).where(
+            ContentRiskFlag.content_id.in_(list(by_id.keys()))
+        )
+        flags_rows = (await self._session.execute(flags_stmt)).all()
+        flags_map: dict[int, list[str]] = defaultdict(list)
+        for cid, flag in flags_rows:
+            flags_map[int(cid)].append(str(flag))
+
+        result_by_id: dict[int, ContentProfileDTO] = {}
+        for cid, payload in by_id.items():
+            content: Content = payload["content"]
+            profile: ContentProfile = payload["profile"]
+            text: str = payload["text"]
+
+            dto = ContentProfileDTO(
+                content_id=cid,
+                text=text,
+                stage=profile.stage,  # type: ignore[arg-type]
+                emotion_score=float(profile.emotion_score) if profile.emotion_score is not None else None,
+                context_suitability=normalize_suitability(profile.context_suitability_json, keys=CONTEXT_KEYS),
+                need_suitability=normalize_suitability(profile.need_suitability_json, keys=NEED_KEYS),
+                personalization_power=normalize_personalization_power(profile.personalization_power),
+                risk_flags=normalize_risk_flags(flags_map.get(cid)),
+                author_id=content.author_id,
+                template_id=content.template_id,
+                review_confidence=normalize_review_confidence(profile.review_confidence),
+            )
+            result_by_id[cid] = dto
+
+        # 按输入顺序返回（跳过缺失/被过滤的）
+        out: list[ContentProfileDTO] = []
+        for cid in unique_ids:
+            dto = result_by_id.get(cid)
+            if dto is not None:
+                out.append(dto)
+        return out
+
+    async def fetch_candidates(
+        self,
+        *,
+        scene: str,
+        user_profile: object,
+        fallback_level: int,
+        limit: int,
+        locale: str,
+        exclude_content_ids: list[int] | None = None,
+    ) -> list[ContentProfileDTO]:
+        """
+        两段式候选召回：
+        1) 先查候选 content_id 列表（含粗过滤、locale 过滤、limit*multiplier）
+        2) 再批量补全字段（复用 fetch_contents_by_ids）
+        """
+
+        if limit <= 0:
+            return []
+
+        signals = _extract_user_signals(user_profile)
+        effective_fallback = int(fallback_level)
+        if signals.missing_need or signals.missing_context or signals.missing_emotion:
+            effective_fallback = max(effective_fallback, 1)
+
+        # locale 文本存在性过滤（不允许语言回退）
+        from app.features.personalized_reco.content_repository.types import normalize_locale
+
+        loc = normalize_locale(locale)
+        text_filter = Content.text_en.is_not(None) if loc == "en" else Content.text_tc.is_not(None)
+
+        filters: list[Any] = [text_filter]
+        if exclude_content_ids:
+            filters.append(not_(Content.content_id.in_(exclude_content_ids)))
+
+        # fallback 约束（repository 只做“降级约束”，不做 hard filter）
+        if effective_fallback >= 1:
+            # personalization_power <= 5 代表 <= 0.5
+            filters.append(ContentProfile.personalization_power <= 5)
+        if effective_fallback >= 2:
+            filters.append(ContentProfile.personalization_power == 0)
+            filters.append(ContentProfile.stage == "general")
+        if effective_fallback >= 3:
+            filters.append(ContentProfile.is_safe_pool.is_(True))
+            filters.append(ContentProfile.personalization_power == 0)
+            filters.append(ContentProfile.stage == "general")
+
+        # stage 粗过滤（仅 L0/L1 才做“用户阶段 + general”；L2/L3 已强制 general）
+        if effective_fallback < 2:
+            user_stage = signals.stage
+            if user_stage in {"expecting", "parenting"}:
+                filters.append(ContentProfile.stage.in_([user_stage, "general"]))
+            else:
+                # unknown 或无法判定：仅取 general，避免误推
+                filters.append(ContentProfile.stage == "general")
+
+        multiplier = 5
+        raw_limit = max(limit * multiplier, limit)
+
+        stmt_ids = (
+            select(Content.content_id)
+            .join(ContentProfile, Content.content_id == ContentProfile.content_id)
+            .where(and_(*filters))
+            .order_by(desc(ContentProfile.updated_at))
+            .limit(raw_limit)
+        )
+
+        candidate_ids_rows = (await self._session.execute(stmt_ids)).scalars().all()
+        candidate_ids = [int(x) for x in candidate_ids_rows]
+        if not candidate_ids:
+            return []
+
+        # 复用按 ID 批量补全（会再次做 locale 过滤，但成本可接受，且可保证一致行为）
+        items = await self.fetch_contents_by_ids(content_ids=candidate_ids, locale=locale)
+        return items[:limit]
+

server/app/features/personalized_reco/content_repository/types.py

@@ -0,0 +1,64 @@
+from __future__ import annotations
+
+from typing import Literal, Optional
+
+from pydantic import BaseModel, Field
+
+
+# 当前阶段仅支持 EN / TC（繁体中文）
+Locale = Literal["en", "tc"]
+
+
+def normalize_locale(locale: str) -> Locale:
+    """
+    将客户端传入的 locale 归一化为内部枚举（仅 EN / TC）。
+
+    约定：
+    - 任何以 "en" 开头的 locale 归一化为 "en"（例如 en、en-US）
+    - "tc"/"zh-TW"/"zh-HK" 归一化为 "tc"
+    - 其他 locale 视为不支持
+    """
+
+    raw = (locale or "").strip()
+    if not raw:
+        raise ValueError("locale 不能为空（当前仅支持 en/tc）")
+
+    low = raw.lower()
+    if low.startswith("en"):
+        return "en"
+    if low in {"tc", "zh-tw", "zh-hk", "zh_tw", "zh_hk"}:
+        return "tc"
+
+    raise ValueError(f"不支持的 locale：{locale!r}（当前仅支持 en/tc）")
+
+
+ContentStage = Literal["general", "expecting", "parenting", "unknown"]
+
+
+class ContentProfileDTO(BaseModel):
+    """
+    推荐模块消费的内容画像（稳定字段契约）。
+
+    注意：
+    - text 已按 locale 选择，不允许语言回退（缺语言文本的内容不返回）
+    - emotion_score 为 None 表示 general
+    - personalization_power 对上统一为 0/0.5/1
+    - review_confidence 缺失时兜底 0.7
+    """
+
+    content_id: int
+    text: str
+    stage: ContentStage
+    emotion_score: Optional[float] = None
+
+    context_suitability: dict[str, float] = Field(default_factory=dict)
+    need_suitability: dict[str, float] = Field(default_factory=dict)
+
+    personalization_power: float
+    risk_flags: list[str] = Field(default_factory=list)
+
+    # 可选字段
+    author_id: Optional[str] = None
+    template_id: Optional[str] = None
+    review_confidence: float = 0.7
+

server/app/features/personalized_reco/observability/__init__.py

@@ -0,0 +1,20 @@
+"""
+个性化推荐｜Observability 子模块（可观测性与打点载荷）
+
+说明：
+- 只负责统一 `RecoMeta` 结构与构建（builder），不负责埋点 SDK/落库/上报实现。
+- `RecoMeta` 需要同时被 `reco-engine` 与 `integration-api-worker` 使用。
+"""
+
+from .builder import RecoMetaBuilder
+from .types import MissingFields, RecoMeta
+from .utils import compute_empty_reason, compute_missing_fields
+
+__all__ = [
+    "MissingFields",
+    "RecoMeta",
+    "RecoMetaBuilder",
+    "compute_empty_reason",
+    "compute_missing_fields",
+]
+

server/app/features/personalized_reco/observability/builder.py

@@ -0,0 +1,136 @@
+from __future__ import annotations
+
+import logging
+from datetime import datetime
+from typing import Any, Optional
+
+from app.features.personalized_reco.observability.types import MissingFields, RecoMeta, Scene
+from app.features.personalized_reco.observability.utils import compute_empty_reason, compute_missing_fields
+
+logger = logging.getLogger(__name__)
+
+
+def _non_negative_int(value: Any, *, default: int = 0) -> int:
+    try:
+        n = int(value)
+    except Exception:
+        return int(default)
+    return max(0, int(n))
+
+
+class RecoMetaBuilder:
+    """
+    在推荐 pipeline 中逐阶段填充 RecoMeta，避免“散落字段/散落日志”。
+
+    说明（V1）：
+    - set 调用允许任意顺序；build 时会做防御式兜底与单调性修正
+    - 单调性约束：raw >= after_hard_filter >= after_dedup >= after_freqcap >= served_k
+    """
+
+    def __init__(self, *, scene: Scene, user_profile: object, k: int, now: Optional[datetime] = None) -> None:
+        self.scene: Scene = scene
+        self.user_profile = user_profile
+        self.k = _non_negative_int(k, default=0)
+        self.now = now
+
+        self._raw: Optional[int] = None
+        self._after_hard: Optional[int] = None
+        self._after_dedup: Optional[int] = None
+        self._after_freqcap: Optional[int] = None
+        self._served_k: Optional[int] = None
+        self._fallback_level_final: Optional[int] = None
+
+        self._risk_filtered_count_by_flag: dict[str, int] = {}
+        self._freqcap_filtered_counts: dict[str, int] = {}
+        self._config_snapshot: dict[str, Any] = {}
+
+    def set_candidate_pool_size_raw(self, n: Any) -> "RecoMetaBuilder":
+        self._raw = _non_negative_int(n)
+        return self
+
+    def set_after_hard_filter(self, n: Any, *, risk_filtered_count_by_flag: Optional[dict[str, Any]] = None) -> "RecoMetaBuilder":
+        self._after_hard = _non_negative_int(n)
+        if risk_filtered_count_by_flag:
+            self._risk_filtered_count_by_flag = {str(k): _non_negative_int(v) for k, v in risk_filtered_count_by_flag.items()}
+        return self
+
+    def set_after_dedup(self, n: Any) -> "RecoMetaBuilder":
+        self._after_dedup = _non_negative_int(n)
+        return self
+
+    def set_after_freqcap(self, n: Any, *, freqcap_filtered_counts: Optional[dict[str, Any]] = None) -> "RecoMetaBuilder":
+        self._after_freqcap = _non_negative_int(n)
+        if freqcap_filtered_counts:
+            self._freqcap_filtered_counts = {str(k): _non_negative_int(v) for k, v in freqcap_filtered_counts.items()}
+        return self
+
+    def set_fallback_level_final(self, level: Any, *, reason: Optional[str] = None) -> "RecoMetaBuilder":
+        # reason 预留，V1 先不入 meta（可放入 config_snapshot 或后续字段）
+        self._fallback_level_final = _non_negative_int(level, default=0)
+        if reason:
+            self._config_snapshot.setdefault("fallback_trigger_reason", str(reason))
+        return self
+
+    def set_served_k(self, n: Any) -> "RecoMetaBuilder":
+        self._served_k = _non_negative_int(n)
+        return self
+
+    def set_config_snapshot(self, snapshot: dict[str, Any]) -> "RecoMetaBuilder":
+        self._config_snapshot = dict(snapshot or {})
+        return self
+
+    def build(self) -> RecoMeta:
+        missing: MissingFields = compute_missing_fields(self.user_profile)
+        conf_u = getattr(self.user_profile, "profile_confidence", 1.0)
+        try:
+            conf_u_f = float(conf_u)
+        except Exception:
+            conf_u_f = 1.0
+        if conf_u_f != conf_u_f:
+            conf_u_f = 1.0
+
+        raw = self._raw if self._raw is not None else 0
+        after_hard = self._after_hard if self._after_hard is not None else raw
+        after_dedup = self._after_dedup if self._after_dedup is not None else after_hard
+        after_freqcap = self._after_freqcap if self._after_freqcap is not None else after_dedup
+        served_k = self._served_k if self._served_k is not None else 0
+
+        # 防御式单调性修正（以最保守值输出）
+        if after_hard > raw:
+            logger.debug("after_hard_filter(%s) > raw(%s)，已修正为 raw", after_hard, raw)
+            after_hard = raw
+        if after_dedup > after_hard:
+            logger.debug("after_dedup(%s) > after_hard_filter(%s)，已修正为 after_hard_filter", after_dedup, after_hard)
+            after_dedup = after_hard
+        if after_freqcap > after_dedup:
+            logger.debug("after_freqcap(%s) > after_dedup(%s)，已修正为 after_dedup", after_freqcap, after_dedup)
+            after_freqcap = after_dedup
+        if served_k > after_freqcap:
+            logger.debug("served_k(%s) > after_freqcap(%s)，已修正为 after_freqcap", served_k, after_freqcap)
+            served_k = after_freqcap
+
+        fallback_level_final = self._fallback_level_final if self._fallback_level_final is not None else 0
+
+        empty_reason = compute_empty_reason(
+            served_k=served_k,
+            candidate_pool_size_raw=raw,
+            candidate_pool_size_after_hard_filter=after_hard,
+            candidate_pool_size_after_freqcap=after_freqcap,
+        )
+
+        return RecoMeta(
+            scene=self.scene,
+            candidate_pool_size_raw=int(raw),
+            candidate_pool_size_after_hard_filter=int(after_hard),
+            candidate_pool_size_after_dedup=int(after_dedup),
+            candidate_pool_size_after_freqcap=int(after_freqcap),
+            fallback_level_final=int(fallback_level_final),
+            served_k=int(served_k),
+            empty_reason=empty_reason,
+            conf_U=float(conf_u_f),
+            missing_fields=missing,
+            risk_filtered_count_by_flag=dict(self._risk_filtered_count_by_flag),
+            freqcap_filtered_counts=dict(self._freqcap_filtered_counts),
+            config_snapshot=dict(self._config_snapshot),
+        )
+

server/app/features/personalized_reco/observability/types.py

@@ -0,0 +1,51 @@
+from __future__ import annotations
+
+from typing import Any, Literal, Optional
+
+from pydantic import BaseModel, Field
+
+Scene = Literal["feed", "push", "widget"]
+
+EmptyReason = Literal["hard_filter_all", "freqcap_all", "pool_empty", "unknown"]
+
+
+class MissingFields(BaseModel):
+    """
+    画像字段缺失情况（布尔结构）。
+    """
+
+    need: bool = False
+    context: bool = False
+    emotion: bool = False
+
+
+class RecoMeta(BaseModel):
+    """
+    推荐模块统一可观测载荷（返回给调用方；调用方负责上报/落库/打点）。
+    """
+
+    scene: Scene
+
+    candidate_pool_size_raw: int = 0
+    candidate_pool_size_after_hard_filter: int = 0
+    candidate_pool_size_after_dedup: int = 0
+    candidate_pool_size_after_freqcap: int = 0
+
+    fallback_level_final: int = 0
+    served_k: int = 0
+
+    # served_k=0 时必填；served_k>0 时建议为 None
+    empty_reason: Optional[EmptyReason] = None
+
+    conf_U: float = 1.0
+    missing_fields: MissingFields = Field(default_factory=MissingFields)
+
+    # 可选：Hard Filter 风险命中统计（按 flag 聚合）
+    risk_filtered_count_by_flag: dict[str, int] = Field(default_factory=dict)
+
+    # 可选：Freqcap 过滤统计（sentence/author/template）
+    freqcap_filtered_counts: dict[str, int] = Field(default_factory=dict)
+
+    # 可选：调参快照（V1 可先只在内部事件使用）
+    config_snapshot: dict[str, Any] = Field(default_factory=dict)
+

server/app/features/personalized_reco/observability/utils.py

@@ -0,0 +1,61 @@
+from __future__ import annotations
+
+from typing import Optional
+
+from app.features.personalized_reco.observability.types import EmptyReason, MissingFields
+
+
+def compute_missing_fields(user_profile: object) -> MissingFields:
+    """
+    判定用户画像缺失字段（对齐算法规则 V1.2 口径）。
+
+    规则：
+    - need：user_profile.need 为空对象 {} 或不存在
+    - context：user_profile.context 为空对象 {} 或不存在
+    - emotion：user_profile.emotion_score 为 None 或不存在
+    """
+
+    need = getattr(user_profile, "need", None)
+    context = getattr(user_profile, "context", None)
+    emotion_score = getattr(user_profile, "emotion_score", None)
+
+    need_missing = not bool(need)
+    context_missing = not bool(context)
+    emotion_missing = emotion_score is None
+
+    return MissingFields(need=need_missing, context=context_missing, emotion=emotion_missing)
+
+
+def compute_empty_reason(
+    *,
+    served_k: int,
+    candidate_pool_size_raw: int,
+    candidate_pool_size_after_hard_filter: int,
+    candidate_pool_size_after_freqcap: int,
+) -> Optional[EmptyReason]:
+    """
+    判定 empty_reason（served_k=0 必填）。
+
+    规则（对齐 plan）：
+    - served_k>0 -> None
+    - raw==0 -> pool_empty
+    - raw>0 且 after_hard_filter==0 -> hard_filter_all
+    - after_freqcap==0 -> freqcap_all
+    - 其他 -> unknown
+    """
+
+    if int(served_k) > 0:
+        return None
+
+    raw = int(candidate_pool_size_raw)
+    after_hard = int(candidate_pool_size_after_hard_filter)
+    after_freqcap = int(candidate_pool_size_after_freqcap)
+
+    if raw == 0:
+        return "pool_empty"
+    if raw > 0 and after_hard == 0:
+        return "hard_filter_all"
+    if after_freqcap == 0:
+        return "freqcap_all"
+    return "unknown"
+

server/app/features/personalized_reco/reco_engine/__init__.py

@@ -0,0 +1,11 @@
+"""
+Reco Engine（推荐引擎编排）。
+
+该模块负责将候选拉取、硬过滤、软打分、重排/频控、回退梯度串成一个稳定 Pipeline，
+并输出统一结构：items + meta（可观测字段）。
+"""
+
+from app.features.personalized_reco.reco_engine.orchestrator import recommend
+
+__all__ = ["recommend"]
+

server/app/features/personalized_reco/reco_engine/defaults.py

@@ -0,0 +1,14 @@
+from __future__ import annotations
+
+from app.features.personalized_reco.reco_engine.types import RecoEngineConfig, Scene
+
+
+def get_default_engine_config(scene: Scene) -> RecoEngineConfig:
+    """
+    获取推荐引擎默认配置（返回副本，避免被意外修改）。
+    """
+
+    # V1：三种场景目前共用一套默认值；保留 scene 参数便于后续按场景拆分
+    base = RecoEngineConfig()
+    return RecoEngineConfig.model_validate(base.model_dump())
+

server/app/features/personalized_reco/reco_engine/hard_filter.py

@@ -0,0 +1,128 @@
+from __future__ import annotations
+
+from collections import defaultdict
+from typing import Any, Iterable, Optional
+
+from app.features.personalized_reco.content_repository.types import ContentProfileDTO
+from app.features.personalized_reco.reco_engine.types import HardFilterResult, RecoConstraints, Scene
+
+
+def _user_stage_key(user_profile: object) -> str:
+    """
+    从 user_profile.stage(one-hot) 提取用户阶段。
+    约定：unknown 通常必填，但这里做防御。
+    """
+
+    stage_obj = getattr(user_profile, "stage", None)
+    if stage_obj is None:
+        return "unknown"
+    if getattr(stage_obj, "expecting", 0) == 1:
+        return "expecting"
+    if getattr(stage_obj, "parenting", 0) == 1:
+        return "parenting"
+    return "unknown"
+
+
+def _user_emotion_score(user_profile: object) -> Optional[float]:
+    v = getattr(user_profile, "emotion_score", None)
+    if v is None:
+        return None
+    try:
+        f = float(v)
+    except Exception:
+        return None
+    if f != f:
+        return None
+    return f
+
+
+def _count_hits(counter: dict[str, int], hits: Iterable[str]) -> None:
+    for h in hits:
+        counter[str(h)] += 1
+
+
+def hard_filter(
+    *,
+    scene: Scene,
+    user_profile: object,
+    candidates: list[ContentProfileDTO],
+    constraints: Optional[RecoConstraints] = None,
+) -> HardFilterResult:
+    """
+    Hard Filter（硬过滤）。
+
+    V1：仅实现硬规则集合（不做软惩罚，不做扩展 hard_rules）。
+    """
+
+    cons = constraints or RecoConstraints()
+
+    exclude_author_ids = set([a for a in (cons.exclude_author_ids or []) if a is not None and str(a).strip() != ""])
+    exclude_template_ids = set([t for t in (cons.exclude_template_ids or []) if t is not None and str(t).strip() != ""])
+    exclude_content_ids = set([int(x) for x in (cons.exclude_content_ids or []) if x is not None])
+
+    u_stage = _user_stage_key(user_profile)
+    u_emotion = _user_emotion_score(user_profile)
+    emotion_low = u_emotion is not None and float(u_emotion) <= 0.2
+
+    kept: list[ContentProfileDTO] = []
+    removed_count = 0
+
+    # 统计：按命中 key 聚合计数（risk_flags 直接用 flag 字符串；跨维度/约束用 rule:* / constraint:* 前缀）
+    hit_counts: dict[str, int] = defaultdict(int)
+    hits_by_content_id: dict[int, list[str]] = {}
+
+    for c in candidates or []:
+        cid = int(c.content_id)
+        hits: list[str] = []
+
+        # 约束：按 content_id/author_id/template_id 排除（视为硬过滤）
+        if cid in exclude_content_ids:
+            hits.append("constraint:exclude_content_id")
+        if c.author_id and c.author_id in exclude_author_ids:
+            hits.append("constraint:exclude_author_id")
+        if c.template_id and c.template_id in exclude_template_ids:
+            hits.append("constraint:exclude_template_id")
+
+        flags = set([str(x) for x in (c.risk_flags or []) if x is not None and str(x).strip() != ""])
+
+        # 全场景必挡
+        if "block_health_medical" in flags:
+            hits.append("block_health_medical")
+
+        # 与用户阶段相关
+        if u_stage == "unknown" and "unsafe_for_stage_unknown" in flags:
+            hits.append("unsafe_for_stage_unknown")
+        if u_stage == "parenting" and "unsafe_for_stage_parenting" in flags:
+            hits.append("unsafe_for_stage_parenting")
+
+        # 与用户情绪相关
+        if emotion_low and "unsafe_for_emotion_low" in flags:
+            hits.append("unsafe_for_emotion_low")
+
+        # 跨维度规则：unknown stage + parenting_pressure 强命中 + 高个性化
+        if u_stage == "unknown":
+            try:
+                need_val = float(c.need_suitability.get("parenting_pressure", 0.0))
+            except Exception:
+                need_val = 0.0
+            if need_val >= 1.0 and float(getattr(c, "personalization_power", 0.0)) >= 1.0:
+                hits.append("rule:unknown_stage_parenting_pressure_power1")
+
+        if hits:
+            removed_count += 1
+            # 单条去重后再计数，避免同 key 重复
+            uniq_hits = sorted(set(hits))
+            hits_by_content_id[cid] = uniq_hits
+            _count_hits(hit_counts, uniq_hits)
+            continue
+
+        hits_by_content_id[cid] = []
+        kept.append(c)
+
+    return HardFilterResult(
+        kept_items=kept,
+        removed_count=int(removed_count),
+        risk_filtered_count_by_flag=dict(hit_counts),
+        hits_by_content_id=hits_by_content_id,
+    )
+

server/app/features/personalized_reco/reco_engine/orchestrator.py

@@ -0,0 +1,396 @@
+from __future__ import annotations
+
+import logging
+from collections import defaultdict
+from datetime import datetime
+from typing import Any, Optional
+
+from app.features.personalized_reco.content_repository.interface import ContentRepository
+from app.features.personalized_reco.content_repository.types import ContentProfileDTO, normalize_locale
+from app.features.personalized_reco.observability.builder import RecoMetaBuilder
+from app.features.personalized_reco.reco_engine.defaults import get_default_engine_config
+from app.features.personalized_reco.reco_engine.hard_filter import hard_filter
+from app.features.personalized_reco.reco_engine.types import RecoConstraints, RecoEngineConfig, RecoEngineResult, RecommendedItem, Scene
+from app.features.personalized_reco.reco_engine.utils import (
+    clamp_personalization_power,
+    merge_exclude_ids,
+    normalize_or_default_locale,
+)
+from app.features.personalized_reco.rerank_freqcap.rerank import rerank_and_freqcap
+from app.features.personalized_reco.rerank_freqcap.types import ScoredCandidate
+from app.features.personalized_reco.scoring.defaults import get_default_config as get_default_score_config
+from app.features.personalized_reco.scoring.score import score_content
+
+logger = logging.getLogger(__name__)
+
+
+def _safe_int(value: Any, *, default: int = 0) -> int:
+    try:
+        n = int(value)
+    except Exception:
+        return int(default)
+    return int(n)
+
+
+def _light_score_summary(score_result: Any) -> dict[str, Any]:
+    """
+    轻量 explanations：只保留少量关键字段，避免 payload 过大。
+    """
+
+    bd = getattr(score_result, "breakdown", None)
+    if bd is None:
+        return {}
+
+    def _get(name: str) -> Optional[float]:
+        v = getattr(bd, name, None)
+        if v is None:
+            return None
+        try:
+            f = float(v)
+        except Exception:
+            return None
+        if f != f:
+            return None
+        return f
+
+    out: dict[str, Any] = {
+        "missing_fields": list(getattr(bd, "missing_fields", []) or []),
+        "S_core": _get("S_core"),
+        "S_personal": _get("S_personal"),
+        "P_uncertainty": _get("P_uncertainty"),
+        "P_risk": _get("P_risk"),
+        "P_widget_emotion_out_of_range": _get("P_widget_emotion_out_of_range"),
+    }
+    # 删除 None，减少噪音
+    return {k: v for k, v in out.items() if v is not None and v != []}
+
+
+def _apply_fallback_level_to_content(content: ContentProfileDTO, *, fallback_level: int) -> ContentProfileDTO:
+    """
+    对内容做防御式一致性处理（与回退梯度一致）。
+    """
+
+    p2 = clamp_personalization_power(content.personalization_power, fallback_level=fallback_level)
+    if p2 == content.personalization_power:
+        return content
+    return content.model_copy(update={"personalization_power": float(p2)})
+
+
+def _merge_counter(dst: dict[str, int], src: dict[str, Any]) -> None:
+    for k, v in (src or {}).items():
+        try:
+            n = int(v)
+        except Exception:
+            n = 0
+        dst[str(k)] = int(dst.get(str(k), 0)) + max(0, int(n))
+
+
+async def recommend(
+    *,
+    repo: ContentRepository,
+    scene: Scene,
+    user_profile: object,
+    already_recommended_ids: list[Any],
+    touched_or_viewed_ids: list[Any],
+    k: int,
+    now: datetime,
+    locale: Optional[str] = None,
+    constraints: Optional[RecoConstraints] = None,
+    config: Optional[RecoEngineConfig] = None,
+) -> RecoEngineResult:
+    """
+    Reco Engine 主入口：编排候选→过滤→打分→重排→回退，并输出 items + meta。
+    """
+
+    cfg = config or get_default_engine_config(scene)
+    cons = constraints or RecoConstraints()
+
+    k_i = max(0, _safe_int(k, default=0))
+    meta_builder = RecoMetaBuilder(scene=scene, user_profile=user_profile, k=k_i, now=now)
+
+    if k_i <= 0:
+        meta_builder.set_candidate_pool_size_raw(0).set_after_hard_filter(0).set_after_dedup(0).set_after_freqcap(0).set_served_k(0).set_fallback_level_final(0)
+        meta_builder.set_config_snapshot({"engine_note": "k<=0，直接返回空结果"})
+        return RecoEngineResult(items=[], meta=meta_builder.build())
+
+    # locale：默认 en；严格校验仅支持 en/tc
+    raw_locale = normalize_or_default_locale(locale)
+    try:
+        effective_locale = normalize_locale(raw_locale)
+    except Exception as e:
+        meta_builder.set_config_snapshot({"error": str(e), "stage": "normalize_locale", "locale": raw_locale})
+        meta_builder.set_candidate_pool_size_raw(0).set_after_hard_filter(0).set_after_dedup(0).set_after_freqcap(0).set_served_k(0).set_fallback_level_final(0)
+        return RecoEngineResult(items=[], meta=meta_builder.build())
+
+    # 聚合统计（跨回退层级累加，确保 meta 单调性成立）
+    raw_total = 0
+    after_hard_total = 0
+    after_dedup_total = 0
+    after_freqcap_total = 0
+
+    risk_counts_total: dict[str, int] = defaultdict(int)
+    freqcap_counts_total: dict[str, int] = defaultdict(int)
+
+    fallback_trace: list[dict[str, Any]] = []
+    selected: list[ScoredCandidate] = []
+    selected_level_by_id: dict[int, int] = {}
+
+    last_fallback_level = 0
+    last_reason = None
+
+    for level in [0, 1, 2, 3]:
+        last_fallback_level = int(level)
+        k_remaining = max(0, k_i - len(selected))
+        if k_remaining <= 0:
+            break
+
+        # Feed：允许不足且不补齐时，拿到任何结果就停止
+        if scene == "feed" and cfg.feed_allow_partial and (not cfg.feed_fill_with_fallback) and len(selected) > 0:
+            break
+
+        # exclude_ids：already/touched + constraints.exclude + 已选内容（避免跨层重复）
+        exclude_ids = merge_exclude_ids(
+            already_recommended_ids=list(already_recommended_ids or []) + [int(x.content_id) for x in selected],
+            touched_or_viewed_ids=list(touched_or_viewed_ids or []),
+            extra_exclude_content_ids=list(cons.exclude_content_ids or []),
+        )
+
+        multiplier = int(cfg.candidate_multiplier_feed if scene == "feed" else cfg.candidate_multiplier_push_widget)
+        base_limit = max(int(cfg.min_candidates_per_level), int(k_remaining) * max(1, int(multiplier)))
+        if cons.max_candidates_limit is not None and int(cons.max_candidates_limit) > 0:
+            limit = min(base_limit, int(cons.max_candidates_limit))
+        else:
+            limit = base_limit
+
+        # 1) Candidate
+        try:
+            cands = await repo.fetch_candidates(
+                scene=scene,
+                user_profile=user_profile,
+                fallback_level=int(level),
+                limit=int(limit),
+                locale=str(effective_locale),
+                exclude_content_ids=exclude_ids,
+            )
+        except Exception as e:
+            logger.exception("fetch_candidates 失败：%s", e)
+            last_reason = "error:fetch_candidates"
+            fallback_trace.append(
+                {
+                    "level": int(level),
+                    "raw": 0,
+                    "after_hard": 0,
+                    "after_dedup": 0,
+                    "after_freqcap": 0,
+                    "served_total": len(selected),
+                    "error": str(e),
+                }
+            )
+            continue
+
+        raw_total += len(cands)
+
+        if not cands:
+            last_reason = "pool_empty"
+            fallback_trace.append(
+                {
+                    "level": int(level),
+                    "raw": 0,
+                    "after_hard": 0,
+                    "after_dedup": 0,
+                    "after_freqcap": 0,
+                    "served_total": len(selected),
+                    "reason": "pool_empty",
+                }
+            )
+            continue
+
+        # 2) Hard Filter
+        hf = hard_filter(scene=scene, user_profile=user_profile, candidates=cands, constraints=cons)
+        kept = [x for x in hf.kept_items if isinstance(x, ContentProfileDTO)]
+        after_hard_total += len(kept)
+        _merge_counter(risk_counts_total, hf.risk_filtered_count_by_flag)
+
+        if not kept:
+            last_reason = "hard_filter_all"
+            fallback_trace.append(
+                {
+                    "level": int(level),
+                    "raw": len(cands),
+                    "after_hard": 0,
+                    "after_dedup": 0,
+                    "after_freqcap": 0,
+                    "served_total": len(selected),
+                    "reason": "hard_filter_all",
+                }
+            )
+            continue
+
+        # 3) Soft Scoring
+        score_cfg = get_default_score_config(scene)
+        if scene == "push":
+            # Push：强制启用不确定性惩罚（与 spec 对齐）
+            score_cfg = score_cfg.model_copy(update={"enable_uncertainty_penalty": True})
+
+        scored: list[ScoredCandidate] = []
+        for c in kept:
+            c2 = _apply_fallback_level_to_content(c, fallback_level=int(level))
+            try:
+                s = score_content(scene=scene, user_profile=user_profile, content_profile=c2, config=score_cfg, pass_filters=True, now=now)
+            except Exception as e:
+                # 单条异常不影响整体
+                logger.exception("score_content 失败 content_id=%s：%s", getattr(c2, "content_id", None), e)
+                continue
+
+            cid = int(c2.content_id)
+            hits = hf.hits_by_content_id.get(cid, [])
+            extra: dict[str, Any] = {
+                "text": c2.text,
+                "fallback_level_used": int(level),
+            }
+            if cfg.enable_explanations:
+                extra["hard_filter_hits"] = hits
+                extra["score_summary"] = _light_score_summary(s)
+
+            scored.append(
+                ScoredCandidate(
+                    content_id=cid,
+                    final_score=float(getattr(s, "final_score", 0.0)),
+                    author_id=c2.author_id,
+                    template_id=c2.template_id,
+                    content_profile=c2,
+                    extra=extra,
+                )
+            )
+
+        if not scored:
+            last_reason = "empty_after_scoring"
+            fallback_trace.append(
+                {
+                    "level": int(level),
+                    "raw": len(cands),
+                    "after_hard": len(kept),
+                    "after_dedup": 0,
+                    "after_freqcap": 0,
+                    "served_total": len(selected),
+                    "reason": "empty_after_scoring",
+                }
+            )
+            continue
+
+        # 4) Rerank/Freqcap
+        try:
+            rer = rerank_and_freqcap(
+                scene=scene,
+                scored_candidates=scored,
+                already_recommended_ids=list(already_recommended_ids or []) + [int(x.content_id) for x in selected],
+                touched_or_viewed_ids=list(touched_or_viewed_ids or []),
+                k=int(k_remaining),
+                recent_author_ids=cons.recent_author_ids,
+                recent_template_ids=cons.recent_template_ids,
+            )
+        except Exception as e:
+            logger.exception("rerank_and_freqcap 失败：%s", e)
+            last_reason = "error:rerank_and_freqcap"
+            fallback_trace.append(
+                {
+                    "level": int(level),
+                    "raw": len(cands),
+                    "after_hard": len(kept),
+                    "after_dedup": 0,
+                    "after_freqcap": 0,
+                    "served_total": len(selected),
+                    "error": str(e),
+                }
+            )
+            continue
+
+        after_dedup_total += int(rer.meta.candidate_pool_size_after_dedup)
+        after_freqcap_total += int(rer.meta.candidate_pool_size_after_freqcap)
+        _merge_counter(freqcap_counts_total, rer.meta.freqcap_filtered_counts)
+
+        served_level = list(rer.ranked_items or [])[:k_remaining]
+        if not served_level:
+            last_reason = "freqcap_all"
+            fallback_trace.append(
+                {
+                    "level": int(level),
+                    "raw": len(cands),
+                    "after_hard": len(kept),
+                    "after_dedup": int(rer.meta.candidate_pool_size_after_dedup),
+                    "after_freqcap": int(rer.meta.candidate_pool_size_after_freqcap),
+                    "served_total": len(selected),
+                    "reason": "freqcap_all",
+                }
+            )
+            continue
+
+        for it in served_level:
+            cid = int(it.content_id)
+            selected.append(it)
+            selected_level_by_id[cid] = int(level)
+
+        last_reason = None
+        fallback_trace.append(
+            {
+                "level": int(level),
+                "raw": len(cands),
+                "after_hard": len(kept),
+                "after_dedup": int(rer.meta.candidate_pool_size_after_dedup),
+                "after_freqcap": int(rer.meta.candidate_pool_size_after_freqcap),
+                "served_total": len(selected),
+                "served_added": len(served_level),
+            }
+        )
+
+        if len(selected) >= k_i:
+            break
+
+    # 组装输出 items（按 selected 顺序）
+    items: list[RecommendedItem] = []
+    for c in selected[:k_i]:
+        cid = int(c.content_id)
+        text = ""
+        if isinstance(c.extra, dict):
+            text = str(c.extra.get("text") or "")
+
+        explanations = None
+        if cfg.enable_explanations and isinstance(c.extra, dict):
+            explanations = {
+                "fallback_level_used": c.extra.get("fallback_level_used"),
+                "hard_filter_hits": c.extra.get("hard_filter_hits"),
+                "score_summary": c.extra.get("score_summary"),
+            }
+
+        items.append(
+            RecommendedItem(
+                content_id=cid,
+                text=text,
+                final_score=float(c.final_score),
+                fallback_level_final=int(selected_level_by_id.get(cid, last_fallback_level)),
+                explanations=explanations,
+            )
+        )
+
+    served_k = len(items)
+
+    # meta：使用聚合统计，确保单调性约束成立（raw>=after_hard>=after_dedup>=after_freqcap>=served_k）
+    # 注意：聚合统计理论上可能出现 after_* > raw_total（例如 repo 返回重复/异常），此处交由 builder 防御修正
+    meta_builder.set_candidate_pool_size_raw(int(raw_total))
+    meta_builder.set_after_hard_filter(int(after_hard_total), risk_filtered_count_by_flag=dict(risk_counts_total))
+    meta_builder.set_after_dedup(int(after_dedup_total))
+    meta_builder.set_after_freqcap(int(after_freqcap_total), freqcap_filtered_counts=dict(freqcap_counts_total))
+    meta_builder.set_served_k(int(served_k))
+    meta_builder.set_fallback_level_final(int(last_fallback_level), reason=last_reason)
+
+    meta_builder.set_config_snapshot(
+        {
+            "fallback_trace": fallback_trace,
+            "engine_config": cfg.model_dump(),
+            "constraints": cons.model_dump(),
+            "locale": effective_locale,
+        }
+    )
+
+    return RecoEngineResult(items=items, meta=meta_builder.build())
+

server/app/features/personalized_reco/reco_engine/types.py

@@ -0,0 +1,101 @@
+from __future__ import annotations
+
+from datetime import datetime
+from typing import Any, Literal, Optional
+
+from pydantic import BaseModel, Field
+
+from app.features.personalized_reco.observability.types import RecoMeta
+
+Scene = Literal["feed", "push", "widget"]
+
+
+class RecoConstraints(BaseModel):
+    """
+    推荐请求的可选约束（调用方可按需传入）。
+    """
+
+    exclude_content_ids: list[int] = Field(default_factory=list)
+    exclude_author_ids: list[str] = Field(default_factory=list)
+    exclude_template_ids: list[str] = Field(default_factory=list)
+
+    # 候选池上限（用于资源保护）
+    max_candidates_limit: Optional[int] = None
+
+    # Push/Widget 作者/模板冷却窗口内的历史集合（增强频控输入）
+    # 说明：若不提供（None），rerank_freqcap 会记录缺失并跳过该维度过滤
+    recent_author_ids: Optional[list[str]] = None
+    recent_template_ids: Optional[list[str]] = None
+
+
+class RecoEngineConfig(BaseModel):
+    """
+    引擎级配置（V1 可调参项）。
+    """
+
+    # Feed 是否允许 served_k < k（允许不足）
+    feed_allow_partial: bool = True
+    # Feed 是否在不足时继续回退补齐
+    feed_fill_with_fallback: bool = True
+
+    # 候选拉取倍率（limit = min(max_candidates_limit, k * multiplier)）
+    candidate_multiplier_feed: int = 10
+    candidate_multiplier_push_widget: int = 30
+
+    # 每层回退的最大候选数量下限（避免 k=1 但候选过少）
+    min_candidates_per_level: int = 30
+
+    # explanations 默认开启（但应保持轻量）
+    enable_explanations: bool = True
+
+
+class RecommendedItem(BaseModel):
+    """
+    引擎最终下发的推荐项。
+    """
+
+    content_id: int
+    text: str
+    final_score: float
+    fallback_level_final: int
+
+    # 解释信息：默认开启，但建议保持轻量（避免 payload 过大）
+    explanations: Optional[dict[str, Any]] = None
+
+
+class RecoEngineResult(BaseModel):
+    """
+    引擎输出容器：items + meta。
+    """
+
+    items: list[RecommendedItem] = Field(default_factory=list)
+    meta: RecoMeta
+
+
+class HardFilterResult(BaseModel):
+    """
+    Hard Filter 输出。
+    """
+
+    kept_items: list[Any] = Field(default_factory=list)
+    removed_count: int = 0
+    risk_filtered_count_by_flag: dict[str, int] = Field(default_factory=dict)
+    # 每条内容的命中信息（仅用于 explanations；默认可为空）
+    hits_by_content_id: dict[int, list[str]] = Field(default_factory=dict)
+
+
+class RecommendRequest(BaseModel):
+    """
+    内部便捷结构（单测/集成时可用）。
+    """
+
+    scene: Scene
+    user_profile: Any
+    already_recommended_ids: list[Any] = Field(default_factory=list)
+    touched_or_viewed_ids: list[Any] = Field(default_factory=list)
+    k: int = 1
+    now: datetime
+    locale: Optional[str] = None
+    constraints: Optional[RecoConstraints] = None
+    config: Optional[RecoEngineConfig] = None
+

server/app/features/personalized_reco/reco_engine/utils.py

@@ -0,0 +1,90 @@
+from __future__ import annotations
+
+from typing import Any, Iterable, Optional
+
+
+def normalize_int_id_list(mixed_ids: Iterable[Any]) -> list[int]:
+    """
+    将混合类型的 id 列表归一化为 int 列表。
+
+    规则：
+    - int/可转 int 的 str -> int
+    - 其他（None/空字符串/不可解析）忽略
+    """
+
+    out: list[int] = []
+    for x in mixed_ids or []:
+        if x is None:
+            continue
+        if isinstance(x, bool):
+            # 避免 True/False 被当作 1/0
+            continue
+        try:
+            s = str(x).strip()
+            if s == "":
+                continue
+            out.append(int(s))
+        except Exception:
+            continue
+    return out
+
+
+def merge_exclude_ids(
+    *,
+    already_recommended_ids: Iterable[Any],
+    touched_or_viewed_ids: Iterable[Any],
+    extra_exclude_content_ids: Optional[Iterable[int]] = None,
+) -> list[int]:
+    """
+    合并并去重排除 id（保持首次出现顺序）。
+    """
+
+    merged = list(normalize_int_id_list(list(already_recommended_ids or []) + list(touched_or_viewed_ids or [])))
+    if extra_exclude_content_ids:
+        merged += [int(x) for x in extra_exclude_content_ids if x is not None]
+
+    seen: set[int] = set()
+    out: list[int] = []
+    for cid in merged:
+        if cid in seen:
+            continue
+        seen.add(cid)
+        out.append(cid)
+    return out
+
+
+def normalize_or_default_locale(locale: Optional[str]) -> str:
+    """
+    locale 防御式归一化：
+    - 未传/空 -> 默认 "en"
+    - 其他 -> 原样返回，由下游 normalize_locale 做严格校验
+    """
+
+    if locale is None:
+        return "en"
+    raw = str(locale).strip()
+    return raw or "en"
+
+
+def clamp_personalization_power(power: Any, *, fallback_level: int) -> float:
+    """
+    按回退层级对 personalization_power 做防御式约束。
+
+    - L0：不改
+    - L1：<= 0.5
+    - L2/L3：= 0
+    """
+
+    try:
+        p = float(power)
+    except Exception:
+        p = 0.0
+    if p != p:
+        p = 0.0
+
+    if int(fallback_level) >= 2:
+        return 0.0
+    if int(fallback_level) >= 1:
+        return min(p, 0.5)
+    return p
+

server/app/features/personalized_reco/rerank_freqcap/__init__.py

@@ -0,0 +1,22 @@
+"""
+个性化推荐｜Rerank & Freqcap 子模块（重排 / 去重 / 频控）
+
+说明（V1）：
+- 本模块在 Soft Scoring 后执行，消费候选的 `final_score`，输出可下发的排序结果。
+- 仅做 Dedup / Freqcap / Feed MMR，不做 Soft Scoring 与 Hard Filter。
+"""
+
+from .defaults import get_default_config
+from .rerank import rerank_and_freqcap
+from .types import RerankConfig, RerankMeta, RerankResult, ScoredCandidate, Scene
+
+__all__ = [
+    "RerankConfig",
+    "RerankMeta",
+    "RerankResult",
+    "ScoredCandidate",
+    "Scene",
+    "get_default_config",
+    "rerank_and_freqcap",
+]
+

server/app/features/personalized_reco/rerank_freqcap/defaults.py

@@ -0,0 +1,41 @@
+from __future__ import annotations
+
+from app.features.personalized_reco.rerank_freqcap.types import RerankConfig, Scene
+
+
+_DEFAULTS: dict[Scene, RerankConfig] = {
+    # Feed：MMR λ=0.7；冷却参数不强制使用
+    "feed": RerankConfig(
+        mmr_lambda=0.7,
+        top_n_for_mmr=200,
+        cooldown_sentence_days=0,
+        cooldown_author_days=0,
+        cooldown_template_days=0,
+    ),
+    # Push：工程默认（来自算法规则的建议参数）
+    "push": RerankConfig(
+        mmr_lambda=0.7,
+        top_n_for_mmr=200,
+        cooldown_sentence_days=14,
+        cooldown_author_days=7,
+        cooldown_template_days=7,
+    ),
+    # Widget：工程默认
+    "widget": RerankConfig(
+        mmr_lambda=0.7,
+        top_n_for_mmr=200,
+        cooldown_sentence_days=7,
+        cooldown_author_days=7,
+        cooldown_template_days=7,
+    ),
+}
+
+
+def get_default_config(scene: Scene) -> RerankConfig:
+    """
+    获取指定场景的默认参数（返回副本，避免被意外修改）。
+    """
+
+    base = _DEFAULTS[scene]
+    return RerankConfig.model_validate(base.model_dump())
+

server/app/features/personalized_reco/rerank_freqcap/rerank.py

@@ -0,0 +1,208 @@
+from __future__ import annotations
+
+from typing import Any, Iterable, Optional
+
+from app.features.personalized_reco.rerank_freqcap.defaults import get_default_config
+from app.features.personalized_reco.rerank_freqcap.types import RerankConfig, RerankMeta, RerankResult, ScoredCandidate, Scene
+from app.features.personalized_reco.rerank_freqcap.utils import as_finite_float, build_tags, clamp, jaccard, normalize_int_id_set
+
+
+def _sort_by_score_desc(cands: list[ScoredCandidate]) -> list[ScoredCandidate]:
+    return sorted(cands, key=lambda x: as_finite_float(x.final_score, default=float("-inf")), reverse=True)
+
+
+def _dedup_by_seen_ids(
+    cands: list[ScoredCandidate],
+    *,
+    seen_ids: set[int],
+) -> tuple[list[ScoredCandidate], int]:
+    kept: list[ScoredCandidate] = []
+    removed = 0
+    for c in cands:
+        if int(c.content_id) in seen_ids:
+            removed += 1
+            continue
+        kept.append(c)
+    return kept, removed
+
+
+def _apply_author_template_freqcap(
+    cands: list[ScoredCandidate],
+    *,
+    recent_author_ids: Optional[Iterable[str]],
+    recent_template_ids: Optional[Iterable[str]],
+) -> tuple[list[ScoredCandidate], dict[str, int], list[str]]:
+    """
+    V1 策略：
+    - 若 recent_*_ids 未提供（None），不执行该维度过滤，但在 meta 记录缺失
+    - 若提供，则执行硬过滤
+    """
+
+    filtered_counts: dict[str, int] = {"author": 0, "template": 0}
+    missing: list[str] = []
+
+    author_set: set[str] | None
+    if recent_author_ids is None:
+        author_set = None
+        missing.append("author")
+    else:
+        author_set = set([a for a in recent_author_ids if a is not None and str(a).strip() != ""])
+
+    template_set: set[str] | None
+    if recent_template_ids is None:
+        template_set = None
+        missing.append("template")
+    else:
+        template_set = set([t for t in recent_template_ids if t is not None and str(t).strip() != ""])
+
+    out: list[ScoredCandidate] = []
+    for c in cands:
+        if author_set is not None and c.author_id and c.author_id in author_set:
+            filtered_counts["author"] += 1
+            continue
+        if template_set is not None and c.template_id and c.template_id in template_set:
+            filtered_counts["template"] += 1
+            continue
+        out.append(c)
+
+    # 只返回真正生效的维度计数（避免 meta 噪音）
+    effective_counts: dict[str, int] = {}
+    if author_set is not None:
+        effective_counts["author"] = int(filtered_counts["author"])
+    if template_set is not None:
+        effective_counts["template"] = int(filtered_counts["template"])
+
+    missing_sorted = sorted(set(missing))
+    return out, effective_counts, missing_sorted
+
+
+def _sim(a: ScoredCandidate, b: ScoredCandidate, *, tags_a: set[str], tags_b: set[str]) -> float:
+    # 离散特征版（V1 推荐），对齐 plan.md
+    if int(a.content_id) == int(b.content_id):
+        return 1.0
+
+    sim = 0.0
+    if a.template_id and b.template_id and a.template_id == b.template_id:
+        sim += 0.6
+    if a.author_id and b.author_id and a.author_id == b.author_id:
+        sim += 0.3
+
+    sim += 0.1 * jaccard(tags_a, tags_b)
+    return clamp(sim, 0.0, 1.0)
+
+
+def _mmr_rerank(
+    *,
+    candidates: list[ScoredCandidate],
+    k: int,
+    lam: float,
+) -> list[ScoredCandidate]:
+    if k <= 0:
+        return []
+
+    if not candidates:
+        return []
+
+    lam_f = clamp(as_finite_float(lam, default=0.7), 0.0, 1.0)
+
+    # 预计算 tags，避免重复构造
+    tags_map: dict[int, set[str]] = {}
+    for c in candidates:
+        tags_map[int(c.content_id)] = build_tags(c)
+
+    remaining = _sort_by_score_desc(list(candidates))
+    selected: list[ScoredCandidate] = []
+
+    # Top1：最高分
+    selected.append(remaining.pop(0))
+
+    while remaining and len(selected) < k:
+        best_idx = 0
+        best_val = float("-inf")
+
+        for idx, c in enumerate(remaining):
+            rel = as_finite_float(c.final_score, default=float("-inf"))
+
+            tags_c = tags_map.get(int(c.content_id), set())
+            max_sim = 0.0
+            for s in selected:
+                tags_s = tags_map.get(int(s.content_id), set())
+                max_sim = max(max_sim, _sim(c, s, tags_a=tags_c, tags_b=tags_s))
+
+            val = lam_f * float(rel) - (1.0 - lam_f) * float(max_sim)
+            if val > best_val:
+                best_val = val
+                best_idx = idx
+
+        selected.append(remaining.pop(best_idx))
+
+    return selected
+
+
+def rerank_and_freqcap(
+    *,
+    scene: Scene,
+    scored_candidates: list[ScoredCandidate],
+    already_recommended_ids: list[Any],
+    touched_or_viewed_ids: list[Any],
+    k: int,
+    config: Optional[RerankConfig] = None,
+    recent_author_ids: Optional[list[str]] = None,
+    recent_template_ids: Optional[list[str]] = None,
+) -> RerankResult:
+    """
+    主入口：对 scored_candidates 做去重/频控/重排，输出最终可下发序列。
+
+    V1 约定：
+    - 冷却窗口“按天”由调用方保证输入集合已经裁剪到窗口内，本模块以“集合代表窗口内历史”为准
+    - Feed 默认只做 dedup + MMR；Push/Widget 做 dedup + freqcap + TopK
+    """
+
+    cfg = config or get_default_config(scene)
+
+    # seen_ids = already_recommended_ids ∪ touched_or_viewed_ids
+    seen_ids = normalize_int_id_set(list(already_recommended_ids) + list(touched_or_viewed_ids))
+
+    # 先按分数降序，保证 Top1 与 TopK 一致
+    base_sorted = _sort_by_score_desc(list(scored_candidates))
+
+    after_dedup, removed_sentence = _dedup_by_seen_ids(base_sorted, seen_ids=seen_ids)
+    candidate_pool_size_after_dedup = len(after_dedup)
+
+    missing_history_fields: list[str] = []
+    freqcap_counts: dict[str, int] = {"sentence": int(removed_sentence)}
+
+    after_freqcap = after_dedup
+
+    # Push/Widget：作者/模板冷却（增强项）
+    if scene in {"push", "widget"}:
+        after_freqcap, dim_counts, missing = _apply_author_template_freqcap(
+            after_freqcap,
+            recent_author_ids=recent_author_ids,
+            recent_template_ids=recent_template_ids,
+        )
+        missing_history_fields = missing
+        freqcap_counts.update(dim_counts)
+    else:
+        # Feed：不强制作者/模板冷却（V1 可选，这里默认跳过）
+        missing_history_fields = []
+
+    candidate_pool_size_after_freqcap = len(after_freqcap)
+
+    ranked: list[ScoredCandidate]
+    if scene == "feed":
+        # MMR 前截断，避免性能问题
+        top_n = int(cfg.top_n_for_mmr) if int(cfg.top_n_for_mmr) > 0 else len(after_freqcap)
+        mmr_pool = after_freqcap[:top_n]
+        ranked = _mmr_rerank(candidates=mmr_pool, k=int(k), lam=cfg.mmr_lambda)
+    else:
+        ranked = after_freqcap[: max(0, int(k))]
+
+    meta = RerankMeta(
+        candidate_pool_size_after_dedup=int(candidate_pool_size_after_dedup),
+        candidate_pool_size_after_freqcap=int(candidate_pool_size_after_freqcap),
+        missing_history_fields=missing_history_fields,
+        freqcap_filtered_counts=freqcap_counts,
+    )
+    return RerankResult(ranked_items=ranked, meta=meta)
+

server/app/features/personalized_reco/rerank_freqcap/types.py

@@ -0,0 +1,61 @@
+from __future__ import annotations
+
+from typing import Any, Literal, Optional
+
+from pydantic import BaseModel, Field
+
+from app.features.personalized_reco.content_repository.types import ContentProfileDTO
+
+Scene = Literal["feed", "push", "widget"]
+
+
+class ScoredCandidate(BaseModel):
+    """
+    Soft Scoring 后的候选项（本模块消费的最小字段集合）。
+
+    说明：
+    - `content_profile` 用于 Feed 的标签/相似度计算；缺失时需降级为仅使用 author/template 等字段
+    """
+
+    content_id: int
+    final_score: float
+
+    author_id: Optional[str] = None
+    template_id: Optional[str] = None
+
+    content_profile: Optional[ContentProfileDTO] = None
+
+    # 允许透传额外字段（例如 text、breakdown 等），便于上层直接下发
+    extra: dict[str, Any] = Field(default_factory=dict)
+
+
+class RerankConfig(BaseModel):
+    """
+    重排/频控配置（可调参）。
+    """
+
+    # Feed：MMR
+    mmr_lambda: float = 0.7
+    top_n_for_mmr: int = 200
+
+    # Push/Widget：冷却窗口（V1 主要用于配置与可观测；真正按天需要带时间戳的历史）
+    cooldown_sentence_days: int = 14
+    cooldown_author_days: int = 7
+    cooldown_template_days: int = 7
+
+
+class RerankMeta(BaseModel):
+    candidate_pool_size_after_dedup: int
+    candidate_pool_size_after_freqcap: int
+
+    # 例如未提供 recent_author_ids/recent_template_ids 时记录 ["author","template"]
+    missing_history_fields: list[str] = Field(default_factory=list)
+
+    # 可选但建议：按维度统计被过滤数量
+    freqcap_filtered_counts: dict[str, int] = Field(default_factory=dict)
+
+
+class RerankResult(BaseModel):
+    ranked_items: list[ScoredCandidate] = Field(default_factory=list)
+    meta: RerankMeta
+

server/app/features/personalized_reco/rerank_freqcap/utils.py

@@ -0,0 +1,107 @@
+from __future__ import annotations
+
+import logging
+from typing import Any, Iterable
+
+logger = logging.getLogger(__name__)
+
+
+def clamp(value: float, min_value: float, max_value: float) -> float:
+    if value != value:  # NaN
+        return min_value
+    return max(min_value, min(max_value, value))
+
+
+def as_finite_float(value: Any, *, default: float) -> float:
+    try:
+        f = float(value)
+    except Exception:
+        return float(default)
+    if f != f:
+        return float(default)
+    if f == float("inf") or f == float("-inf"):
+        return float(default)
+    return f
+
+
+def normalize_int_id_set(values: Iterable[Any]) -> set[int]:
+    """
+    将历史 ID 列表归一化为 int 集合（支持 str/int 混用）。
+
+    说明：
+    - 无法转换的值会被忽略，并记录 debug 日志（不影响主流程）
+    """
+
+    out: set[int] = set()
+    for v in values:
+        try:
+            if isinstance(v, bool):
+                # 避免 True/False 被当作 1/0
+                raise ValueError("bool 不是合法 id")
+            out.add(int(v))
+        except Exception:
+            logger.debug("历史 id 无法转为 int，已忽略：%r", v)
+    return out
+
+
+def jaccard(a: set[str], b: set[str]) -> float:
+    if not a and not b:
+        return 0.0
+    inter = len(a & b)
+    union = len(a | b)
+    return float(inter) / float(union) if union > 0 else 0.0
+
+
+def argmax_key(d: dict[str, Any] | None) -> str | None:
+    """
+    从 suitability 字典中取最大值 key（V1 用作代表标签）。
+    - 空字典/None -> None
+    - 值非法 -> 按 default=0 处理
+    """
+
+    if not d:
+        return None
+    best_k: str | None = None
+    best_v = float("-inf")
+    for k, v in d.items():
+        fv = as_finite_float(v, default=0.0)
+        if fv > best_v:
+            best_v = fv
+            best_k = k
+    return best_k
+
+
+def build_tags(candidate: Any) -> set[str]:
+    """
+    构造离散标签集合（V1 写死）：
+    - stage:<stage>
+    - need:<argmax_key>
+    - context:<argmax_key>
+
+    说明：
+    - candidate 可能是 ScoredCandidate 或具备 content_profile 的对象
+    - 字段缺失时自动降级（只返回可得标签）
+    """
+
+    tags: set[str] = set()
+
+    cp = getattr(candidate, "content_profile", None)
+    if cp is None:
+        return tags
+
+    stage = getattr(cp, "stage", None)
+    if stage:
+        tags.add(f"stage:{stage}")
+
+    need = getattr(cp, "need_suitability", None)
+    need_k = argmax_key(need)
+    if need_k:
+        tags.add(f"need:{need_k}")
+
+    ctx = getattr(cp, "context_suitability", None)
+    ctx_k = argmax_key(ctx)
+    if ctx_k:
+        tags.add(f"context:{ctx_k}")
+
+    return tags
+

server/app/features/personalized_reco/scoring/__init__.py

@@ -0,0 +1,22 @@
+"""
+个性化推荐｜Scoring 子模块（软打分与惩罚项）
+
+说明：
+- 本模块只做软打分与本模块定义的惩罚项（P_uncertainty、Widget 情绪软降权）。
+- Hard Filter / 频控重排 / 新鲜度等由其他模块产出，通过入参注入（缺省按 0）。
+"""
+
+from .defaults import get_default_config
+from .score import score_content
+from .types import ExternalTerms, Scene, ScoreBreakdown, ScoreConfig, ScoreResult
+
+__all__ = [
+    "ExternalTerms",
+    "Scene",
+    "ScoreBreakdown",
+    "ScoreConfig",
+    "ScoreResult",
+    "get_default_config",
+    "score_content",
+]
+

server/app/features/personalized_reco/scoring/defaults.py

@@ -0,0 +1,44 @@
+from __future__ import annotations
+
+from app.features.personalized_reco.scoring.types import Scene, ScoreConfig
+
+
+_DEFAULTS: dict[Scene, ScoreConfig] = {
+    # 来源：设计说明文档/個性化推薦算法規則.md（V1 建议权重）
+    "feed": ScoreConfig(
+        w_need=0.35,
+        w_emotion=0.20,
+        w_stage=0.15,
+        w_context=0.30,
+        # Feed 默认不启用不确定性惩罚（可按需开启）
+        enable_uncertainty_penalty=False,
+    ),
+    "push": ScoreConfig(
+        w_need=0.45,
+        w_emotion=0.35,
+        w_stage=0.15,
+        w_context=0.05,
+        # Push 默认启用不确定性惩罚
+        enable_uncertainty_penalty=True,
+    ),
+    "widget": ScoreConfig(
+        w_need=0.25,
+        w_emotion=0.25,
+        w_stage=0.30,
+        w_context=0.20,
+        # Widget 默认不启用不确定性惩罚（可按需开启）
+        enable_uncertainty_penalty=False,
+        widget_emotion_soft_range=(0.4, 0.8),
+        widget_emotion_penalty_gamma=0.25,
+    ),
+}
+
+
+def get_default_config(scene: Scene) -> ScoreConfig:
+    """
+    获取指定场景的默认打分参数（返回副本，避免被意外修改）。
+    """
+
+    base = _DEFAULTS[scene]
+    return ScoreConfig.model_validate(base.model_dump())
+

server/app/features/personalized_reco/scoring/score.py

@@ -0,0 +1,201 @@
+from __future__ import annotations
+
+from datetime import datetime
+from typing import Optional
+
+from app.features.personalized_reco.content_repository.types import ContentProfileDTO
+from app.features.personalized_reco.scoring.defaults import get_default_config
+from app.features.personalized_reco.scoring.types import ExternalTerms, Scene, ScoreBreakdown, ScoreConfig, ScoreResult
+from app.features.personalized_reco.scoring.utils import as_finite_float, clamp, pick_one_hot_key
+from app.features.user_profile_scoring.types import UserProfileV1_2
+
+
+def _missing_fields(user_profile: UserProfileV1_2) -> list[str]:
+    missing: list[str] = []
+    if not user_profile.need:
+        missing.append("need")
+    if not user_profile.context:
+        missing.append("context")
+    if user_profile.emotion_score is None:
+        missing.append("emotion")
+    return missing
+
+
+def _score_need(user_profile: UserProfileV1_2, content: ContentProfileDTO) -> float:
+    key = pick_one_hot_key(user_profile.need)  # type: ignore[arg-type]
+    if key is None:
+        return 0.5
+    raw = content.need_suitability.get(key, 0.5)
+    return clamp(as_finite_float(raw, default=0.5), 0.0, 1.0)
+
+
+def _score_context(user_profile: UserProfileV1_2, content: ContentProfileDTO) -> float:
+    key = pick_one_hot_key(user_profile.context)  # type: ignore[arg-type]
+    if key is None:
+        return 0.5
+    raw = content.context_suitability.get(key, 0.5)
+    return clamp(as_finite_float(raw, default=0.5), 0.0, 1.0)
+
+
+def _score_emotion(user_profile: UserProfileV1_2, content: ContentProfileDTO) -> float:
+    # V1.2：用户情绪缺失 -> 0.8
+    if user_profile.emotion_score is None:
+        return 0.8
+
+    # 文案 general（emotion_score=None）-> 0.8
+    if content.emotion_score is None:
+        return 0.8
+
+    u = clamp(as_finite_float(user_profile.emotion_score, default=0.8), 0.0, 1.0)
+    c = clamp(as_finite_float(content.emotion_score, default=0.8), 0.0, 1.0)
+    return clamp(1.0 - abs(u - c), 0.0, 1.0)
+
+
+def _user_stage_key(user_profile: UserProfileV1_2) -> str:
+    # 约定：UserStageOneHot.unknown 必填；但这里仍做防御
+    stage = user_profile.stage
+    if getattr(stage, "expecting", 0) == 1:
+        return "expecting"
+    if getattr(stage, "parenting", 0) == 1:
+        return "parenting"
+    if getattr(stage, "unknown", 1) == 1:
+        return "unknown"
+    return "unknown"
+
+
+def _score_stage(user_profile: UserProfileV1_2, content: ContentProfileDTO) -> float:
+    # 对齐算法规则：
+    # - general=1；命中=1；unknown对非unknown=0.7；其余=0
+    if content.stage == "general":
+        return 1.0
+
+    u_stage = _user_stage_key(user_profile)
+    if content.stage == u_stage:
+        return 1.0
+
+    if u_stage == "unknown" and content.stage != "unknown":
+        return 0.7
+
+    return 0.0
+
+
+def _score_personal(alpha: float, personalization_power: float, s_need: float, s_context: float) -> float:
+    power = clamp(as_finite_float(personalization_power, default=0.0), 0.0, 1.0)
+    a = as_finite_float(alpha, default=0.0)
+    return float(a) * float(power) * max(float(s_need), float(s_context))
+
+
+def _penalty_uncertainty(beta: float, user_profile: UserProfileV1_2, content: ContentProfileDTO) -> float:
+    b = as_finite_float(beta, default=0.0)
+    power = clamp(as_finite_float(content.personalization_power, default=0.0), 0.0, 1.0)
+
+    # V1 约定：conf_U 缺失时按 1.0（避免过惩罚）
+    conf_u = clamp(as_finite_float(getattr(user_profile, "profile_confidence", 1.0), default=1.0), 0.0, 1.0)
+    conf_c = clamp(as_finite_float(getattr(content, "review_confidence", 0.7), default=0.7), 0.0, 1.0)
+
+    return float(b) * (1.0 - float(conf_u)) * (1.0 - float(conf_c)) * float(power)
+
+
+def _widget_emotion_penalty(scene: Scene, content: ContentProfileDTO, config: ScoreConfig) -> float:
+    if scene != "widget":
+        return 0.0
+    if content.emotion_score is None:
+        return 0.0
+
+    lo, hi = config.widget_emotion_soft_range
+    lo_f = as_finite_float(lo, default=0.4)
+    hi_f = as_finite_float(hi, default=0.8)
+    width = hi_f - lo_f
+    if width <= 0:
+        return 0.0
+
+    e = clamp(as_finite_float(content.emotion_score, default=0.6), 0.0, 1.0)
+    if e < lo_f:
+        d = lo_f - e
+    elif e > hi_f:
+        d = e - hi_f
+    else:
+        d = 0.0
+
+    gamma = as_finite_float(config.widget_emotion_penalty_gamma, default=0.25)
+    raw = float(gamma) * float(d) / float(width)
+    return clamp(raw, 0.0, float(gamma))
+
+
+def score_content(
+    *,
+    scene: Scene,
+    user_profile: UserProfileV1_2,
+    content_profile: ContentProfileDTO,
+    config: Optional[ScoreConfig] = None,
+    pass_filters: bool = True,
+    external_terms: Optional[ExternalTerms] = None,
+    now: Optional[datetime] = None,  # 预留：V1 不使用
+) -> ScoreResult:
+    """
+    主入口：对单条内容 Cᵢ 进行软打分，返回 final_score 与 breakdown。
+
+    说明（V1）：
+    - `pass_filters` 来自 Hard Filter（本模块不做硬过滤）
+    - `external_terms` 可注入 S_fresh / P_fatigue / P_repeat / P_risk（缺省按 0）
+    - `now` 预留给未来的 freshness/时间衰减（V1 不实现）
+    """
+
+    cfg = config or get_default_config(scene)
+    ext = external_terms or ExternalTerms()
+
+    missing = _missing_fields(user_profile)
+
+    s_need = _score_need(user_profile, content_profile)
+    s_context = _score_context(user_profile, content_profile)
+    s_emotion = _score_emotion(user_profile, content_profile)
+    s_stage = _score_stage(user_profile, content_profile)
+
+    w_need = as_finite_float(cfg.w_need, default=0.0)
+    w_emotion = as_finite_float(cfg.w_emotion, default=0.0)
+    w_stage = as_finite_float(cfg.w_stage, default=0.0)
+    w_context = as_finite_float(cfg.w_context, default=0.0)
+
+    s_core = float(w_need) * s_need + float(w_emotion) * s_emotion + float(w_stage) * s_stage + float(w_context) * s_context
+
+    s_personal = _score_personal(cfg.alpha, content_profile.personalization_power, s_need, s_context)
+
+    p_uncertainty = 0.0
+    if cfg.enable_uncertainty_penalty:
+        p_uncertainty = _penalty_uncertainty(cfg.beta, user_profile, content_profile)
+
+    p_widget = _widget_emotion_penalty(scene, content_profile, cfg)
+
+    s_fresh = as_finite_float(ext.S_fresh, default=0.0)
+    p_fatigue = as_finite_float(ext.P_fatigue, default=0.0)
+    p_repeat = as_finite_float(ext.P_repeat, default=0.0)
+    p_risk_external = as_finite_float(ext.P_risk, default=0.0)
+
+    # Widget 软降权并入 P_risk（但在 breakdown 中单独暴露，便于打点）
+    p_risk = float(p_risk_external) + float(p_widget)
+
+    raw_final = s_core + s_personal + float(s_fresh) - float(p_fatigue) - float(p_repeat) - float(p_risk) - float(p_uncertainty)
+    final_score = float(raw_final) if pass_filters else 0.0
+
+    breakdown = ScoreBreakdown(
+        scene=scene,
+        **{
+            "pass": bool(pass_filters),
+        },
+        missing_fields=missing,
+        S_need=float(s_need),
+        S_context=float(s_context),
+        S_stage=float(s_stage),
+        S_emotion=float(s_emotion),
+        S_core=float(s_core),
+        S_personal=float(s_personal),
+        S_fresh=float(s_fresh),
+        P_fatigue=float(p_fatigue),
+        P_repeat=float(p_repeat),
+        P_risk=float(p_risk),
+        P_uncertainty=float(p_uncertainty),
+        P_widget_emotion_out_of_range=float(p_widget),
+    )
+
+    return ScoreResult(final_score=float(final_score), breakdown=breakdown)
+

server/app/features/personalized_reco/scoring/types.py

@@ -0,0 +1,84 @@
+from __future__ import annotations
+
+from typing import Literal
+
+from pydantic import BaseModel, Field
+
+Scene = Literal["feed", "push", "widget"]
+
+
+class ScoreConfig(BaseModel):
+    """
+    打分配置（可调参）。
+
+    说明：
+    - 默认值由 `defaults.get_default_config(scene)` 提供
+    - 本模块不负责回退梯度（fallback_level）策略；仅做防御式 clamp
+    """
+
+    w_need: float
+    w_emotion: float
+    w_stage: float
+    w_context: float
+
+    alpha: float = 0.15
+    beta: float = 0.30
+
+    enable_uncertainty_penalty: bool = False
+
+    # Widget 情绪软区间与软降权强度
+    widget_emotion_soft_range: tuple[float, float] = (0.4, 0.8)
+    widget_emotion_penalty_gamma: float = 0.25
+
+
+class ExternalTerms(BaseModel):
+    """
+    外部注入项（V1 可选）。
+
+    说明：
+    - 由 `rerank-freqcap` 或 `reco-engine` 产出
+    - 本模块缺省按 0，保证可排序与输出结构稳定
+    """
+
+    S_fresh: float = 0.0
+    P_fatigue: float = 0.0
+    P_repeat: float = 0.0
+    P_risk: float = 0.0
+
+
+class ScoreBreakdown(BaseModel):
+    """
+    可观测分解项（用于调参与回归测试）。
+    """
+
+    scene: Scene
+    passed: bool = Field(alias="pass")
+
+    missing_fields: list[str] = Field(default_factory=list)
+
+    S_need: float
+    S_context: float
+    S_stage: float
+    S_emotion: float
+
+    S_core: float
+    S_personal: float
+    S_fresh: float
+
+    P_fatigue: float
+    P_repeat: float
+    P_risk: float
+    P_uncertainty: float
+
+    # Widget 专用：区间外软降权（建议保留，便于打点）
+    P_widget_emotion_out_of_range: float = 0.0
+
+    model_config = {
+        "populate_by_name": True,
+    }
+
+
+class ScoreResult(BaseModel):
+    final_score: float
+    breakdown: ScoreBreakdown
+

server/app/features/personalized_reco/scoring/utils.py

@@ -0,0 +1,59 @@
+from __future__ import annotations
+
+import logging
+from typing import Any
+
+logger = logging.getLogger(__name__)
+
+
+def clamp(value: float, min_value: float, max_value: float) -> float:
+    """
+    将值裁剪到区间内，并对 NaN 做兜底。
+    """
+
+    if value != value:  # NaN
+        return min_value
+    return max(min_value, min(max_value, value))
+
+
+def as_finite_float(value: Any, *, default: float) -> float:
+    """
+    将任意值尽量转为有限 float；失败则返回 default。
+    """
+
+    try:
+        f = float(value)
+    except Exception:
+        return float(default)
+
+    # NaN / inf 都视为不可用
+    if f != f:
+        return float(default)
+    if f == float("inf") or f == float("-inf"):
+        return float(default)
+    return f
+
+
+def pick_one_hot_key(one_hot: dict[str, Any] | None) -> str | None:
+    """
+    从稀疏 one-hot（{key: 1}）中取唯一 key。
+
+    约定：
+    - None / {} → 缺失，返回 None
+    - 单 key → 返回该 key
+    - 多 key → 取“字典序最小”的 key，并记录 debug 日志（避免静默歧义）
+    """
+
+    if not one_hot:
+        return None
+
+    keys = [k for k, v in one_hot.items() if v == 1 or v is True]
+    if not keys:
+        return None
+    if len(keys) == 1:
+        return keys[0]
+
+    chosen = sorted(keys)[0]
+    logger.debug("one-hot 出现多个 key=1，已按字典序选择：chosen=%s keys=%s", chosen, keys)
+    return chosen
+

server/app/features/user_profile_scoring/__init__.py

@@ -0,0 +1,9 @@
+"""
+User Profile Scoring（用户画像打分）V1.2
+
+说明：
+- 提供“问卷答案（可跳过）→ 用户画像（可计算、可观测、可版本化）”的服务端实现
+- 规则以 `spec_kit/User Profile Scoring/spec.md`（V1.2）与
+  `设计说明文档/客戶端問卷打分規則.md`（V1.2）为准
+"""
+

server/app/features/user_profile_scoring/scoring.py

@@ -0,0 +1,194 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+from datetime import datetime, timezone
+from typing import Optional
+
+from app.features.user_profile_scoring.types import (
+    HardRules,
+    ProfileAnswered,
+    QuestionnaireAnswersV1_2,
+    UserProfileV1_2_Extended,
+    UserStageOneHot,
+)
+
+
+def _clamp(value: float, min_value: float, max_value: float) -> float:
+    if value != value:  # NaN
+        return min_value
+    return max(min_value, min(max_value, value))
+
+
+def normalize_answers(raw: QuestionnaireAnswersV1_2) -> QuestionnaireAnswersV1_2:
+    """
+    归一化答案：
+    - Pydantic 已对枚举做了校验；此处仅统一 None/缺失的语义为“跳过”
+    """
+
+    # 直接返回一份拷贝，保持纯函数语义
+    return QuestionnaireAnswersV1_2.model_validate(raw.model_dump())
+
+
+def compute_profile_answered(answers: QuestionnaireAnswersV1_2) -> ProfileAnswered:
+    return ProfileAnswered(
+        stage=answers.mom_stage is not None,
+        emotion=answers.emotion is not None,
+        context=answers.context is not None,
+        need=answers.need is not None,
+    )
+
+
+def compute_time_confidence(generated_at: datetime, now: datetime) -> float:
+    """
+    时间衰减置信度（conf_time）
+    - 0–7 天：1.0
+    - 7–30 天：线性衰减到 0.7（含第 30 天）
+    - 30 天以上：0.5
+    """
+
+    delta = (now - generated_at).total_seconds()
+    if delta <= 0:
+        return 1.0
+
+    days = delta / (24 * 60 * 60)
+    if days <= 7:
+        return 1.0
+    if days <= 30:
+        t = (days - 7) / (30 - 7)  # 0..1
+        return 1.0 - 0.3 * t
+    return 0.5
+
+
+def compute_profile_confidence(conf_time: float, answered: ProfileAnswered) -> float:
+    """
+    V1.2：profile_confidence（conf_U）
+    conf = clamp(conf_time * (0.5 + 0.5 * completion), 0.2, 1.0)
+    """
+
+    answered_count = sum(
+        [
+            1 if answered.stage else 0,
+            1 if answered.emotion else 0,
+            1 if answered.context else 0,
+            1 if answered.need else 0,
+        ]
+    )
+    completion = answered_count / 4
+    completion_factor = 0.5 + 0.5 * completion
+    return _clamp(float(conf_time) * float(completion_factor), 0.2, 1.0)
+
+
+def _build_stage_one_hot(mom_stage: Optional[str]) -> UserStageOneHot:
+    # V1.2：mom_stage 跳过按安全策略输出 unknown=1
+    if mom_stage is None:
+        return UserStageOneHot(unknown=1)
+
+    return UserStageOneHot(
+        expecting=1 if mom_stage == "expecting" else 0,
+        parenting=1 if mom_stage == "parenting" else 0,
+        unknown=1 if mom_stage == "unknown" else 0,
+    )
+
+
+def _map_emotion_score(emotion: Optional[str]) -> Optional[float]:
+    if emotion is None:
+        return None
+    mapping = {
+        "low": 0.0,
+        "overwhelmed": 0.2,
+        "tired": 0.4,
+        "neutral": 0.6,
+        "calm": 0.8,
+        "joyful": 1.0,
+    }
+    return mapping.get(emotion)
+
+
+def _build_sparse_one_hot(value: Optional[str]) -> dict[str, int]:
+    if value is None:
+        return {}
+    return {value: 1}
+
+
+@dataclass(frozen=True)
+class _RuleOutput:
+    rule_hits: list[str]
+    hard_rules: HardRules
+
+
+def _compute_rule_output(stage: UserStageOneHot, emotion_score: Optional[float]) -> _RuleOutput:
+    rule_hits: list[str] = []
+    forbidden_risk_flags: list[str] = []
+
+    stage_unknown = stage.unknown == 1
+    stage_parenting = stage.parenting == 1
+
+    if stage_unknown:
+        rule_hits.append("unsafe_for_stage_unknown")
+        forbidden_risk_flags.append("unsafe_for_stage_unknown")
+
+    if stage_parenting:
+        rule_hits.append("unsafe_for_stage_parenting")
+        forbidden_risk_flags.append("unsafe_for_stage_parenting")
+
+    if emotion_score is not None and emotion_score <= 0.2:
+        rule_hits.append("unsafe_for_emotion_low")
+        forbidden_risk_flags.append("unsafe_for_emotion_low")
+
+    forbidden_content_predicates = []
+    if stage_unknown:
+        forbidden_content_predicates.append(
+            {
+                "id": "unknown_block_parenting_pressure_personalized",
+                "when_user": {"stage_unknown": True},
+                "forbid_content": {"need": "parenting_pressure", "personalization_power": 1},
+            }
+        )
+
+    return _RuleOutput(
+        rule_hits=rule_hits,
+        hard_rules=HardRules(
+            forbidden_risk_flags=forbidden_risk_flags,
+            forbidden_content_predicates=forbidden_content_predicates,
+        ),
+    )
+
+
+def build_user_profile_from_questionnaire(
+    raw_answers: QuestionnaireAnswersV1_2,
+    *,
+    generated_at: Optional[datetime] = None,
+    now: Optional[datetime] = None,
+) -> UserProfileV1_2_Extended:
+    """
+    主入口：问卷答案（可跳过）→ 用户画像（V1.2）+ 硬规则输出
+    """
+
+    answers = normalize_answers(raw_answers)
+    answered = compute_profile_answered(answers)
+
+    now_dt = now or datetime.now(tz=timezone.utc)
+    gen_dt = generated_at or now_dt
+
+    conf_time = compute_time_confidence(gen_dt, now_dt)
+    conf_u = compute_profile_confidence(conf_time, answered)
+
+    stage = _build_stage_one_hot(answers.mom_stage)
+    emotion_score = _map_emotion_score(answers.emotion)
+    context = _build_sparse_one_hot(answers.context)
+    need = _build_sparse_one_hot(answers.need)
+
+    rule_out = _compute_rule_output(stage, emotion_score)
+
+    return UserProfileV1_2_Extended(
+        profile_generated_at=gen_dt,
+        profile_confidence=conf_u,
+        profile_answered=answered,
+        stage=stage,
+        emotion_score=emotion_score,
+        context=context,  # type: ignore[arg-type]
+        need=need,  # type: ignore[arg-type]
+        rule_hits=rule_out.rule_hits,
+        hard_rules=rule_out.hard_rules,
+    )
+

server/app/features/user_profile_scoring/types.py

@@ -0,0 +1,89 @@
+from __future__ import annotations
+
+from datetime import datetime
+from typing import Any, Literal, Optional
+
+from pydantic import BaseModel, Field
+
+
+MomStageAnswer = Literal["expecting", "parenting", "unknown"]
+EmotionAnswer = Literal["low", "overwhelmed", "tired", "neutral", "calm", "joyful"]
+ContextAnswer = Literal["family", "work", "relationship", "friends", "health"]
+NeedAnswer = Literal[
+    "emotional_support",
+    "parenting_pressure",
+    "self_worth",
+    "anxiety_relief",
+    "rest_balance",
+]
+
+
+class QuestionnaireAnswersV1_2(BaseModel):
+    """
+    V1.2：每题可跳过
+
+    说明：
+    - `None` 表示题目被跳过/无值（与客户端的 `null` 对齐）
+    - 字段缺失（未传）也视为跳过
+    """
+
+    mom_stage: Optional[MomStageAnswer] = None
+    emotion: Optional[EmotionAnswer] = None
+    context: Optional[ContextAnswer] = None
+    need: Optional[NeedAnswer] = None
+
+
+class ProfileAnswered(BaseModel):
+    stage: bool
+    emotion: bool
+    context: bool
+    need: bool
+
+
+class UserStageOneHot(BaseModel):
+    expecting: Optional[Literal[0, 1]] = None
+    parenting: Optional[Literal[0, 1]] = None
+    unknown: Literal[0, 1]
+
+
+class ForbiddenContentPredicate(BaseModel):
+    """
+    用于表达“需要同时看用户与内容字段才能执行”的规则（跨维度规则）。
+    """
+
+    id: str
+    when_user: dict[str, Any] = Field(default_factory=dict)
+    forbid_content: dict[str, Any] = Field(default_factory=dict)
+
+
+class HardRules(BaseModel):
+    forbidden_risk_flags: list[str] = Field(default_factory=list)
+    forbidden_content_predicates: list[ForbiddenContentPredicate] = Field(default_factory=list)
+
+
+class UserProfileV1_2(BaseModel):
+    profile_version: Literal["v1.2"] = "v1.2"
+    profile_source: Literal["questionnaire"] = "questionnaire"
+    profile_generated_at: datetime
+    profile_confidence: float
+    profile_answered: ProfileAnswered
+    stage: UserStageOneHot
+    emotion_score: Optional[float] = None
+    context: dict[str, Literal[1]] = Field(default_factory=dict)
+    need: dict[str, Literal[1]] = Field(default_factory=dict)
+
+
+class UserProfileV1_2_Extended(UserProfileV1_2):
+    rule_hits: list[str] = Field(default_factory=list)
+    hard_rules: HardRules = Field(default_factory=HardRules)
+
+
+class BuildUserProfileRequest(BaseModel):
+    """
+    API 请求体：问卷答案 + 可选时间注入（便于回归测试/服务端批处理）
+    """
+
+    answers: QuestionnaireAnswersV1_2 = Field(default_factory=QuestionnaireAnswersV1_2)
+    generated_at: Optional[datetime] = None
+    now: Optional[datetime] = None
+

server/app/main.py

@@ -1,6 +1,8 @@
 from fastapi import FastAPI
 
 from app.core.config import get_settings
+from app.api.v1.reco import router as reco_router
+from app.api.v1.user_profile_scoring import router as user_profile_router
 
 
 def create_app() -> FastAPI:
@@ -14,6 +16,10 @@ def create_app() -> FastAPI:
 
     app = FastAPI(title=settings.app_name)
 
+    # 业务路由
+    app.include_router(user_profile_router)
+    app.include_router(reco_router)
+
     @app.get("/healthz")
     async def healthz() -> dict:
         return {"status": "ok", "env": settings.app_env}

server/app/tasks/reco.py

@@ -0,0 +1,168 @@
+from __future__ import annotations
+
+import asyncio
+from datetime import datetime, timezone
+from typing import Any, Optional
+
+from celery import shared_task
+
+from app.db.session import AsyncSessionLocal
+from app.features.personalized_reco.content_repository.sqlalchemy_repo import SqlAlchemyContentRepository
+from app.features.personalized_reco.content_repository.types import normalize_locale
+from app.features.personalized_reco.reco_engine import recommend
+from app.features.personalized_reco.reco_engine.types import RecoConstraints, RecoEngineResult, Scene
+from app.features.user_profile_scoring.types import UserProfileV1_2
+
+
+def _ensure_now(now: Optional[datetime]) -> datetime:
+    if now is None:
+        return datetime.now(timezone.utc)
+    if now.tzinfo is None:
+        return now.replace(tzinfo=timezone.utc)
+    return now
+
+
+def _ensure_locale(locale: Optional[str]) -> str:
+    raw = (locale or "").strip() or "en"
+    # 严格校验只支持 en/tc（允许 en-US 等在 normalize_locale 内归一化）
+    return str(normalize_locale(raw))
+
+
+async def _run_reco_async(
+    *,
+    scene: Scene,
+    user_profile: UserProfileV1_2,
+    already_recommended_ids: list[Any],
+    touched_or_viewed_ids: list[Any],
+    k: int,
+    now: datetime,
+    locale: str,
+) -> RecoEngineResult:
+    async with AsyncSessionLocal() as session:
+        repo = SqlAlchemyContentRepository(session)
+        return await recommend(
+            repo=repo,
+            scene=scene,
+            user_profile=user_profile,
+            already_recommended_ids=list(already_recommended_ids or []),
+            touched_or_viewed_ids=list(touched_or_viewed_ids or []),
+            k=int(k),
+            now=now,
+            locale=locale,
+            constraints=RecoConstraints(),
+        )
+
+
+def _run_reco_sync(
+    *,
+    scene: Scene,
+    user_profile: UserProfileV1_2,
+    already_recommended_ids: list[Any],
+    touched_or_viewed_ids: list[Any],
+    k: int,
+    now: Optional[datetime],
+    locale: Optional[str],
+) -> dict[str, Any]:
+    effective_now = _ensure_now(now)
+    effective_locale = _ensure_locale(locale)
+    result = asyncio.run(
+        _run_reco_async(
+            scene=scene,
+            user_profile=user_profile,
+            already_recommended_ids=already_recommended_ids,
+            touched_or_viewed_ids=touched_or_viewed_ids,
+            k=int(k),
+            now=effective_now,
+            locale=effective_locale,
+        )
+    )
+    # 默认不存结果，但返回值可用于开发调试（worker 通常 ignore_result）
+    return result.model_dump()
+
+
+@shared_task(name="tasks.reco.generate")
+def generate(
+    *,
+    scene: Scene,
+    user_profile: dict[str, Any],
+    already_recommended_ids: Optional[list[Any]] = None,
+    touched_or_viewed_ids: Optional[list[Any]] = None,
+    k: Optional[int] = None,
+    now: Optional[str] = None,
+    locale: Optional[str] = None,
+) -> dict[str, Any]:
+    """
+    推荐生成任务（通用入口）。
+
+    说明：
+    - 入参尽量保持小（避免 Redis 队列膨胀）
+    - 默认 worker 配置为 ignore_result，但这里仍返回结构，便于本地调试
+    """
+
+    # 解析 user_profile（严格按 V1.2）
+    u = UserProfileV1_2.model_validate(user_profile or {})
+
+    # k 默认按场景（与 API 一致）
+    if k is None:
+        k_i = 30 if scene == "feed" else 1
+    else:
+        k_i = int(k)
+
+    # now 支持 ISO 字符串
+    dt: Optional[datetime]
+    if not now:
+        dt = None
+    else:
+        raw = str(now).strip()
+        if raw.endswith("Z"):
+            raw = raw[:-1] + "+00:00"
+        try:
+            dt = datetime.fromisoformat(raw)
+        except Exception:
+            dt = None
+
+    return _run_reco_sync(
+        scene=scene,
+        user_profile=u,
+        already_recommended_ids=list(already_recommended_ids or []),
+        touched_or_viewed_ids=list(touched_or_viewed_ids or []),
+        k=k_i,
+        now=dt,
+        locale=locale,
+    )
+
+
+def _deliver_push_placeholder(payload: dict[str, Any]) -> None:
+    """
+    Push 下游写入占位函数（V1 不接真实推送系统）。
+    """
+
+    _ = payload
+    return None
+
+
+@shared_task(name="tasks.reco.push_once")
+def push_once(
+    *,
+    user_profile: dict[str, Any],
+    already_recommended_ids: Optional[list[Any]] = None,
+    touched_or_viewed_ids: Optional[list[Any]] = None,
+    now: Optional[str] = None,
+    locale: Optional[str] = None,
+) -> dict[str, Any]:
+    """
+    单次 Push 生成（占位任务）。
+    """
+
+    payload = generate(
+        scene="push",
+        user_profile=user_profile,
+        already_recommended_ids=already_recommended_ids,
+        touched_or_viewed_ids=touched_or_viewed_ids,
+        k=1,
+        now=now,
+        locale=locale,
+    )
+    _deliver_push_placeholder(payload)
+    return payload
+

server/requirements.txt

@@ -4,6 +4,8 @@ uvicorn[standard]>=0.27
 # 数据库（SQLAlchemy 2.x 异步 + MySQL）
 SQLAlchemy>=2.0
 aiomysql>=0.2
+greenlet>=3.0
+aiosqlite>=0.20
 
 # 配置
 pydantic>=2.6
@@ -18,3 +20,5 @@ redis>=5.0
 
 # 测试
 pytest>=8.0
+pytest-asyncio>=0.23
+httpx>=0.27

server/run.sh

@@ -0,0 +1,145 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# 一键启动 FastAPI 后端：
+# - 自动创建/复用虚拟环境（.venv）
+# - 自动安装 requirements.txt 依赖
+# - 自动启动 uvicorn（默认开启 --reload）
+#
+# 用法示例：
+#   ./run.sh                 # 默认 host=0.0.0.0 port=8000 env=dev reload=on
+#   ./run.sh --env prod      # 使用 .env.prod（若存在且可被 source）
+#   ./run.sh --port 9000     # 改端口
+#   ./run.sh --no-reload     # 关闭热更新
+#   ./run.sh --install-only  # 只安装依赖，不启动
+
+usage() {
+  cat <<'EOF'
+用法：
+  ./run.sh [--env dev|prod] [--host 0.0.0.0] [--port 8000] [--no-reload] [--skip-install] [--install-only]
+
+参数：
+  --env dev|prod     优先尝试加载 .env.dev 或 .env.prod（如果存在）。
+  --host <host>      uvicorn host（默认 0.0.0.0）
+  --port <port>      uvicorn port（默认 8000）
+  --no-reload        关闭 uvicorn --reload
+  --skip-install     跳过依赖安装（默认会安装/更新 requirements.txt）
+  --install-only     只安装依赖，不启动服务
+  -h, --help         显示帮助
+
+说明：
+  - 若你的 .env.* 不是 shell 可 source 的格式（例如包含空格/特殊字符未加引号），建议改成 KEY=value 形式。
+  - 启动后访问：
+      /healthz  健康检查
+      /docs     OpenAPI 文档
+EOF
+}
+
+# 始终从脚本所在目录运行（避免在别处执行导致路径错）
+SCRIPT_DIR="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)"
+cd "$SCRIPT_DIR"
+
+ENV_NAME="dev"
+HOST="0.0.0.0"
+PORT="8000"
+RELOAD="1"
+SKIP_INSTALL="0"
+INSTALL_ONLY="0"
+
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --env)
+      ENV_NAME="${2:-}"
+      shift 2
+      ;;
+    --host)
+      HOST="${2:-}"
+      shift 2
+      ;;
+    --port)
+      PORT="${2:-}"
+      shift 2
+      ;;
+    --no-reload)
+      RELOAD="0"
+      shift 1
+      ;;
+    --skip-install)
+      SKIP_INSTALL="1"
+      shift 1
+      ;;
+    --install-only)
+      INSTALL_ONLY="1"
+      shift 1
+      ;;
+    -h|--help)
+      usage
+      exit 0
+      ;;
+    *)
+      echo "未知参数：$1" >&2
+      echo "" >&2
+      usage >&2
+      exit 2
+      ;;
+  esac
+done
+
+if [[ "$ENV_NAME" != "dev" && "$ENV_NAME" != "prod" ]]; then
+  echo "--env 仅支持 dev 或 prod，当前：$ENV_NAME" >&2
+  exit 2
+fi
+
+ENV_FILE=".env.${ENV_NAME}"
+if [[ -f "$ENV_FILE" ]]; then
+  # 让 source 进来的变量自动 export（供 pydantic-settings/应用读取）
+  set -a
+  # shellcheck disable=SC1090
+  source "$ENV_FILE"
+  set +a
+fi
+
+# 选择 python 命令（优先 python3）
+PY_BIN=""
+if command -v python3 >/dev/null 2>&1; then
+  PY_BIN="python3"
+elif command -v python >/dev/null 2>&1; then
+  PY_BIN="python"
+else
+  echo "未找到 python/python3，请先安装 Python 3.11+。" >&2
+  exit 1
+fi
+
+VENV_DIR=".venv"
+if [[ ! -d "$VENV_DIR" ]]; then
+  echo "创建虚拟环境：$VENV_DIR"
+  "$PY_BIN" -m venv "$VENV_DIR"
+fi
+
+# 激活虚拟环境
+# shellcheck disable=SC1091
+source "$VENV_DIR/bin/activate"
+
+if [[ "$SKIP_INSTALL" == "0" ]]; then
+  if [[ -f "requirements.txt" ]]; then
+    echo "升级 pip 并安装依赖（requirements.txt）"
+    python -m pip install -U pip
+    python -m pip install -r requirements.txt
+  else
+    echo "未找到 requirements.txt，跳过依赖安装。" >&2
+  fi
+fi
+
+if [[ "$INSTALL_ONLY" == "1" ]]; then
+  echo "依赖安装完成（install-only），退出。"
+  exit 0
+fi
+
+UVICORN_ARGS=(app.main:app --host "$HOST" --port "$PORT")
+if [[ "$RELOAD" == "1" ]]; then
+  UVICORN_ARGS+=(--reload)
+fi
+
+echo "启动服务：uvicorn ${UVICORN_ARGS[*]}"
+exec uvicorn "${UVICORN_ARGS[@]}"
+

server/tests/conftest.py

@@ -0,0 +1,223 @@
+from __future__ import annotations
+
+import os
+import sys
+from pathlib import Path
+from typing import AsyncIterator, Callable
+from urllib.parse import urlparse
+
+import pytest
+import pytest_asyncio
+import sqlalchemy as sa
+from alembic import command
+from alembic.config import Config
+from sqlalchemy import event
+from sqlalchemy.ext.asyncio import AsyncEngine, AsyncSession, async_sessionmaker, create_async_engine
+
+# 确保在任何 pytest rootdir 下都能 `import app.*`
+SERVER_DIR = Path(__file__).resolve().parents[1]  # .../server
+if str(SERVER_DIR) not in sys.path:
+    sys.path.insert(0, str(SERVER_DIR))
+
+
+def _read_env_kv(env_path: Path) -> dict[str, str]:
+    """
+    读取 .env 文件中的 KEY=VALUE（最小实现，避免引入额外依赖）。
+    """
+
+    data: dict[str, str] = {}
+    if not env_path.exists():
+        return data
+    for raw in env_path.read_text(encoding="utf-8").splitlines():
+        line = raw.strip()
+        if not line or line.startswith("#"):
+            continue
+        if "=" not in line:
+            continue
+        k, v = line.split("=", 1)
+        k = k.strip()
+        v = v.strip().strip('"').strip("'")
+        if k:
+            data[k] = v
+    return data
+
+
+def _get_database_url() -> str:
+    """
+    获取测试用数据库连接串。
+
+    约定（与 alembic/env.py 保持一致）：
+    - 优先读取环境变量 `DATABASE_URL`
+    - 若未设置，则按 `APP_ENV`（默认 dev）读取 `server/.env.dev` 或 `server/.env.prod`
+    """
+
+    env_url = (os.getenv("DATABASE_URL") or "").strip()
+    if env_url:
+        return env_url
+
+    server_dir = Path(__file__).resolve().parents[1]  # .../server
+    app_env = (os.getenv("APP_ENV") or "dev").strip() or "dev"
+    env_file = server_dir / (".env.prod" if app_env == "prod" else ".env.dev")
+    kv = _read_env_kv(env_file)
+    url = (kv.get("DATABASE_URL") or "").strip()
+    if url:
+        return url
+
+    raise RuntimeError(
+        "缺少 DATABASE_URL：请设置环境变量 DATABASE_URL，或在 server/.env.dev（或 .env.prod）中配置 DATABASE_URL。"
+    )
+
+
+def _assert_safe_mysql_test_db(url: str) -> None:
+    """
+    为了避免对开发库造成破坏性影响，集成测试只允许连接到“测试库”。
+
+    规则（V1）：
+    - 必须是 mysql+aiomysql://...
+    - 为避免误连生产库：不允许数据库名为 'mindfulness'（prod 默认库名）
+    - 建议使用独立测试库（例如 mindfulness_dev_test）
+    """
+
+    if not url.startswith("mysql+"):
+        raise RuntimeError(f"当前仅允许 MySQL 集成测试（mysql+aiomysql）。实际：{url!r}")
+
+    parsed = urlparse(url.replace("mysql+aiomysql://", "mysql://", 1))
+    db_name = (parsed.path or "").lstrip("/")
+    if db_name.lower() == "mindfulness":
+        raise RuntimeError(
+            "为避免误连生产库，集成测试不允许连接到数据库 'mindfulness'。"
+            "请改用 dev 测试库（例如 mindfulness_dev_test 或 mindfulness_dev）。"
+        )
+
+
+def _run_alembic_upgrade_head() -> None:
+    """
+    使用 Alembic 将测试库升级到最新 schema。
+
+    说明：
+    - 依赖 env.py 内部读取 DATABASE_URL
+    - 仅在 session 级别执行一次，避免每个测试都跑迁移
+    """
+
+    server_dir = Path(__file__).resolve().parents[1]  # .../server
+    alembic_ini = server_dir / "alembic.ini"
+    cfg = Config(str(alembic_ini))
+    # 确保脚本路径正确（alembic.ini 里一般已配置，这里兜底）
+    cfg.set_main_option("script_location", "alembic")
+    command.upgrade(cfg, "head")
+
+def _assert_schema_exists(url: str) -> None:
+    """
+    非破坏性检查：要求目标库已经存在所需表。
+
+    说明：
+    - 默认不在测试中运行 Alembic（避免任何 schema 变更）
+    - 若要自动迁移，请设置环境变量 ALLOW_SCHEMA_MIGRATION=1
+    """
+
+    allow_migration = (os.getenv("ALLOW_SCHEMA_MIGRATION") or "").strip() == "1"
+    if allow_migration:
+        _run_alembic_upgrade_head()
+        return
+
+    # 使用 PyMySQL 做同步检查，避免依赖 MySQLdb（不要求系统安装 mysqlclient）
+    sync_url = url.replace("mysql+aiomysql://", "mysql+pymysql://", 1)
+    engine = sa.create_engine(sync_url, future=True)
+    try:
+        insp = sa.inspect(engine)
+        tables = set(insp.get_table_names())
+        required = {"contents", "content_profiles", "content_risk_flags"}
+        missing = sorted(required - tables)
+        if missing:
+            raise RuntimeError(
+                "集成测试检测到 schema 不完整（缺少表："
+                + ", ".join(missing)
+                + "）。为避免破坏性操作，测试不会自动迁移。"
+                "请先手动在该库执行 `alembic upgrade head`，或设置 ALLOW_SCHEMA_MIGRATION=1 允许测试自动迁移。"
+            )
+    finally:
+        engine.dispose()
+
+
+@pytest.fixture(scope="session", autouse=True)
+def _migrate_db_once() -> None:
+    """
+    Session 级 schema 检查（仅在配置了数据库连接时启用）。
+
+    说明：
+    - 纯函数单元测试不需要 MySQL；若未配置 DATABASE_URL，则跳过检查
+    - 集成测试（依赖 db_session/async_engine）仍会在获取 DATABASE_URL 时失败，从而提示用户配置
+    """
+
+    try:
+        url = _get_database_url()
+    except RuntimeError:
+        return
+
+    _assert_safe_mysql_test_db(url)
+    _assert_schema_exists(url)
+
+
+@pytest_asyncio.fixture
+async def async_engine() -> AsyncIterator[AsyncEngine]:
+    url = _get_database_url()
+    _assert_safe_mysql_test_db(url)
+    engine = create_async_engine(url, pool_pre_ping=True)
+    try:
+        yield engine
+    finally:
+        await engine.dispose()
+
+
+@pytest_asyncio.fixture
+async def db_session(async_engine: AsyncEngine) -> AsyncIterator[AsyncSession]:
+    """
+    提供一个干净的 AsyncSession。
+
+    清理策略：每个测试都在事务中执行，并在结束时回滚（不做 DELETE/TRUNCATE）。
+
+    说明：
+    - 用例里严禁调用 session.commit()，只允许 flush()
+    - 这样不会对测试库产生持久化写入，更不会影响开发库
+    """
+
+    SessionLocal: async_sessionmaker[AsyncSession] = async_sessionmaker(
+        bind=async_engine,
+        expire_on_commit=False,
+        autoflush=False,
+        autocommit=False,
+    )
+
+    async with SessionLocal() as session:
+        trans = await session.begin()
+        try:
+            yield session
+        finally:
+            await trans.rollback()
+
+
+@pytest.fixture
+def query_counter(async_engine: AsyncEngine) -> Callable[[], int]:
+    """
+    返回一个函数：调用可获得当前累计查询次数。
+    """
+
+    count = {"n": 0}
+
+    def before_cursor_execute(*args, **kwargs):  # type: ignore[no-untyped-def]
+        count["n"] += 1
+
+    event.listen(async_engine.sync_engine, "before_cursor_execute", before_cursor_execute)
+
+    def get_count() -> int:
+        return int(count["n"])
+
+    def fin() -> None:
+        event.remove(async_engine.sync_engine, "before_cursor_execute", before_cursor_execute)
+
+    # 用 yield 确保测试后移除监听，避免重复绑定导致统计偏大
+    try:
+        yield get_count  # type: ignore[misc]
+    finally:
+        fin()
+

server/tests/test_content_repository.py

@@ -0,0 +1,178 @@
+from __future__ import annotations
+
+from datetime import datetime
+
+import pytest
+
+from app.db.models.content import Content
+from app.db.models.content_profile import ContentProfile
+from app.db.models.content_risk_flag import ContentRiskFlag
+from app.features.personalized_reco.content_repository.sqlalchemy_repo import SqlAlchemyContentRepository
+
+
+def _ctx_json() -> dict:
+    return {"family": 0.5, "work": 0.5, "relationship": 0.5, "friends": 0.5, "health": 0.5}
+
+
+def _need_json() -> dict:
+    return {
+        "emotional_support": 0.5,
+        "parenting_pressure": 0.5,
+        "self_worth": 0.5,
+        "anxiety_relief": 0.5,
+        "rest_balance": 0.5,
+    }
+
+
+@pytest.mark.asyncio
+async def test_fetch_contents_by_ids_locale_no_fallback(db_session, query_counter):
+    # content 1: 只有英文
+    c1 = Content(text_en="hello", text_tc=None, author_id="a1", template_id="t1")
+    db_session.add(c1)
+    await db_session.flush()
+    db_session.add(
+        ContentProfile(
+            content_id=c1.content_id,
+            stage="general",
+            emotion_score=None,
+            context_suitability_json=_ctx_json(),
+            need_suitability_json=_need_json(),
+            personalization_power=10,
+            review_confidence=None,
+            is_safe_pool=False,
+        )
+    )
+    db_session.add(ContentRiskFlag(content_id=c1.content_id, flag="block_stage_unknown"))
+
+    # content 2: 只有繁中
+    c2 = Content(text_en=None, text_tc="繁體中文", author_id="a2", template_id="t2")
+    db_session.add(c2)
+    await db_session.flush()
+    db_session.add(
+        ContentProfile(
+            content_id=c2.content_id,
+            stage="general",
+            emotion_score=None,
+            context_suitability_json=_ctx_json(),
+            need_suitability_json=_need_json(),
+            personalization_power=0,
+            review_confidence=0.9,
+            is_safe_pool=True,
+        )
+    )
+    db_session.add(ContentRiskFlag(content_id=c2.content_id, flag="block_health_sensitive"))
+
+    await db_session.flush()
+
+    repo = SqlAlchemyContentRepository(db_session)
+    start = query_counter()
+
+    # en：只能拿到有 text_en 的内容（不允许回退到 text_tc）
+    en_items = await repo.fetch_contents_by_ids(content_ids=[c2.content_id, c1.content_id], locale="en")
+    assert [x.content_id for x in en_items] == [c1.content_id]
+    assert en_items[0].text == "hello"
+    assert en_items[0].personalization_power == 1.0
+    assert en_items[0].review_confidence == 0.7  # NULL -> 0.7
+    assert "unsafe_for_stage_unknown" in en_items[0].risk_flags
+    assert "block_stage_unknown" not in en_items[0].risk_flags
+
+    # tc：只能拿到有 text_tc 的内容（不允许回退到 text_en）
+    tc_items = await repo.fetch_contents_by_ids(content_ids=[c1.content_id, c2.content_id], locale="tc")
+    assert [x.content_id for x in tc_items] == [c2.content_id]
+    assert tc_items[0].text == "繁體中文"
+    assert tc_items[0].review_confidence == 0.9
+    assert "block_health_medical" in tc_items[0].risk_flags
+    assert "block_health_sensitive" not in tc_items[0].risk_flags
+
+    # 两次调用各自 2 次查询（主体+画像一次，flags 一次），总计应为常数级
+    end = query_counter()
+    assert (end - start) <= 4
+
+
+@pytest.mark.asyncio
+async def test_fetch_candidates_fallback_and_locale_filter(db_session, query_counter):
+    # 构造 4 条英文内容：power 0/5/10，安全池标记不同
+    contents = []
+    for i, (power, safe) in enumerate([(0, True), (5, False), (10, False), (0, False)], start=1):
+        c = Content(text_en=f"en_{i}", text_tc=None, author_id=f"a{i}", template_id=f"t{i}")
+        db_session.add(c)
+        await db_session.flush()
+        db_session.add(
+            ContentProfile(
+                content_id=c.content_id,
+                stage="general",
+                emotion_score=None,
+                context_suitability_json=_ctx_json(),
+                need_suitability_json=_need_json(),
+                personalization_power=power,
+                review_confidence=None,
+                is_safe_pool=safe,
+                # 让测试数据在候选排序中排到最前，避免依赖“库为空”
+                updated_at=datetime(2099, 1, 1, 0, 0, 0),
+            )
+        )
+        contents.append(c)
+    await db_session.flush()
+    inserted_ids = {int(c.content_id) for c in contents}
+
+    repo = SqlAlchemyContentRepository(db_session)
+
+    class _MinimalUser:
+        # need/context/emotion_score 全缺失 -> effective_fallback 至少 L1
+        need = {}
+        context = {}
+        emotion_score = None
+        stage = {"unknown": 1}
+
+    start = query_counter()
+
+    # 入参 L0，但因为缺失字段，effective_fallback=1 -> power<=5（排除 power=10）
+    items_l0 = await repo.fetch_candidates(
+        scene="feed",
+        user_profile=_MinimalUser(),
+        fallback_level=0,
+        limit=3,
+        locale="en-US",
+    )
+    powers = [x.personalization_power for x in items_l0]
+    assert 1.0 not in powers
+    assert all(x.text.startswith("en_") for x in items_l0)
+
+    # L2：强制 power=0 且 stage=general（这里都 general），只剩 power=0 的两条
+    items_l2 = await repo.fetch_candidates(
+        scene="feed",
+        user_profile=_MinimalUser(),
+        fallback_level=2,
+        limit=2,
+        locale="en",
+    )
+    assert all(x.personalization_power == 0.0 for x in items_l2)
+    assert all(x.text.startswith("en_") for x in items_l2)
+
+    # L3：只安全池（is_safe_pool=true）且 power=0
+    items_l3 = await repo.fetch_candidates(
+        scene="feed",
+        user_profile=_MinimalUser(),
+        fallback_level=3,
+        limit=1,
+        locale="en",
+    )
+    assert len(items_l3) == 1
+    assert items_l3[0].text.startswith("en_")
+    assert items_l3[0].personalization_power == 0.0
+
+    # locale 过滤：tc 请求下这些内容都没有 text_tc -> 返回空
+    items_tc = await repo.fetch_candidates(
+        scene="feed",
+        user_profile=_MinimalUser(),
+        fallback_level=0,
+        limit=10,
+        locale="tc",
+    )
+    # 不要求库为空：只断言“不会把本次插入的 en-only 测试数据返回出来”
+    assert not any(x.content_id in inserted_ids for x in items_tc)
+
+    end = query_counter()
+    # 期望为常数级（每次 fetch_candidates：1 次取 ids + 2 次补全），这里 4 次调用 -> <= 12
+    assert (end - start) <= 12
+

server/tests/test_integration_api_worker.py

@@ -0,0 +1,165 @@
+from __future__ import annotations
+
+import importlib
+from datetime import datetime, timezone
+from typing import Any
+
+import pytest
+from fastapi.testclient import TestClient
+
+
+def _set_min_env(monkeypatch: pytest.MonkeyPatch) -> None:
+    # 让 Settings 可构造（引擎不会在单测中真正连接 DB/Redis）
+    monkeypatch.setenv(
+        "DATABASE_URL",
+        "mysql+aiomysql://u:p@127.0.0.1:3306/mindfulness_dev_test?charset=utf8mb4",
+    )
+    monkeypatch.setenv("REDIS_URL", "redis://127.0.0.1:6379/0")
+    monkeypatch.setenv("CELERY_BROKER_URL", "redis://127.0.0.1:6379/0")
+
+
+def _user_profile_dict() -> dict[str, Any]:
+    return {
+        "profile_version": "v1.2",
+        "profile_source": "questionnaire",
+        "profile_generated_at": "2026-02-02T12:00:00Z",
+        "profile_confidence": 1.0,
+        "profile_answered": {"stage": True, "emotion": False, "context": False, "need": False},
+        "stage": {"unknown": 1},
+        "emotion_score": None,
+        "context": {},
+        "need": {},
+    }
+
+
+@pytest.fixture
+def client(monkeypatch: pytest.MonkeyPatch):
+    _set_min_env(monkeypatch)
+
+    # 清理 settings cache，避免被其他测试污染
+    from app.core import config as config_mod
+
+    config_mod.get_settings.cache_clear()
+
+    # 重新加载 main，确保使用最新 env
+    import app.main as main_mod
+
+    importlib.reload(main_mod)
+
+    app = main_mod.create_app()
+
+    # override repo（避免依赖真实 DB）
+    from app.api.v1 import reco as reco_mod
+    from app.features.personalized_reco.content_repository.types import ContentProfileDTO
+
+    class _FakeRepo:
+        async def fetch_candidates(self, **kwargs):  # type: ignore[no-untyped-def]
+            # 返回一条可下发内容
+            return [
+                ContentProfileDTO(
+                    content_id=1,
+                    text="t1",
+                    stage="general",
+                    emotion_score=None,
+                    context_suitability={},
+                    need_suitability={},
+                    personalization_power=0.0,
+                    risk_flags=[],
+                    author_id=None,
+                    template_id=None,
+                    review_confidence=0.7,
+                )
+            ]
+
+        async def fetch_contents_by_ids(self, **kwargs):  # type: ignore[no-untyped-def]
+            return []
+
+    async def _override_repo():  # type: ignore[no-untyped-def]
+        return _FakeRepo()
+
+    app.dependency_overrides[reco_mod.get_reco_repo] = _override_repo
+
+    # 清空限流计数，避免跨测试污染
+    import app.api.limits as limits_mod
+
+    limits_mod._reco_rate_limiter._counters.clear()  # type: ignore[attr-defined]
+    limits_mod._reco_rate_limiter._last_gc_bucket = 0  # type: ignore[attr-defined]
+
+    return TestClient(app)
+
+
+def test_accept_language_mapping_to_tc(client: TestClient, monkeypatch: pytest.MonkeyPatch) -> None:
+    # 固定时间，避免跨分钟 flake
+    import app.api.limits as limits_mod
+
+    monkeypatch.setattr(limits_mod.time, "time", lambda: 1738497600.0)  # 2025-02-02 12:00:00Z 的某个时间戳
+
+    resp = client.post(
+        "/v1/reco/feed",
+        json={"user_profile": _user_profile_dict(), "already_recommended_ids": [], "touched_or_viewed_ids": []},
+        headers={"Accept-Language": "zh-TW,zh;q=0.9"},
+    )
+    assert resp.status_code == 200
+    data = resp.json()
+    assert data["meta"]["config_snapshot"]["locale"] == "tc"
+
+
+def test_x_now_header_priority_over_body_now(client: TestClient, monkeypatch: pytest.MonkeyPatch) -> None:
+    import app.api.limits as limits_mod
+
+    monkeypatch.setattr(limits_mod.time, "time", lambda: 1738497600.0)
+
+    resp = client.post(
+        "/v1/reco/push",
+        json={
+            "user_profile": _user_profile_dict(),
+            "now": "2026-02-01T00:00:00Z",
+            "already_recommended_ids": [],
+            "touched_or_viewed_ids": [],
+        },
+        headers={"X-Now": "2026-02-02T12:00:00Z", "Accept-Language": "en-US"},
+    )
+    assert resp.status_code == 200
+    data = resp.json()
+    # meta.config_snapshot 里没有 now，但 served_k 应该正常
+    assert data["meta"]["served_k"] == 1
+
+
+def test_rate_limit_10_per_minute(client: TestClient, monkeypatch: pytest.MonkeyPatch) -> None:
+    import app.api.limits as limits_mod
+
+    monkeypatch.setattr(limits_mod.time, "time", lambda: 1738497600.0)
+
+    body = {"user_profile": _user_profile_dict(), "already_recommended_ids": [], "touched_or_viewed_ids": []}
+    for _ in range(10):
+        r = client.post("/v1/reco/widget", json=body)
+        assert r.status_code == 200
+
+    r = client.post("/v1/reco/widget", json=body)
+    assert r.status_code == 429
+    assert r.json()["detail"] == "rate_limited"
+
+
+def test_celery_tasks_can_call_generate(monkeypatch: pytest.MonkeyPatch) -> None:
+    _set_min_env(monkeypatch)
+    from app.core import config as config_mod
+
+    config_mod.get_settings.cache_clear()
+
+    import app.tasks.reco as reco_tasks
+
+    # monkeypatch async runner，避免依赖 DB
+    async def _fake_run_reco_async(**kwargs):  # type: ignore[no-untyped-def]
+        from app.features.personalized_reco.observability.types import RecoMeta
+        from app.features.personalized_reco.reco_engine.types import RecoEngineResult, RecommendedItem
+
+        return RecoEngineResult(
+            items=[RecommendedItem(content_id=1, text="t1", final_score=1.0, fallback_level_final=0, explanations={})],
+            meta=RecoMeta(scene="push", served_k=1),
+        )
+
+    monkeypatch.setattr(reco_tasks, "_run_reco_async", _fake_run_reco_async)
+
+    out = reco_tasks.generate(scene="push", user_profile=_user_profile_dict(), k=1)
+    assert out["meta"]["served_k"] == 1
+

server/tests/test_observability.py

@@ -0,0 +1,129 @@
+from __future__ import annotations
+
+from datetime import datetime, timezone
+
+import pytest
+
+from app.features.personalized_reco.observability.builder import RecoMetaBuilder
+from app.features.personalized_reco.observability.utils import compute_empty_reason, compute_missing_fields
+from app.features.user_profile_scoring.types import ProfileAnswered, UserProfileV1_2, UserStageOneHot
+
+
+def _u(*, need: dict | None = None, context: dict | None = None, emotion_score=None, conf_u: float = 0.9) -> UserProfileV1_2:
+    now = datetime.now(tz=timezone.utc)
+    return UserProfileV1_2(
+        profile_generated_at=now,
+        profile_confidence=conf_u,
+        profile_answered=ProfileAnswered(stage=True, emotion=True, context=True, need=True),
+        stage=UserStageOneHot(unknown=1),
+        emotion_score=emotion_score,
+        context=context or {},
+        need=need or {},
+    )
+
+
+def test_compute_missing_fields() -> None:
+    u1 = _u(need={}, context={}, emotion_score=None)
+    m1 = compute_missing_fields(u1)
+    assert m1.need is True
+    assert m1.context is True
+    assert m1.emotion is True
+
+    u2 = _u(need={"x": 1}, context={"y": 1}, emotion_score=0.6)
+    m2 = compute_missing_fields(u2)
+    assert m2.need is False
+    assert m2.context is False
+    assert m2.emotion is False
+
+
+def test_compute_empty_reason_branches() -> None:
+    assert (
+        compute_empty_reason(
+            served_k=1,
+            candidate_pool_size_raw=0,
+            candidate_pool_size_after_hard_filter=0,
+            candidate_pool_size_after_freqcap=0,
+        )
+        is None
+    )
+
+    assert (
+        compute_empty_reason(
+            served_k=0,
+            candidate_pool_size_raw=0,
+            candidate_pool_size_after_hard_filter=0,
+            candidate_pool_size_after_freqcap=0,
+        )
+        == "pool_empty"
+    )
+
+    assert (
+        compute_empty_reason(
+            served_k=0,
+            candidate_pool_size_raw=10,
+            candidate_pool_size_after_hard_filter=0,
+            candidate_pool_size_after_freqcap=0,
+        )
+        == "hard_filter_all"
+    )
+
+    assert (
+        compute_empty_reason(
+            served_k=0,
+            candidate_pool_size_raw=10,
+            candidate_pool_size_after_hard_filter=5,
+            candidate_pool_size_after_freqcap=0,
+        )
+        == "freqcap_all"
+    )
+
+    assert (
+        compute_empty_reason(
+            served_k=0,
+            candidate_pool_size_raw=10,
+            candidate_pool_size_after_hard_filter=5,
+            candidate_pool_size_after_freqcap=3,
+        )
+        == "unknown"
+    )
+
+
+def test_builder_outputs_stable_fields_and_monotonic_counts() -> None:
+    u = _u(need={"emotional_support": 1}, context={}, emotion_score=None, conf_u=0.2)
+
+    # 故意设置“非单调”的输入，验证 builder 的防御修正
+    meta = (
+        RecoMetaBuilder(scene="feed", user_profile=u, k=30)
+        .set_candidate_pool_size_raw(10)
+        .set_after_hard_filter(12)  # 非法：大于 raw
+        .set_after_dedup(20)  # 非法：大于 after_hard
+        .set_after_freqcap(15)  # 非法：大于 after_dedup（修正后会与 after_dedup 对齐）
+        .set_served_k(99)  # 非法：大于 after_freqcap
+        .set_fallback_level_final(1, reason="freqcap_all")
+        .build()
+    )
+
+    d = meta.model_dump()
+    for k in [
+        "scene",
+        "candidate_pool_size_raw",
+        "candidate_pool_size_after_hard_filter",
+        "candidate_pool_size_after_dedup",
+        "candidate_pool_size_after_freqcap",
+        "fallback_level_final",
+        "served_k",
+        "empty_reason",
+        "conf_U",
+        "missing_fields",
+    ]:
+        assert k in d
+
+    assert meta.candidate_pool_size_raw == 10
+    assert meta.candidate_pool_size_after_hard_filter == 10
+    assert meta.candidate_pool_size_after_dedup == 10
+    assert meta.candidate_pool_size_after_freqcap == 10
+    assert meta.served_k == 10
+    assert meta.conf_U == pytest.approx(0.2)
+    assert meta.missing_fields.context is True
+    assert meta.missing_fields.emotion is True
+

server/tests/test_reco_engine.py

@@ -0,0 +1,173 @@
+from __future__ import annotations
+
+from datetime import datetime, timezone
+
+import pytest
+
+from app.features.personalized_reco.content_repository.interface import ContentRepository
+from app.features.personalized_reco.content_repository.types import ContentProfileDTO
+from app.features.personalized_reco.reco_engine.orchestrator import recommend
+from app.features.personalized_reco.reco_engine.types import RecoConstraints
+from app.features.user_profile_scoring.types import ProfileAnswered, UserProfileV1_2, UserStageOneHot
+
+
+class _FakeRepo(ContentRepository):
+    def __init__(self, candidates_by_level: dict[int, list[ContentProfileDTO]]):
+        self._candidates_by_level = candidates_by_level
+
+    async def fetch_candidates(  # type: ignore[override]
+        self,
+        *,
+        scene: str,
+        user_profile: object,
+        fallback_level: int,
+        limit: int,
+        locale: str,
+        exclude_content_ids: list[int] | None = None,
+    ) -> list[ContentProfileDTO]:
+        # 单测：简化实现，只按 level 返回，忽略 limit/locale/exclude
+        return list(self._candidates_by_level.get(int(fallback_level), []))[: int(limit)]
+
+    async def fetch_contents_by_ids(self, *, content_ids: list[int], locale: str) -> list[ContentProfileDTO]:  # type: ignore[override]
+        raise NotImplementedError
+
+
+def _now() -> datetime:
+    return datetime(2026, 2, 2, 12, 0, 0, tzinfo=timezone.utc)
+
+
+def _user_profile(*, stage: str = "unknown", emotion_score: float | None = 0.5) -> UserProfileV1_2:
+    if stage == "expecting":
+        st = UserStageOneHot(expecting=1, parenting=0, unknown=0)  # type: ignore[arg-type]
+    elif stage == "parenting":
+        st = UserStageOneHot(expecting=0, parenting=1, unknown=0)  # type: ignore[arg-type]
+    else:
+        st = UserStageOneHot(expecting=0, parenting=0, unknown=1)  # type: ignore[arg-type]
+
+    return UserProfileV1_2(
+        profile_generated_at=_now(),
+        profile_confidence=1.0,
+        profile_answered=ProfileAnswered(stage=True, emotion=emotion_score is not None, context=False, need=False),
+        stage=st,
+        emotion_score=emotion_score,
+        context={},
+        need={},
+    )
+
+
+def _content(
+    *,
+    content_id: int,
+    text: str = "hello",
+    stage: str = "general",
+    personalization_power: float = 0.0,
+    risk_flags: list[str] | None = None,
+    author_id: str | None = None,
+    template_id: str | None = None,
+    need_suitability: dict[str, float] | None = None,
+) -> ContentProfileDTO:
+    return ContentProfileDTO(
+        content_id=int(content_id),
+        text=text,
+        stage=stage,  # type: ignore[arg-type]
+        emotion_score=None,
+        context_suitability={},
+        need_suitability=need_suitability or {},
+        personalization_power=float(personalization_power),
+        risk_flags=risk_flags or [],
+        author_id=author_id,
+        template_id=template_id,
+        review_confidence=0.7,
+    )
+
+
+@pytest.mark.asyncio
+async def test_reco_engine_k_zero_returns_empty() -> None:
+    repo = _FakeRepo({0: [_content(content_id=1)]})
+    res = await recommend(
+        repo=repo,
+        scene="feed",
+        user_profile=_user_profile(),
+        already_recommended_ids=[],
+        touched_or_viewed_ids=[],
+        k=0,
+        now=_now(),
+        locale=None,
+    )
+    assert res.items == []
+    assert res.meta.served_k == 0
+
+
+@pytest.mark.asyncio
+async def test_reco_engine_pool_empty_sets_empty_reason() -> None:
+    repo = _FakeRepo({0: []})
+    res = await recommend(
+        repo=repo,
+        scene="feed",
+        user_profile=_user_profile(),
+        already_recommended_ids=[],
+        touched_or_viewed_ids=[],
+        k=1,
+        now=_now(),
+        locale="en",
+    )
+    assert res.items == []
+    assert res.meta.empty_reason == "pool_empty"
+
+
+@pytest.mark.asyncio
+async def test_reco_engine_hard_filter_all_sets_empty_reason() -> None:
+    repo = _FakeRepo({0: [_content(content_id=1, risk_flags=["block_health_medical"])]})
+    res = await recommend(
+        repo=repo,
+        scene="push",
+        user_profile=_user_profile(stage="unknown"),
+        already_recommended_ids=[],
+        touched_or_viewed_ids=[],
+        k=1,
+        now=_now(),
+        locale="en",
+    )
+    assert res.items == []
+    assert res.meta.empty_reason == "hard_filter_all"
+    assert res.meta.risk_filtered_count_by_flag.get("block_health_medical", 0) == 1
+
+
+@pytest.mark.asyncio
+async def test_reco_engine_freqcap_all_sets_empty_reason() -> None:
+    # Push 场景：作者冷却命中，导致 after_freqcap=0
+    repo = _FakeRepo({0: [_content(content_id=1, author_id="a1")]})
+    res = await recommend(
+        repo=repo,
+        scene="push",
+        user_profile=_user_profile(stage="unknown"),
+        already_recommended_ids=[],
+        touched_or_viewed_ids=[],
+        k=1,
+        now=_now(),
+        locale="en",
+        constraints=RecoConstraints(recent_author_ids=["a1"]),
+    )
+    assert res.items == []
+    assert res.meta.empty_reason == "freqcap_all"
+
+
+@pytest.mark.asyncio
+async def test_reco_engine_returns_item_and_explanations_enabled() -> None:
+    repo = _FakeRepo({0: [_content(content_id=1, text="t1", personalization_power=0.0)]})
+    res = await recommend(
+        repo=repo,
+        scene="feed",
+        user_profile=_user_profile(stage="unknown"),
+        already_recommended_ids=[],
+        touched_or_viewed_ids=[],
+        k=1,
+        now=_now(),
+        locale="en",
+    )
+    assert len(res.items) == 1
+    assert res.items[0].content_id == 1
+    assert res.items[0].text == "t1"
+    assert res.items[0].explanations is not None
+
+

server/tests/test_rerank_freqcap.py

@@ -0,0 +1,130 @@
+from __future__ import annotations
+
+import pytest
+
+from app.features.personalized_reco.content_repository.types import ContentProfileDTO
+from app.features.personalized_reco.rerank_freqcap.rerank import rerank_and_freqcap
+from app.features.personalized_reco.rerank_freqcap.types import ScoredCandidate
+
+
+def _c(
+    *,
+    content_id: int,
+    score: float,
+    author_id: str | None = None,
+    template_id: str | None = None,
+    stage: str = "general",
+    need_key: str = "emotional_support",
+    context_key: str = "family",
+) -> ScoredCandidate:
+    cp = ContentProfileDTO(
+        content_id=content_id,
+        text="t",
+        stage=stage,  # type: ignore[arg-type]
+        emotion_score=None,
+        context_suitability={context_key: 1.0},
+        need_suitability={need_key: 1.0},
+        personalization_power=0.0,
+        risk_flags=[],
+        author_id=author_id,
+        template_id=template_id,
+    )
+    return ScoredCandidate(
+        content_id=content_id,
+        final_score=score,
+        author_id=author_id,
+        template_id=template_id,
+        content_profile=cp,
+    )
+
+
+def test_dedup_normalizes_str_int_ids() -> None:
+    cands = [_c(content_id=1, score=0.9), _c(content_id=2, score=0.8), _c(content_id=3, score=0.7)]
+    r = rerank_and_freqcap(
+        scene="push",
+        scored_candidates=cands,
+        already_recommended_ids=["1", "bad"],
+        touched_or_viewed_ids=[2],
+        k=10,
+    )
+    got_ids = [x.content_id for x in r.ranked_items]
+    assert got_ids == [3]
+    assert r.meta.candidate_pool_size_after_dedup == 1
+    assert r.meta.freqcap_filtered_counts["sentence"] == 2
+
+
+def test_freqcap_missing_recent_author_template_is_recorded() -> None:
+    cands = [
+        _c(content_id=1, score=0.9, author_id="a1", template_id="t1"),
+        _c(content_id=2, score=0.8, author_id="a2", template_id="t2"),
+    ]
+    r = rerank_and_freqcap(
+        scene="push",
+        scored_candidates=cands,
+        already_recommended_ids=[],
+        touched_or_viewed_ids=[],
+        k=10,
+        recent_author_ids=None,
+        recent_template_ids=None,
+    )
+    assert r.meta.missing_history_fields == ["author", "template"]
+    assert "author" not in r.meta.freqcap_filtered_counts  # 未提供则不执行该维度
+    assert "template" not in r.meta.freqcap_filtered_counts
+
+
+def test_freqcap_filters_by_recent_author_when_provided() -> None:
+    cands = [
+        _c(content_id=1, score=0.9, author_id="a1", template_id="t1"),
+        _c(content_id=2, score=0.8, author_id="a2", template_id="t2"),
+    ]
+    r = rerank_and_freqcap(
+        scene="widget",
+        scored_candidates=cands,
+        already_recommended_ids=[],
+        touched_or_viewed_ids=[],
+        k=10,
+        recent_author_ids=["a1"],
+        recent_template_ids=None,
+    )
+    got_ids = [x.content_id for x in r.ranked_items]
+    assert got_ids == [2]
+    assert r.meta.missing_history_fields == ["template"]
+    assert r.meta.freqcap_filtered_counts["author"] == 1
+
+
+def test_feed_mmr_picks_diverse_second_item() -> None:
+    # 构造：Top1 是 a1；c2 分数略高但同作者；c3 分数略低但不同作者/阶段
+    c1 = _c(content_id=1, score=1.0, author_id="a1", template_id="t1", stage="general")
+    c2 = _c(content_id=2, score=0.99, author_id="a1", template_id="t2", stage="general")
+    c3 = _c(content_id=3, score=0.95, author_id="a2", template_id="t3", stage="expecting")
+
+    r = rerank_and_freqcap(
+        scene="feed",
+        scored_candidates=[c1, c2, c3],
+        already_recommended_ids=[],
+        touched_or_viewed_ids=[],
+        k=2,
+    )
+    got_ids = [x.content_id for x in r.ranked_items]
+    assert got_ids[0] == 1
+    assert got_ids[1] == 3
+
+
+def test_push_topk_sorted_by_score_after_filters() -> None:
+    cands = [
+        _c(content_id=1, score=0.1),
+        _c(content_id=2, score=0.9),
+        _c(content_id=3, score=0.8),
+    ]
+    r = rerank_and_freqcap(
+        scene="push",
+        scored_candidates=cands,
+        already_recommended_ids=[],
+        touched_or_viewed_ids=[],
+        k=2,
+        recent_author_ids=[],
+        recent_template_ids=[],
+    )
+    got_ids = [x.content_id for x in r.ranked_items]
+    assert got_ids == [2, 3]
+

server/tests/test_scoring.py

@@ -0,0 +1,133 @@
+from __future__ import annotations
+
+from datetime import datetime, timezone
+
+import pytest
+
+from app.features.personalized_reco.content_repository.types import ContentProfileDTO
+from app.features.personalized_reco.scoring.defaults import get_default_config
+from app.features.personalized_reco.scoring.score import score_content
+from app.features.personalized_reco.scoring.types import ExternalTerms, ScoreConfig
+from app.features.user_profile_scoring.types import ProfileAnswered, UserProfileV1_2, UserStageOneHot
+
+
+def _u(
+    *,
+    stage: str = "unknown",
+    emotion_score: float | None = None,
+    need: dict[str, int] | None = None,
+    context: dict[str, int] | None = None,
+    profile_confidence: float = 1.0,
+) -> UserProfileV1_2:
+    now = datetime.now(tz=timezone.utc)
+    if stage == "expecting":
+        s = UserStageOneHot(expecting=1, parenting=0, unknown=0)
+    elif stage == "parenting":
+        s = UserStageOneHot(expecting=0, parenting=1, unknown=0)
+    else:
+        s = UserStageOneHot(unknown=1)
+
+    return UserProfileV1_2(
+        profile_generated_at=now,
+        profile_confidence=profile_confidence,
+        profile_answered=ProfileAnswered(stage=True, emotion=True, context=True, need=True),
+        stage=s,
+        emotion_score=emotion_score,
+        context=context or {},
+        need=need or {},
+    )
+
+
+def _c(
+    *,
+    stage: str = "general",
+    emotion_score: float | None = None,
+    need_key: str = "emotional_support",
+    context_key: str = "family",
+    need_value: float = 1.0,
+    context_value: float = 1.0,
+    personalization_power: float = 1.0,
+    review_confidence: float = 0.7,
+) -> ContentProfileDTO:
+    return ContentProfileDTO(
+        content_id=1,
+        text="hello",
+        stage=stage,  # type: ignore[arg-type]
+        emotion_score=emotion_score,
+        context_suitability={context_key: context_value},
+        need_suitability={need_key: need_value},
+        personalization_power=personalization_power,
+        risk_flags=[],
+        review_confidence=review_confidence,
+    )
+
+
+def test_missing_fields_defaults_are_applied() -> None:
+    u = _u(emotion_score=None, need={}, context={})
+    c = _c(stage="general", emotion_score=None)
+    r = score_content(scene="feed", user_profile=u, content_profile=c)
+
+    assert r.breakdown.S_need == pytest.approx(0.5)
+    assert r.breakdown.S_context == pytest.approx(0.5)
+    assert r.breakdown.S_emotion == pytest.approx(0.8)
+    assert set(r.breakdown.missing_fields) == {"need", "context", "emotion"}
+
+
+def test_uncertainty_penalty_enabled_for_push_by_default_and_can_be_disabled() -> None:
+    u = _u(stage="unknown", emotion_score=0.6, need={"emotional_support": 1}, context={"family": 1}, profile_confidence=0.2)
+    c = _c(stage="general", emotion_score=0.6, personalization_power=1.0, review_confidence=0.2)
+
+    r_on = score_content(scene="push", user_profile=u, content_profile=c)
+    assert r_on.breakdown.P_uncertainty > 0
+
+    cfg_off = ScoreConfig.model_validate(get_default_config("push").model_dump() | {"enable_uncertainty_penalty": False})
+    r_off = score_content(scene="push", user_profile=u, content_profile=c, config=cfg_off)
+    assert r_off.breakdown.P_uncertainty == pytest.approx(0.0)
+    assert r_off.final_score > r_on.final_score
+
+
+def test_widget_emotion_soft_penalty_is_applied_outside_range() -> None:
+    u = _u(stage="unknown", emotion_score=0.6, need={"emotional_support": 1}, context={"family": 1})
+
+    cfg = get_default_config("widget")
+    assert cfg.widget_emotion_soft_range == (0.4, 0.8)
+    assert cfg.widget_emotion_penalty_gamma == pytest.approx(0.25)
+
+    c_in = _c(stage="general", emotion_score=0.6, personalization_power=0.0)
+    r_in = score_content(scene="widget", user_profile=u, content_profile=c_in, config=cfg)
+    assert r_in.breakdown.P_widget_emotion_out_of_range == pytest.approx(0.0)
+
+    c_out = _c(stage="general", emotion_score=0.0, personalization_power=0.0)
+    r_out = score_content(scene="widget", user_profile=u, content_profile=c_out, config=cfg)
+    assert r_out.breakdown.P_widget_emotion_out_of_range == pytest.approx(0.25)
+    assert r_out.final_score < r_in.final_score
+
+
+def test_pass_false_forces_final_score_zero_but_breakdown_is_present() -> None:
+    u = _u(stage="unknown", emotion_score=0.6, need={"emotional_support": 1}, context={"family": 1})
+    c = _c(stage="general", emotion_score=0.6, personalization_power=1.0)
+    r = score_content(scene="feed", user_profile=u, content_profile=c, pass_filters=False, external_terms=ExternalTerms())
+
+    assert r.final_score == pytest.approx(0.0)
+    assert r.breakdown.passed is False
+    # breakdown 字段集合稳定（至少包含关键分解项）
+    d = r.breakdown.model_dump(by_alias=True)
+    for k in [
+        "scene",
+        "pass",
+        "S_need",
+        "S_context",
+        "S_stage",
+        "S_emotion",
+        "S_core",
+        "S_personal",
+        "S_fresh",
+        "P_fatigue",
+        "P_repeat",
+        "P_risk",
+        "P_uncertainty",
+        "P_widget_emotion_out_of_range",
+        "missing_fields",
+    ]:
+        assert k in d
+

spec_kit/Client User Identity/spec.md

@@ -0,0 +1,151 @@
+# Client User Identity｜客户端用户标识建立（用于 PUSH Token 绑定）｜Spec
+
+> 阶段：高层规范（spec）
+>
+> 目标：在**无账号体系或账号可选**的前提下，为客户端生成一个稳定的“客户端用户标识”（下称 `client_user_id`），用于与 APNs/FCM 的 Push Token 建立绑定关系，便于后端精准下发 PUSH，并支持 Token 变更/多设备/环境隔离等场景。
+
+---
+
+## 1. 背景与动机（摘要）
+
+Push Token（APNs device token / FCM registration token）会发生变化（重装、系统升级、重新授权、token rotate 等），且同一用户可能多设备。为了能稳定地“找到这个客户端实例/用户侧主体”并维护 Token 映射，需要一个**与 Token 解耦**、可持久化、不可推断的标识。
+
+---
+
+## 2. 目标（Goals）
+
+- **建立 `client_user_id`**：客户端可生成并持久化一个稳定标识，作为后端 Push Token 绑定的主键之一。
+- **Token 绑定可更新**：支持 Token 变更时“同一 `client_user_id` 重新上报即可更新”。
+- **多设备兼容**：同一个账号（若未来引入）可关联多个 `client_user_id`；一个 `client_user_id` 可存在多个 Token（例如同设备多渠道/多应用包形态）时需可扩展。
+- **环境隔离**：dev/prod、iOS/Android、bundle id / package name 维度隔离，避免串绑。
+- **隐私友好**：不使用可追踪的硬件标识（IMEI/IDFA/Android ID 等），不引入额外合规风险。
+
+---
+
+## 3. 非目标（Non-goals）
+
+- 不在本阶段引入完整账号体系、登录态、用户合并策略（如“同一人多设备合并为一个 user_id”）。
+- 不在本阶段强制接入设备证明（App Attest/Play Integrity）；仅在安全章节提出可选增强方向。
+- 不定义具体数据库表结构与迁移脚本（属于 plan 阶段细化）。
+
+---
+
+## 4. 术语与对象（Definitions）
+
+- **`client_user_id`**：客户端生成并持久化的随机标识，代表“一个客户端安装实例（或一段时间内的用户侧主体）”，用于 Push 绑定。
+- **`push_token`**：系统/厂商下发的推送 token（iOS/APNs，Android/FCM），可能变化。
+- **`account_id`（可选）**：若未来存在登录账号，则用于把多个 `client_user_id` 归属到同一账号。
+
+---
+
+## 5. 关键决策：使用 UUID 做 `client_user_id` 是否合适？
+
+结论：**合适**，推荐用**随机 UUID（UUID v4 为默认）**，并把它当作后端与客户端都不解析的**不透明字符串**。
+
+### 5.1 为什么 UUID 合适
+
+- **唯一性足够**：v4 基于随机数，碰撞概率极低，满足全局唯一需求。
+- **不可推断**：相较自增 ID，不易被枚举；相较设备硬件标识，更隐私友好。
+- **跨端易实现**：iOS/Android/JS 都可稳定生成与序列化（字符串）。
+
+### 5.2 需要明确的边界与注意事项
+
+- **UUID 不等于“真实用户”**：它更像“安装实例 ID”。用户重装/清数据后可能变化；这对 Push 绑定通常可接受（新装产生新 token 与新 id）。
+- **不要用设备硬件/系统可追踪 ID 替代**：避免隐私与合规风险，也避免系统限制导致的不稳定。
+- **安全边界**：如果后端完全信任客户端上报的 `client_user_id`，存在“伪造绑定”风险；应结合登录态或签名/证明（见第 9 节）降低滥用。
+
+### 5.3 UUID 版本建议
+
+- **默认**：UUID v4（实现最简单、兼容最好）。
+- **可选增强**：UUID v7（有时间有序性，利于日志/索引与写入局部性），但需要确保两端实现一致与依赖可控。
+
+---
+
+## 6. 客户端行为规范（Client Contract）
+
+### 6.1 生成与持久化
+
+- 首次启动（或首次需要注册 Push 时）：
+  - 若本地不存在 `client_user_id`：生成一个新的 UUID（字符串），写入持久化存储。
+  - 若已存在：直接复用。
+- 存储建议（不做强约束，但必须“尽量稳定”）：
+  - iOS：Keychain
+  - Android：Keystore 保护的加密存储/SharedPreferences（或等价方案）
+  - React Native/Expo：使用安全存储能力（例如 SecureStore/Keychain wrapper）
+
+### 6.2 Push Token 获取与上报时机
+
+- 在以下任一时机触发“注册/更新”：
+  - 用户同意 Push 权限后获得 token
+  - App 冷启动获取到 token（含 token 变更）
+  - 账号登录/登出（若存在账号）
+  - 环境切换（dev/prod）或应用更新（可选）
+
+---
+
+## 7. 后端接口契约（API Contract，摘要）
+
+> 具体路由/鉴权方式在 plan 阶段落地；此处先定义字段语义与幂等行为。
+
+### 7.1 注册/更新绑定
+
+- `POST /v1/push/register`
+- 请求体（最小集）：
+  - `client_user_id`: string（UUID 字符串）
+  - `platform`: `"ios" | "android"`
+  - `push_token`: string
+  - `app_id`: string（bundle id / package name，用于隔离）
+  - `env`: `"dev" | "prod"`
+  - （可选）`account_id`: string
+  - （可选）`device_meta`: `{ model, os_version, app_version, locale, timezone }`
+- 行为要求（幂等）：
+  - 以 `push_token + env + app_id` 维度做唯一性约束，避免重复记录。
+  - 若同一 `client_user_id` 上报了新 token：应更新/新增映射，旧 token 进入失效或保留历史（由实现决定，但必须可控）。
+
+### 7.2 解绑（可选但建议）
+
+- `POST /v1/push/unregister`
+- 请求体：
+  - `client_user_id`
+  - `platform`
+  - `push_token`（或让后端按 `client_user_id` 批量解绑，二选一）
+  - `app_id`
+  - `env`
+
+---
+
+## 8. 数据模型（逻辑约束）
+
+最小需要表达的关系：
+
+- 一个 `client_user_id` 可对应 0..N 个 `push_token`（考虑多端、多渠道、token rotate）。
+- 一个 `push_token` 在同一 `env + app_id` 下应只对应一个“当前归属”（避免重复推送）。
+- 若存在 `account_id`：
+  - 一个 `account_id` 可关联 0..N 个 `client_user_id`（多设备）。
+
+---
+
+## 9. 安全与滥用防护（高层约束）
+
+- **最小要求**：接口需具备基本鉴权与频率限制（例如基于设备指纹/匿名 session/应用侧签名的任一组合），避免被脚本批量绑定垃圾 token。
+- **若存在登录态**：推荐绑定写入需要登录态（或在登录后把 `client_user_id` 归属到 `account_id`），降低“抢绑”风险。
+- **可选增强（后续）**：接入 iOS App Attest / Android Play Integrity，或对注册请求做一次性挑战签名。
+
+---
+
+## 10. 边界场景与处理原则
+
+- **用户拒绝 Push 权限**：允许只有 `client_user_id`，不产生 token 绑定；后端不应报错。
+- **token 变化**：客户端重新调用 `register`，后端必须幂等更新，避免重复推送。
+- **重装/清数据**：`client_user_id` 变化可接受；若未来有 `account_id`，可在登录后重新建立关联。
+- **多环境**：dev/prod token 不可混用；必须以 `env + app_id` 隔离。
+
+---
+
+## 11. 验收标准（Acceptance Criteria）
+
+- 客户端能稳定生成并持久化 `client_user_id`（重复启动不变）。
+- 在 token 获取/变更后，调用注册接口可在后端建立（或更新）绑定关系，且接口幂等。
+- 同一 `push_token` 在同一 `env + app_id` 下不会产生多条“当前有效”绑定，避免重复推送。
+- 在用户拒绝 Push 权限、无 token 的情况下，不影响 App 正常使用与后续再次授权后的绑定。
+

spec_kit/Personalized Reco/modules/content-repository/plan.md

@@ -0,0 +1,275 @@
+# Content Repository（候选查询与数据访问层）｜Plan
+
+> 对应规范：`spec_kit/Personalized Reco/modules/content-repository/spec.md`
+>
+> 依赖对齐：
+>
+> - DB 设计：`spec_kit/Personalized Reco/modules/db-design/plan.md`
+> - 回退梯度与场景口径：`spec_kit/Personalized Reco/spec.md`（Fallback Ladder + 场景默认参数）
+
+---
+
+## 1. 目标与交付物
+
+### 1.1 目标
+
+- 为推荐引擎提供**与 ORM/SQL 解耦**的数据访问接口：候选召回与按 ID 批量获取。
+- 将 DB 内部存储形态（JSON、关联表、NULL 语义等）统一“规范化”为上层稳定的 `ContentProfile` 结构。
+- 在候选不足时支持按 `fallback_level (L0~L3)` 进行**可控降级**（降个性化/回退通用池/安全池），并且不产生 N+1 查询。
+
+### 1.2 交付物
+
+- `modules/content-repository/plan.md`：本技术计划（本文件）。
+- 代码实现（后续 tasks 阶段落地）建议位置：
+  - `server/app/features/personalized_reco/content_repository/`（或等价目录）
+  - 需要包含：
+    - 抽象接口 `ContentRepository`（Protocol 或 ABC）
+    - SQLAlchemy 实现 `SqlAlchemyContentRepository`
+    - `ContentProfile`（DTO/数据结构）与规范化工具函数
+- 单元测试与最小集成测试（后续 tasks 阶段落地）：
+  - risk_flags 映射与去重
+  - suitability 缺失兜底
+  - personalization_power 映射
+  - 查询不出现按 `content_id` 循环查 flags（避免 N+1）
+
+---
+
+## 2. 关键技术决策（V1）
+
+### 2.1 存储形态（来自 DB Design 的已对齐结论）
+
+- `content_id`：MySQL 自增主键（`int`）。
+- `context_suitability_json / need_suitability_json`：JSON 存储。
+- `risk_flags`：关联表 `content_risk_flags(content_id, flag)`（同一 content 下 `uniq_content_flag` 去重）。
+- `emotion_score`：`NULL` 表示 general。
+- `review_confidence`：DB 可为 `NULL`；读取层输出时按 `0.7` 兜底（对齐规则口径）。
+- `personalization_power`：DB 推荐存 `0/5/10`；读取层输出稳定为 `0.0/0.5/1.0`。
+
+### 2.2 职责边界（避免耦合）
+
+- `Content Repository` **不负责** Hard Filter/Soft Scoring/Rerank/Freqcap（这些由引擎编排与打分子模块完成）。
+- `Content Repository` **负责**：
+  - 按 `fallback_level` 对候选池做“降级约束”（例如限制 `personalization_power`、回退通用池/安全池）
+  - 输出稳定结构（JSON 解析、默认值兜底、旧 flag 映射）
+
+---
+
+## 3. 接口与数据结构（V1）
+
+### 3.1 接口定义（与 spec 对齐）
+
+- `fetch_candidates(scene, user_profile, fallback_level, limit, locale, exclude_content_ids=None) -> List[ContentProfile]`
+- `fetch_contents_by_ids(content_ids: List[int], locale) -> List[ContentProfile]`
+
+### 3.2 `ContentProfile`（输出契约的推荐形态）
+
+稳定字段（必须输出）：
+
+- `content_id: int`
+- `text: str`
+- `stage: "general" | "expecting" | "parenting" | "unknown"`
+- `emotion_score: float | None`（`None` 表示 general）
+- `context_suitability: Dict[str, float]`
+- `need_suitability: Dict[str, float]`
+- `personalization_power: float`（`0/0.5/1`）
+- `risk_flags: List[str]`
+
+可选字段（尽量输出）：
+
+- `author_id: str | None`
+- `template_id: str | None`
+- `review_confidence: float`（缺失/NULL 按 `0.7` 输出）
+
+---
+
+## 4. 读取层规范化（Normalization）
+
+### 4.1 text 选文案与多语言策略（不允许回退）
+
+数据来源（当前 DB/ORM 约定）：
+
+- `contents.text_en`：英文
+- `contents.text_tc`：繁体中文
+
+规则（**不允许语言回退**）：
+
+- `locale=en*`：仅允许返回存在 `text_en` 的内容；输出 `text = text_en`
+- `locale=tc/zh-TW/zh-HK`：仅允许返回存在 `text_tc` 的内容；输出 `text = text_tc`
+
+若内容缺少目标语言文本（例如 `locale=en*` 但 `text_en` 为空）：该内容视为不可用，必须在候选/按 ID 获取时过滤掉。
+
+### 4.1 suitability JSON 解析与缺失兜底
+
+固定 key 集合（对齐 DB Plan 的最小入库契约）：
+
+- `context_suitability`：`family/work/relationship/friends/health`
+- `need_suitability`：`emotional_support/parenting_pressure/self_worth/anxiety_relief/rest_balance`
+
+规则：
+
+- 若 DB 字段缺失/为 `NULL`/解析失败：**补齐为全 0.5**（以上所有 key 均为 `0.5`）。
+- 若 DB JSON 存在但缺少部分 key：对缺少 key 补 `0.5`，其余按原值。
+- 值域约束：期望值为 `0/0.5/1`；若出现其他值（例如字符串、越界浮点），按 `0.5` 兜底并记录告警日志（V1 可先打 debug，后续接入可观测模块）。
+
+### 4.2 `review_confidence` 兜底
+
+- DB `review_confidence` 为 `NULL` 或缺失：输出 `0.7`。
+
+### 4.3 `personalization_power` 映射
+
+若 DB 存 `0/5/10`：
+
+- `0 -> 0.0`
+- `5 -> 0.5`
+- `10 -> 1.0`
+
+若读到其他值：按 `0.0` 兜底并记录告警日志。
+
+### 4.4 risk_flags 旧→新映射与输出约束
+
+映射表（对齐 spec）：
+
+- `block_stage_unknown` → `unsafe_for_stage_unknown`
+- `block_stage_parenting` → `unsafe_for_stage_parenting`
+- `block_emotion_low` → `unsafe_for_emotion_low`
+- `block_health_sensitive` → `block_health_medical`（V1 保守硬拦截）
+
+输出约束：
+
+- 输出 `risk_flags` 必须去重。
+- 输出不得包含旧命名。
+- 输出建议稳定排序（便于测试与可观测）：按字典序排序或按严重等级排序（V1 可先字典序）。
+
+---
+
+## 5. 查询策略（V1）
+
+> 原则：DB 层先做“粗过滤”，应用层再做“精过滤/打分”。避免在 V1 过早依赖 JSON 路径查询索引。
+
+### 5.1 `fetch_contents_by_ids`（按 ID 批量获取）
+
+目标：
+
+- 输入任意 `content_id` 列表，返回无重复的 `ContentProfile` 列表。
+- 避免 N+1：不得按 `content_id` 循环查 `content_risk_flags`。
+- 返回顺序：必须与输入 `content_ids` 一致（对“缺记录/缺语言文本”的 id 采取跳过策略，见下）。
+
+缺记录/缺语言文本的处理（V1 约定）：
+
+- 若某个 `content_id` 在 DB 中不存在，或按 `locale` 规则无法产出 `text`：该 id 在返回列表中**跳过**（不返回占位对象）。
+
+推荐实现形态（两段式，避免 JOIN 导致重复行）：
+
+1. **批量拉主体与画像**（`contents` JOIN `content_profiles`），限制 `content_id IN (...)`。
+2. **批量拉 risk_flags**：`SELECT content_id, flag FROM content_risk_flags WHERE content_id IN (...)`，在应用层按 `content_id` 聚合为集合，再做旧→新映射与去重。
+
+备注：
+
+- 由于 `content_risk_flags` 是 1:N，直接三表 JOIN 容易导致行膨胀；两段式更便于组装与去重。
+
+### 5.2 `fetch_candidates`（候选召回，支持 L0~L3）
+
+输入：
+
+- `scene: feed | push | widget`
+- `user_profile`（允许字段缺失）
+- `fallback_level: 0|1|2|3`
+- `limit`
+- `exclude_content_ids`（可选）
+
+#### 5.2.1 fallback_level 约束（对齐大规范 Fallback Ladder）
+
+从 `spec_kit/Personalized Reco/spec.md` 对齐：
+
+- **L0**：正常召回配比（不在读取层实现复杂配比，读取层只保证候选池足够大且不过度放宽）
+- **L1**：放宽匹配 + 降个性化：限制 `personalization_power ≤ 0.5`
+- **L2**：回退通用池 + 进一步降个性化：限制 `personalization_power = 0`，且优先 `stage=general`
+- **L3**：兜底安全池：限制 `is_safe_pool = true`（安全池字段来自 DB Design）
+
+缺失字段的最小处理（对齐大规范 Candidate Generation 口径）：
+
+- 若 `user_profile` 缺失明显（need/context/emotion 任一缺失）：读取层按**至少 L1** 的约束执行（即使入参 fallback_level=0）。
+
+#### 5.2.2 stage 粗过滤策略（V1）
+
+读取层可做的最小粗过滤（不引入复杂业务判断）：
+
+- **L2/L3**：只取 `stage=general`（L3 额外 `is_safe_pool=true`）。
+- **L0/L1**：
+  - 优先取 `stage=用户匹配阶段` + `stage=general`
+  - 若无法从 `user_profile` 明确阶段，则仅取 `stage=general`（避免误推）
+
+> 说明：更细粒度的阶段/跨维度规则（例如 unknown+parenting_pressure 的禁推）由 Hard Filter 子模块实现；读取层仅做粗过滤以减少扫描与传输。
+
+#### 5.2.3 查询形态（避免 JOIN 行膨胀 + 保证 limit）
+
+推荐采用“两段式候选召回”：
+
+1. **先只查候选 ID 列表**（`contents` JOIN `content_profiles`），应用粗过滤（stage / personalization_power / is_safe_pool / exclude_content_ids），并增加 **locale 文本存在性过滤**（不允许语言回退），再用 `LIMIT limit * multiplier` 拉一批候选 ID（`multiplier` 例如 3~5，避免后续去重/过滤后不足）。
+2. **再用 `fetch_contents_by_ids` 批量补全字段**（主体+画像+risk_flags），最终在应用层去重并截断到 `limit`。
+
+排序（V1）：
+
+- 若没有更明确的排序字段：使用 `updated_at DESC` 或随机抽样（需谨慎，MySQL `ORDER BY RAND()` 在大表会慢）。
+- 推荐：V1 先用 `content_profiles.updated_at DESC` 或 `contents.created_at DESC`，后续由打分模块决定最终排序。
+
+---
+
+## 6. 性能与可观测（V1）
+
+### 6.1 性能约束
+
+- 单次调用不得出现按 `content_id` 循环查库（避免 N+1）。
+- `fetch_candidates` 必须在 DB 层支持 `limit`，并尽量通过粗过滤减少扫描。
+
+### 6.2 建议打点/日志（为 observability 子模块预留）
+
+在 Repository 层建议输出 debug 级日志（或埋点字段，供上层汇总）：
+
+- `scene`
+- `fallback_level`（入参）与 `effective_fallback_level`（考虑缺失字段自动至少 L1 后的实际约束级别）
+- `limit`、`exclude_content_ids_count`
+- `candidate_ids_size_raw`（第 1 段查到的候选 ID 数）
+- `candidate_size_returned`（最终返回数量）
+
+---
+
+## 7. 测试计划（V1）
+
+### 7.1 单元测试（纯函数）
+
+- risk_flags 映射：
+  - 输入包含旧 flag，输出只包含新命名
+  - 去重与稳定排序
+- suitability 兜底：
+  - DB 字段缺失/NULL/解析失败 → 全 0.5
+  - 部分 key 缺失 → 补齐 0.5
+- personalization_power 映射：
+  - 0/5/10 → 0.0/0.5/1.0
+  - 异常值 → 0.0 兜底
+- review_confidence：
+  - NULL/缺失 → 0.7
+
+### 7.2 最小集成测试（含数据库）
+
+- `fetch_contents_by_ids`：
+  - 输入多个 id 返回无重复
+  - flags 聚合正确（同一 content 多条 flag 行能聚合成 list）
+- 查询次数断言（避免 N+1）：
+  - `fetch_contents_by_ids`：固定 2 次查询（主体+画像一次，flags 一次）
+  - `fetch_candidates`：固定 3 次查询（候选 id 一次 + `fetch_contents_by_ids` 两次），或实现允许的常数级次数
+
+---
+
+## 8. 风险与后续演进
+
+### 8.1 已知风险
+
+- V1 不做 JSON 路径索引：候选量变大后，粗过滤不足可能导致候选池拉取过多、应用层过滤成本上升。
+- `ORDER BY RAND()` 的性能风险：候选大表下不可用，需要替代策略（时间窗口抽样/预生成候选池）。
+
+### 8.2 V1.1 优化方向（与 DB Plan 对齐）
+
+- 为常用 need/context key 增加生成列/函数索引（从 JSON_EXTRACT 提取到 TINYINT）以加速召回。
+- 为强规则风险（如 `block_health_medical`）增加派生布尔列或缓存表，减少 JOIN 成本。
+

spec_kit/Personalized Reco/modules/content-repository/spec.md

@@ -0,0 +1,105 @@
+# 子模块：Content Repository（候选查询与数据访问层）｜Spec
+
+## 1. 目标描述
+
+提供推荐算法可注入的、与 ORM/SQL 解耦的数据访问接口：
+
+- 按场景与用户画像拉取候选内容画像（Cᵢ）。
+- 按 `content_id` 批量获取内容画像（去重/重排/补字段）。
+- 对 risk_flags、suitability JSON 等“存储形态”做统一解析与兼容（对上提供稳定结构）。
+
+> 规则口径：risk_flags 命名与语义严格按 `句子文案打分规则`；若历史数据存在旧 flag，需在读取层做一次映射（避免语义漂移）。
+>
+> 存储形态对齐 `modules/db-design/plan.md`：读取层需要将 `contents` + `content_profiles` + `content_risk_flags` 组装为上层稳定的 `ContentProfile` 结构。
+
+---
+
+## 2. 输入 / 输出定义
+
+### 2.1 输入
+
+- `scene`: `feed | push | widget`
+- `user_profile`: 客户端问卷画像（V1.2；字段允许缺失）
+- `locale`：客户端语言（由请求携带并透传至推荐模块；**当前仅支持 EN/TC**，例如 `en` / `en-US` / `tc` / `zh-TW` / `zh-HK`）
+- `fallback_level`: `0|1|2|3`
+- `limit`: 候选条数上限（由引擎配置）
+- （可选）排除集合：`exclude_content_ids`（用于 DB 层先排一部分，减少传输；类型为 `List[int]`，与 MySQL 自增 `content_id` 对齐）
+- 数据库会话/连接（实现层使用 `AsyncSession` 注入）
+
+### 2.2 输出
+
+- `List[ContentProfile]`（稳定字段契约）：
+  - `content_id`：`int`（MySQL 自增主键）
+  - `text`：按 `locale` 输出的文案文本（**不允许语言回退**）
+    - `locale=en*`：必须从 `contents.text_en` 产出；若该 content 无 `text_en`，则该 content 不可返回（在候选/按 ID 获取时过滤）
+    - `locale=tc/zh-TW/zh-HK`：必须从 `contents.text_tc` 产出；若无 `text_tc`，则该 content 不可返回
+  - `stage`：`general | expecting | parenting | unknown`
+  - `emotion_score`：`float | None`（约定：`None` 表示 general）
+  - `context_suitability`：`Dict[str, float]`（见 4.1 的 key 集合；值为 `0/0.5/1`）
+  - `need_suitability`：`Dict[str, float]`（见 4.1 的 key 集合；值为 `0/0.5/1`）
+  - `personalization_power`：`float`（对上稳定口径为 `0/0.5/1`；若 DB 存 `0/5/10`，读取层需映射）
+  - `risk_flags`：`List[str]`（已做旧→新映射、去重；命名只允许 `unsafe_for_* / block_* / soft_*`）
+  - 可选：`author_id`、`template_id`、`review_confidence`
+
+### 2.3 数据存储形态（对齐 DB 设计）
+
+> 对齐：`spec_kit/Personalized Reco/modules/db-design/plan.md`
+
+- `contents`：提供 `content_id`、`text`、（可选）`author_id`、`template_id`
+- `content_profiles`：提供 `stage`、`emotion_score`、`context_suitability_json`、`need_suitability_json`、`personalization_power`（推荐存 `0/5/10`）、（可选）`review_confidence`
+- `content_risk_flags`：通过关联表提供风险标记集合（同一 content 下按 `uniq_content_flag(content_id, flag)` 去重）
+
+---
+
+## 3. 接口（建议）
+
+推荐模块对该子模块只依赖抽象接口（Python Protocol/ABC 均可）：
+
+- `fetch_candidates(scene, user_profile, fallback_level, limit, exclude_content_ids=None) -> List[ContentProfile]`
+- `fetch_contents_by_ids(content_ids: List[int]) -> List[ContentProfile]`
+
+---
+
+## 4. 关键规则与实现约束
+
+### 4.1 字段缺失与默认值
+
+- 若 `review_confidence` 缺失：输出时默认按 `0.7`（对齐 `句子文案打分规则` V1.2 约定）。
+- `context_suitability/need_suitability` 若缺失：**读取层必须补齐为“全 0.5 的通用可推”结构**（稳定输出，避免上层分支判断）。
+  - `context_suitability` 必须包含 5 个 key：`family/work/relationship/friends/health`
+  - `need_suitability` 必须包含 5 个 key：`emotional_support/parenting_pressure/self_worth/anxiety_relief/rest_balance`
+  - 补齐时上述 key 的默认值均为 `0.5`
+- `personalization_power`：若 DB 采用 `0/5/10` 存储，读取层必须映射为 `0.0/0.5/1.0` 对上输出。
+
+### 4.2 risk_flags 兼容映射（若存在历史旧数据）
+
+对齐 `句子文案打分规则` 的旧→新映射：
+
+- `block_stage_unknown` → `unsafe_for_stage_unknown`
+- `block_stage_parenting` → `unsafe_for_stage_parenting`
+- `block_emotion_low` → `unsafe_for_emotion_low`
+- `block_health_sensitive` → `block_health_medical`（读取层默认采用更保守的硬拦截映射；除非未来引入可判定的细分字段再放宽为 `soft_health_sensitive`）
+
+输出约束：
+
+- 输出的 flags 必须已去重，且不得包含任何旧命名。
+
+### 4.3 性能约束
+
+- 不得产生 N+1 查询：候选与字段必须一次或少量批量查询获取。
+- `fetch_candidates` 必须支持 limit，并在 DB 层尽量过滤（减少应用层扫描）。
+- `fetch_contents_by_ids` / `fetch_candidates` 推荐查询形态：
+  - `contents` JOIN `content_profiles`，再 LEFT JOIN `content_risk_flags`（或先批量取 profiles，再批量取 flags 并在应用层聚合），避免按 `content_id` 循环查 flags。
+
+---
+
+## 5. 验收标准（可验证）
+
+- `fetch_contents_by_ids`：
+  - 输入任意 `content_id` 列表，返回包含完整字段的 `ContentProfile` 列表（无重复、可缺省字段按约定兜底）。
+- `fetch_candidates`：
+  - 在不同 `scene` 与 `fallback_level` 下能返回候选（即便画像缺失也不报错）。
+  - risk_flags 映射正确：输出的 flag 名称集合只包含新命名（`unsafe_for_* / block_* / soft_*`）。
+- 性能：
+  - 单次调用不出现按 content_id 循环查库的行为（可通过日志/测试断言查询次数）。
+

spec_kit/Personalized Reco/modules/content-repository/tasks.md

@@ -0,0 +1,198 @@
+# Content Repository（候选查询与数据访问层）｜Tasks
+
+> 对应计划：`spec_kit/Personalized Reco/modules/content-repository/plan.md`
+>
+> 本清单已对齐确认点：
+>
+> - `text` 按客户端 **locale** 输出（请求携带）
+> - **不允许语言回退**（缺少目标语言文本的内容直接过滤，不返回）
+> - 需要做 **MySQL（dev）DB 集成测试**
+> - 代码与其他推荐子模块放同一目录：`server/app/features/personalized_reco/`
+> - `fetch_contents_by_ids` 返回顺序必须与输入 `content_ids` **一致**
+>
+> 执行说明：
+>
+> - 本次已在 dev MySQL 环境跑通 `pytest`，且测试不做破坏性操作：
+>   - 不清表（不执行 DELETE/TRUNCATE）
+>   - 每个用例使用事务并在结束时 rollback
+>   - 默认不执行 Alembic 迁移（如需自动迁移需显式设置 `ALLOW_SCHEMA_MIGRATION=1`）
+
+---
+
+## 0. 任务标记规则
+
+- 用勾选框标记执行状态：
+  - `[ ]` 未开始
+  - `[x]` 已完成
+- 每个任务都要求可独立验收（有明确产出/可运行的检查方式）。
+
+---
+
+## 1. 文档对齐（先把口径写死，避免实现漂移）
+
+- [x] 1.1 更新 `modules/content-repository/spec.md`，加入 locale 相关契约
+  - **变更点**：
+    - 在输入中新增 `locale`（例如 `en`/`en-US`/`tc`/`zh-TW`/`zh-HK`），声明由客户端请求携带并透传至 repository
+    - 在输出字段 `text` 补充语言选择规则：
+      - `locale=en*`：必须从 `text_en` 产出；若缺失则该内容不可返回（过滤）
+      - `locale=zh-TW|zh-HK`：必须从 `text_tc` 产出；若缺失则该内容不可返回（过滤）
+      - 说明：当前仅支持 EN/TC（不做繁转简）；若未来新增 `zh-CN` 再单独设计转换策略
+    - 明确：`fetch_contents_by_ids` **返回顺序与入参一致**
+  - **验收**：`spec.md` 中输入/输出与接口签名不再缺少 locale，且 `text` 的来源与“不允许语言回退”规则清晰。
+
+- [x] 1.2 更新 `modules/content-repository/plan.md`，补齐 locale 选文与“繁转简”技术方案
+  - **变更点**：
+    - 在“读取层规范化（Normalization）”新增 `text` 规范化章节：语言选择 + 简中转换
+    - 说明：当前仅支持 EN/TC（不做繁转简）；若未来新增 `zh-CN` 再单独设计转换策略
+    - 明确策略：**不允许语言回退**；缺少目标语言文本的内容视为不可用，必须在候选/按 ID 获取时过滤掉
+  - **验收**：`plan.md` 有明确依赖与落地策略（含依赖包/转换时机/兜底策略），不留二义性。
+
+---
+
+## 2. 目录与骨架（与推荐子模块同级）
+
+- [x] 2.1 新建目录 `server/app/features/personalized_reco/content_repository/`
+  - **包含**：
+    - `__init__.py`
+    - `types.py`（DTO：`ContentProfile`、locale 类型等）
+    - `interface.py`（`ContentRepository` Protocol/ABC）
+    - `normalization.py`（解析与兜底：suitability、risk_flags、power、text）
+    - `sqlalchemy_repo.py`（SQLAlchemy 实现）
+  - **验收**：可被 `app.features.personalized_reco.content_repository.*` 正常 import。
+
+- [ ] 2.2 依赖补齐（如采用 OpenCC）
+  - **说明**：当前仅支持 EN/TC，此任务可跳过；若未来新增 `zh-CN` 并需要繁转简，再引入 OpenCC。
+
+---
+
+## 3. 数据结构与接口（面向引擎注入）
+
+- [x] 3.1 定义 `ContentProfile` DTO（稳定字段契约）
+  - **字段**：对齐 `modules/content-repository/spec.md`，并补齐 `review_confidence` 输出兜底为 `0.7`
+  - **注意**：`text` 为最终对外输出文本（已按 locale 选择/转换）
+  - **验收**：DTO 字段齐全；类型清晰；不暴露 ORM 模型。
+
+- [x] 3.2 定义 `ContentRepository` 接口（含 locale）
+  - **建议签名**（示例，最终以 spec 为准）：
+    - `fetch_candidates(scene, user_profile, fallback_level, limit, locale, exclude_content_ids=None) -> List[ContentProfile]`
+    - `fetch_contents_by_ids(content_ids, locale) -> List[ContentProfile]`
+  - **验收**：推荐引擎可以仅依赖该接口，不依赖 SQLAlchemy/FastAPI Depends。
+
+---
+
+## 4. 规范化工具函数（可单测）
+
+- [x] 4.1 suitability 解析与兜底
+  - **规则**：
+    - 缺失/NULL/解析失败 → 全 0.5（固定 key 集合）
+    - 部分 key 缺失 → 对缺失 key 补 0.5
+    - 非法值 → 兜底 0.5
+  - **验收**：单元测试覆盖缺失/部分缺失/非法值。
+
+- [x] 4.2 risk_flags 映射、去重与排序
+  - **规则**：旧→新映射对齐 spec；去重；稳定排序（例如字典序）
+  - **验收**：单元测试断言输出不含旧命名且顺序稳定。
+
+- [x] 4.3 personalization_power 映射
+  - **规则**：`0/5/10 -> 0.0/0.5/1.0`；非法值 -> 0.0
+  - **验收**：单元测试覆盖正常/异常值。
+
+- [x] 4.4 text 选择与简中转换
+  - **输入**：`text_en`、`text_tc`、`locale`
+  - **规则**：按 1.1/1.2 写死的策略执行（当前仅支持 EN/TC，不做繁转简）
+  - **验收**：单元测试覆盖：
+    - `en` 取英文
+    - `zh-TW` 取繁中
+    - 缺失目标语言文本时的行为：返回“不可用”（例如返回空字符串 + 上层过滤，或直接返回 `None` 由调用方过滤；实现中必须一致）
+
+---
+
+## 5. SQLAlchemy 实现（无 N+1、顺序可控）
+
+> 说明：当前 DB 模型为：
+>
+> - `contents`：`text_en` / `text_tc` / `author_id` / `template_id`
+> - `content_profiles`：JSON、power、stage、is_safe_pool、review_confidence
+> - `content_risk_flags`：关联表（1:N）
+
+- [x] 5.1 实现 `fetch_contents_by_ids(content_ids, locale)`
+  - **实现要点**：
+    - 输入去重，但输出必须按原始输入顺序重排（并忽略不存在的 id 或明确行为：不存在则跳过）
+    - 两段式查询避免行膨胀：
+      1) `contents` JOIN `content_profiles` 批量取主体与画像
+      2) `content_risk_flags` 批量取 flags，再按 `content_id` 聚合
+    - 组装 DTO 时执行 normalization（含 text locale 规则）
+  - **验收**：
+    - 返回顺序与输入一致
+    - 缺失目标语言文本的 content_id 不返回（跳过，不做语言回退）
+    - 不产生按 id 循环查 flags 的查询（查询次数为常数级）
+
+- [x] 5.2 实现 `fetch_candidates(scene, user_profile, fallback_level, limit, locale, exclude_content_ids)`
+  - **实现要点**：
+    - 计算 `effective_fallback_level`：
+      - 若画像 need/context/emotion 任一缺失，则 `effective_fallback_level = max(fallback_level, 1)`
+    - DB 粗过滤对齐 plan：
+      - L1：`personalization_power <= 5`
+      - L2：`personalization_power = 0` 且 `stage = general`
+      - L3：`is_safe_pool = true` 且 `stage = general` 且 `personalization_power = 0`（如需更严格可在此明确）
+    - 排序（V1）：按 `content_profiles.updated_at DESC` 或 `contents.updated_at DESC`（择一写死并记录）
+    - 两段式候选：
+      1) 先查候选 id（`LIMIT limit * multiplier`）
+      2) 调用 `fetch_contents_by_ids` 补全字段
+    - locale 文本存在性过滤：
+      - `locale=en*`：`contents.text_en IS NOT NULL`
+      - `locale=zh-*`：`contents.text_tc IS NOT NULL`
+    - 输出顺序：
+      - 返回顺序按候选 id 列表顺序（用于后续引擎打分/重排）；最终截断至 `limit`
+  - **验收**：
+    - 在不同 `effective_fallback_level` 下能返回候选
+    - 不返回缺少目标语言文本的内容（不做语言回退）
+    - 查询次数为常数级（不随 `limit` 线性增长）
+
+---
+
+## 6. DB 集成测试（dev MySQL）
+
+- [x] 6.1 建立测试目录与 pytest 配置
+  - **目标**：在 `server/` 内新增 `tests/`（或 `app/**/__tests__/`，但建议统一为 `server/tests/`）
+  - **内容**：
+    - `server/tests/conftest.py`：提供 AsyncEngine/AsyncSession、清库策略、query count 统计工具
+    - 测试运行约定：通过 `DATABASE_URL` 指向 dev 测试库（建议单独库名，例如 `mindfulness_dev_test`）
+  - **验收**：`pytest` 可在 `server/` 下运行并发现测试。
+
+- [x] 6.2 测试库 schema 初始化（用 Alembic）
+  - **策略**（二选一写死）：
+    - A：测试启动时 `alembic upgrade head`（确保 schema 最新）
+    - B：在 CI/本地提前准备库，仅在测试中清表
+  - **验收**：测试运行前 schema 可用，且不会污染开发主库数据（推荐使用独立 test 库）。
+
+- [x] 6.3 集成测试用例：`fetch_contents_by_ids`
+  - **准备数据**：插入最小内容 2~3 条（覆盖 text_en/text_tc 缺失组合）、profiles、flags（含旧 flag）
+  - **断言**：
+    - 返回顺序与输入一致
+    - `en` 不返回 `text_en` 缺失的内容（不回退 `text_tc`）
+    - risk_flags 映射后不含旧命名
+    - `review_confidence` NULL → 0.7
+  - **验收**：测试稳定通过。
+
+- [x] 6.4 集成测试用例：`fetch_candidates`
+  - **准备数据**：覆盖 `personalization_power` 0/5/10、`is_safe_pool` true/false、不同 stage
+  - **断言**：
+    - L1/L2/L3 粗过滤生效
+    - `exclude_content_ids` 生效
+    - 查询次数为常数级（用 before_cursor_execute 计数）
+  - **验收**：测试稳定通过。
+
+---
+
+## 7. 最终自检清单（合入前）
+
+- [x] 7.1 文档一致性检查
+  - `spec.md` / `plan.md` / 实现接口签名三者一致（尤其是 `locale` 与 `text` 输出规则）
+
+- [x] 7.2 性能检查（最小）
+  - `fetch_contents_by_ids` / `fetch_candidates` 查询次数断言通过（无 N+1）
+
+- [x] 7.3 回归检查
+  - 不影响现有 `user_profile_scoring` 模块与迁移脚本（仅新增模块与测试）
+

spec_kit/Personalized Reco/modules/db-design/plan.md

@@ -0,0 +1,217 @@
+# DB Design & Migrations（数据库设计与迁移）｜Plan
+
+> 对应规范：`spec_kit/Personalized Reco/modules/db-design/spec.md`
+>
+> 前置确认（已对齐）：
+>
+> - `content_id`：MySQL **自增主键**；文案微调时 **content_id 不变**（更新同一条记录）。
+> - `need_suitability/context_suitability`：**JSON** 存储。
+> - `risk_flags`：选择更强扩展性的方案（本计划采用 **关联表**，利于索引与过滤）。
+> - 安全池（L3）：**方式 A**（`is_safe_pool`）。
+> - 迁移：使用 **Alembic**，目录放 `server/alembic/`；dev/pro 两套库均可运行同一套迁移。
+> - 字符集：统一 **utf8mb4**。
+
+---
+
+## 1. 目标与交付物
+
+### 1.1 目标
+
+- 在“库为空”的前提下，落地推荐系统最小可用的数据模型。
+- 保证后续推荐查询可实现：按画像条件召回、按 ID 批量查、按风险标记过滤、支持安全池兜底。
+
+### 1.2 交付物
+
+- `server/alembic/`：Alembic 初始化目录、`alembic.ini`（或等价配置）、迁移脚本。
+- SQLAlchemy ORM 模型（建议放 `server/app/db/models/`）。
+- 初始迁移：创建 `contents`、`content_profiles`、`content_risk_flags`（以及必要索引）。
+
+---
+
+## 2. 技术决策（V1）
+
+### 2.1 表设计原则
+
+- **分离主体与画像**：`contents` 存文本与来源字段；`content_profiles` 存画像字段（便于未来画像重算/回填）。
+- **JSON 存 suitability**：`context_suitability`、`need_suitability` 用 JSON，保持结构与规则文档一致。
+- **risk_flags 关联表**：用 `content_risk_flags(content_id, flag)`，便于：
+  - 快速 Hard Filter（`block_health_medical` 等）
+  - 索引与统计（按 flag 计数）
+  - 兼容旧 flag 映射（在写入/读取层）
+- **emotion_score 的 general 表示**：用 `NULL` 表示 general（与规则文档“可为 general”语义等价）。
+- **personalization_power**：存为 `TINYINT`（0/5/10）或 `DECIMAL(2,1)`（0/0.5/1）。本计划推荐 `TINYINT`（更易索引/更省空间），应用层做映射：
+  - 0 → 0.0
+  - 5 → 0.5
+  - 10 → 1.0
+
+### 2.2 字符集与排序规则
+
+- 数据库与表：`utf8mb4`
+- collation：建议 `utf8mb4_0900_ai_ci`（MySQL 8 默认更常见；若环境不同以实际为准，但必须 utf8mb4）
+
+---
+
+## 3. 表结构（V1 方案）
+
+> 以下为“建议 schema”。实际字段名可调整，但语义必须严格对齐 `句子文案打分规则`。
+
+### 3.1 `contents`（文案主体）
+
+- `content_id` BIGINT UNSIGNED PK AUTO_INCREMENT
+- `text` TEXT NOT NULL
+- `author_id` VARCHAR(64) NULL
+- `template_id` VARCHAR(64) NULL
+- `created_at` DATETIME NOT NULL
+- `updated_at` DATETIME NOT NULL
+
+索引建议：
+
+- `idx_contents_author_id(author_id)`
+- `idx_contents_template_id(template_id)`
+
+### 3.2 `content_profiles`（内容画像）
+
+- `content_id` BIGINT UNSIGNED PK（FK → contents.content_id，ON DELETE CASCADE）
+- `stage` ENUM('general','expecting','parenting','unknown') NOT NULL DEFAULT 'general'
+- `emotion_score` DECIMAL(3,2) NULL
+  - 约定：NULL 表示 general
+- `context_suitability_json` JSON NOT NULL
+- `need_suitability_json` JSON NOT NULL
+- `personalization_power` TINYINT UNSIGNED NOT NULL DEFAULT 0
+  - 约定：只允许 0/5/10
+- `review_confidence` DECIMAL(3,2) NULL
+  - 约定：NULL 由推荐模块按 0.7 兜底（对齐规则文档）
+- `is_safe_pool` BOOLEAN NOT NULL DEFAULT FALSE
+- `updated_at` DATETIME NOT NULL
+
+索引建议：
+
+- `idx_profiles_stage(stage)`
+- `idx_profiles_personalization_power(personalization_power)`
+- `idx_profiles_is_safe_pool(is_safe_pool)`
+
+> 说明：suitability 放 JSON 后，V1 可以先不做 JSON 路径索引；当候选量上来后再加“生成列/函数索引”做加速（见 6.2）。
+
+### 3.3 `content_risk_flags`（风险标记，关联表）
+
+- `id` BIGINT UNSIGNED PK AUTO_INCREMENT
+- `content_id` BIGINT UNSIGNED NOT NULL（FK → contents.content_id，ON DELETE CASCADE）
+- `flag` VARCHAR(64) NOT NULL
+- `created_at` DATETIME NOT NULL
+
+约束与索引：
+
+- UNIQUE：`uniq_content_flag(content_id, flag)`（同一 content 不重复插同 flag）
+- 索引：`idx_flag(flag)`（用于 Hard Filter 与统计）
+- 索引：`idx_content_id(content_id)`（用于按内容批量取 flags）
+
+命名约束（应用层强制，DB 可选）：
+
+- flag 必须以 `unsafe_for_` / `block_` / `soft_` 开头（严格对齐规则文档）
+
+---
+
+## 4. Alembic 迁移落地步骤
+
+### 4.1 依赖与目录
+
+- 后端依赖：`alembic`（加入 `server/requirements.txt`，版本随项目统一管理）
+- 目录：`server/alembic/`（包含 `env.py`、`versions/`）
+- 连接串：复用现有 `DATABASE_URL`（`mysql+aiomysql://...`）
+
+### 4.2 初始化与生成迁移（一次性）
+
+- `alembic init alembic`（在 `server/` 下）
+- 配置 `env.py`：
+  - 从 `app/core/config.py` 读取 `DATABASE_URL`
+  - 引入 ORM Base 与 models，启用 autogenerate
+- 创建初始迁移：
+  - `alembic revision --autogenerate -m "init content tables"`
+  - `alembic upgrade head`
+
+### 4.3 dev/pro 一致性
+
+- 迁移脚本保持同一套；通过不同环境的 `DATABASE_URL` 指向 `mindfulness_dev` 或 `mindfulness`。
+
+---
+
+## 5. 入库流程（写入契约）
+
+### 5.1 一条文案最小入库数据
+
+必须字段（V1 最小可用）：
+
+- `text`
+- `content_profiles.stage`
+- `content_profiles.emotion_score`（可为 NULL 表示 general）
+- `content_profiles.context_suitability_json`（必须包含 5 个 key：family/work/relationship/friends/health，值为 0/0.5/1）
+- `content_profiles.need_suitability_json`（必须包含 5 个 key：emotional_support/parenting_pressure/self_worth/anxiety_relief/rest_balance，值为 0/0.5/1）
+- `content_profiles.personalization_power`（0/5/10）
+- `content_risk_flags`（可为空集合，但若存在必须按命名规范）
+
+强烈建议字段：
+
+- `author_id`、`template_id`
+- `review_confidence`
+- `is_safe_pool`（若要参与 L3 安全池）
+
+### 5.2 写入策略
+
+- 创建文案时：
+  - 先写 `contents` 得到 `content_id`（自增）
+  - 再写 `content_profiles`（同 content_id）
+  - 再批量写 `content_risk_flags`
+- 文案微调时（content_id 不变）：
+  - 更新 `contents.text` 与 `updated_at`
+  - 同步更新 `content_profiles`（若画像变更）
+  - risk_flags 做“全量覆盖”或“差量更新”（plan 实现阶段定）
+
+---
+
+## 6. 查询与性能规划
+
+### 6.1 V1 查询策略（先可用）
+
+- 候选召回：
+  - 先按 `stage`、`personalization_power`、`is_safe_pool` 等可索引字段进行粗过滤
+  - 再在应用层结合 suitability JSON 与 risk_flags 做精过滤/打分
+- Hard Filter：
+  - 通过 `content_risk_flags` join 或子查询排除指定 flags（如 `block_health_medical`）
+- 批量查：
+  - `content_id IN (...)` join `content_profiles` + left join `content_risk_flags`
+
+### 6.2 V1.1 性能增强（候选量上来后再做）
+
+当候选池变大、应用层过滤成本上升时，优先做两类增强：
+
+- **生成列/函数索引**：为常用召回维度（例如 need/context 的某些 key）创建 generated columns（从 JSON_EXTRACT 取值并映射到 TINYINT），再加索引。
+- **风险 flag 位图/派生列**：对 `block_health_medical` 等强规则增加派生布尔列（或维护冗余表），降低 join 成本。
+
+---
+
+## 7. 测试与验收（DB 子模块）
+
+### 7.1 迁移验收
+
+- 在全新库执行 `alembic upgrade head` 成功。
+- 执行 `downgrade`（若实现）可回滚（至少在开发环境可用）。
+
+### 7.2 数据契约验收
+
+插入一条最小文案记录后，能够查询并组装出推荐模块所需的 `ContentProfile` 字段集合：
+
+- `content_id/text/stage/emotion_score/context_suitability/need_suitability/personalization_power/risk_flags`
+
+### 7.3 规则口径验收（写入侧）
+
+- 写入 `risk_flags` 时，若出现旧 flag（如 `block_stage_unknown`）：
+  - 写入层需在入库前映射为新命名（或拒绝写入并提示）
+  - 推荐侧读取层不得再出现旧 flag 名称
+
+---
+
+## 8. 与其他子模块的接口约定
+
+- `Content Repository` 只依赖本模块提供的表与字段语义，不依赖具体迁移实现细节。
+- 推荐引擎/打分模块对 `review_confidence` 的缺省值假设（0.7）在 DB 缺失时依然成立。
+

spec_kit/Personalized Reco/modules/db-design/spec.md

@@ -0,0 +1,91 @@
+# 子模块：DB Design & Migrations（数据库设计与迁移）｜Spec
+
+## 1. 目标描述
+
+在当前“数据库尚未设计且为空”的前提下，为个性化推荐提供最小可用的数据存储与索引能力：
+
+- 存储文案及其内容画像（Content Profile，Cᵢ）。
+- 能按画像条件进行候选召回（need/context/stage/general、personalization_power、risk_flags 等）。
+- 能按 `content_id` 批量查询（补全候选、去重/重排时取字段）。
+- 为后续标注/审核/置信度补齐留出扩展空间。
+
+> 规则口径：字段语义必须严格对齐 `设计说明文档/句子文案打分規則.md`。
+
+---
+
+## 2. 输入 / 输出定义
+
+### 2.1 输入
+
+- 内容侧提供的文案与画像数据（可由运营导入、AI Reviewer 产出、人审修正等方式写入）。
+- 推荐模块对数据访问的需求（召回过滤字段、排序字段、频控字段）。
+
+### 2.2 输出
+
+- 一套 MySQL 表结构（或视图）满足 `ContentProfile` 字段契约：
+  - `content_id`（稳定主键）
+  - `text`
+  - `stage`（`general/expecting/parenting/unknown`）
+  - `emotion_score`（0~1 或 `general` 的等价表示）
+  - `context_suitability`（每个 context 的 {0,0.5,1}）
+  - `need_suitability`（每个 need 的 {0,0.5,1}）
+  - `personalization_power`（0/0.5/1）
+  - `risk_flags`（命名以 `unsafe_for_* / block_* / soft_*` 为准）
+  - 可选：`author_id`、`template_id`、`review_confidence`
+- Alembic 迁移脚本：`alembic revision --autogenerate` / `alembic upgrade head` 可创建上述表。
+- 推荐查询需要的索引（至少支持按场景召回与按 ID 批量查）。
+
+---
+
+## 3. 建议表结构（V1，允许后续调整）
+
+> 说明：本子模块不强制具体表名；但需保证字段语义与索引可用。以下给出一个推荐落地方案，便于后续 plan 直接实现。
+
+### 3.1 `contents`（文案主体）
+
+- `content_id`（PK）
+- `text`
+- `author_id`（可空）
+- `template_id`（可空）
+- `created_at` / `updated_at`
+
+### 3.2 `content_profiles`（内容画像）
+
+- `content_id`（PK/FK → contents）
+- `stage`（枚举：general/expecting/parenting/unknown）
+- `emotion_score`（可为 NULL 表示 general，或用额外字段 `emotion_is_general` 表示）
+- `context_suitability_json`（JSON：每个 context -> 0/0.5/1）
+- `need_suitability_json`（JSON：每个 need -> 0/0.5/1）
+- `personalization_power`（DECIMAL(2,1) 或 TINYINT 映射到 0/0.5/1）
+- `risk_flags_json`（JSON 数组：字符串集合）
+- `review_confidence`（DECIMAL(3,2)，缺省可按 0.7 处理，由推荐模块兜底）
+- `updated_at`
+
+### 3.3 “通用安全池”支持（L3 兜底）
+
+至少提供一种方式能拉到安全池内容：
+
+- 方式 A：`content_profiles` 增加 `is_safe_pool`（boolean）
+- 方式 B：单独 `safe_pool_contents`（content_id 列表）
+
+---
+
+## 4. 索引与查询能力（V1 必需）
+
+- **按 ID 批量查**：`content_id in (...)`
+- **按 stage / personalization_power 过滤**：支持候选召回与回退梯度
+- **按 risk_flags 过滤**：
+  - 推荐做法：将 `risk_flags_json` 冗余为可索引的派生列/位图/多表行（plan 阶段定实现）
+  - V1 最小可用：允许先在应用层过滤（但需控制候选量，避免全表扫）
+
+---
+
+## 5. 验收标准（可验证）
+
+- **迁移可运行**：全新 MySQL 库上执行 `alembic upgrade head` 可成功创建表结构。
+- **字段契约可满足**：能从 DB 读出 `ContentProfile` 需要的字段（至少 content_id/text/stage/emotion_score/context_suitability/need_suitability/personalization_power/risk_flags）。
+- **基本查询可用**：
+  - 能按 `content_id` 批量拉取内容画像
+  - 能按 `stage/general` 与 `personalization_power` 条件召回候选（用于 L0~L3）
+- **安全池可用**：能稳定拉取 L3 兜底候选集（不依赖用户画像）。
+

spec_kit/Personalized Reco/modules/db-design/tasks.md

@@ -0,0 +1,99 @@
+# DB Design & Migrations（数据库设计与迁移）｜Tasks
+
+> 对应计划：`spec_kit/Personalized Reco/modules/db-design/plan.md`
+>
+> 目标：在“库为空”的前提下，落地推荐系统最小可用数据模型 + Alembic 迁移，并通过最小查询/契约验收。
+
+---
+
+## 0. 任务状态约定
+
+- `[ ]`：未开始
+- `[~]`：进行中
+- `[x]`：已完成
+- `[-]`：已取消/不做（需写明原因）
+
+---
+
+## 1. 环境与依赖准备
+
+- [ ] **确认 MySQL 版本与字符集支持**
+  - 验收：MySQL 版本为 8.x；库/表可用 `utf8mb4`（建议 `utf8mb4_0900_ai_ci`）。
+- [x] **后端依赖补齐 Alembic**
+  - 说明：`server/requirements.txt` 已包含 `alembic>=1.13`
+  - 验收：在 `server/.venv` 中可成功 `import alembic`。
+
+---
+
+## 2. ORM 模型落地（推荐最小集合）
+
+- [x] **创建 ORM 模型目录**
+  - 目标路径：`server/app/db/models/`
+  - 验收：目录存在，且可被 Python 正常 import。
+- [x] **实现 `contents` 模型**
+  - 字段：`content_id(PK, 自增)`、`text_en?`、`text_tc?`、`author_id?`、`template_id?`、`created_at`、`updated_at`
+  - 索引：`author_id`、`template_id`
+  - 验收：Alembic autogenerate 能识别表结构。
+- [x] **实现 `content_profiles` 模型**
+  - 字段：`content_id(PK/FK)`、`stage(enum)`、`emotion_score(NULL=general)`、`context_suitability_json(JSON)`、`need_suitability_json(JSON)`、`personalization_power(0/5/10)`、`review_confidence?`、`is_safe_pool`、`updated_at`
+  - 索引：`stage`、`personalization_power`、`is_safe_pool`
+  - 验收：Alembic autogenerate 能识别表结构；`content_id` 具备外键与级联删除。
+- [x] **实现 `content_risk_flags` 模型（关联表）**
+  - 字段：`id(PK)`、`content_id(FK)`、`flag`、`created_at`
+  - 约束：`UNIQUE(content_id, flag)`
+  - 索引：`flag`、`content_id`
+  - 验收：Alembic autogenerate 能识别唯一约束与索引。
+
+---
+
+## 3. Alembic 初始化与迁移生成
+
+- [x] **初始化 Alembic 目录**
+  - 目标位置：`server/alembic/`（含 `versions/`、`env.py`）
+  - 验收：在 `server/` 下可运行 `alembic -h` 且能读取配置。
+- [x] **配置 Alembic 连接串来源**
+  - 要求：复用现有 `DATABASE_URL`（对齐 `server/app/core/config.py`）
+  - 验收：`alembic` 命令可加载 `env.py` 并读取 `DATABASE_URL`（未连 DB 验收留到第 4 章）。
+- [x] **配置 `env.py` 支持 autogenerate**
+  - 要求：引入 ORM `Base` 与 models（确保 metadata 完整）
+  - 验收：Alembic 能识别 `target_metadata`；且已提供初始迁移版本文件。
+- [ ] **生成并执行初始迁移**
+  - 命令：`alembic upgrade head`
+  - 验收：数据库中出现 `contents`、`content_profiles`、`content_risk_flags` 与 Alembic 版本表。
+
+---
+
+## 4. 数据契约验收（最小入库与查询）
+
+- [ ] **准备一条最小文案数据（人工插入或脚本）**
+  - 必须字段（对齐 plan）：`text`、`stage`、`emotion_score(可 NULL)`、`context_suitability_json(5 keys)`、`need_suitability_json(5 keys)`、`personalization_power(0/5/10)`、`risk_flags(可空)`
+  - 验收：可插入成功，不违反约束。
+- [ ] **验证按 `content_id` 批量查询可用**
+  - 目标：能 join 组装出 `ContentProfile` 所需字段集合（含 risk_flags 列表）
+  - 验收：至少验证字段：`content_id/text/stage/emotion_score/context_suitability/need_suitability/personalization_power/risk_flags`。
+- [ ] **验证 Hard Filter 关键 flag 可过滤**
+  - 插入：至少一条带 `block_health_medical` 的记录
+  - 验收：通过 `content_risk_flags` 可在 SQL 层排除该内容（后续供推荐候选召回使用）。
+- [ ] **验证安全池（L3）可用**
+  - 插入：至少一条 `is_safe_pool=true`
+  - 验收：可单独查询出安全池候选集（不依赖画像条件）。
+
+---
+
+## 5. 规则口径验收（risk_flags 严格对齐）
+
+- [ ] **建立 risk_flags 白名单/校验策略（写入侧或读取侧）**
+  - 要求：flag 命名必须以 `unsafe_for_` / `block_` / `soft_` 开头（对齐 `句子文案打分规则`）
+  - 验收：插入非法前缀时被拒绝或被修正（选择其一，并写清策略）。
+- [ ] **旧 flag 兼容映射验收（若存在历史数据导入）**
+  - 覆盖：`block_stage_unknown`→`unsafe_for_stage_unknown` 等（详见 plan）
+  - 验收：系统对外（读出/下游）不再出现旧 flag 名称。
+
+---
+
+## 6. 文档与交接
+
+- [ ] **补齐本子模块 README/说明（可选，但建议）**
+  - 内容：如何初始化 DB、如何跑迁移、如何插入一条最小文案数据、如何验证查询。
+  - 验收：新同学按文档能在 30 分钟内跑通迁移 + 插入 + 查询。
+