merge: 解决第56波与develop的add/add冲突（sticker+subtitle）

test(P3-1): 第56波 worker层更多引擎纯逻辑单测（+108）
- test_subtitle_render_engine.py: 53个（颜色转换/时间格式化/文字换行/ASS转义/样式） - test_sticker_engine.py: 27个（配置类/位置解析/常量/便捷函数） - test_tts_engine.py: 12个（数据类/入口判断/初始化） - test_render_audio_utils.py: 16个（clip_effective_duration/RenderContext）覆盖worker层4个模块的纯逻辑部分
2026-07-25 12:29:34 +08:00 · 2026-07-25 09:37:08 +08:00
1015 changed files with 36257 additions and 132611 deletions
@@ -1,161 +0,0 @@
 name: ACR Cleanup
 on:
  schedule:
    - cron: '0 19 * * *'  # UTC 19:00 = 北京时间凌晨3:00
  workflow_dispatch:
    inputs:
      pr_sha:
        description: "PR commit SHA（仅清理指定PR镜像，留空则全量清理）"
        required: false
        default: ""
      dry_run:
        description: "预览模式（dry-run），不实际删除"
        required: false
        default: "true"
  pull_request_target:
    types: [closed]
    branches: [develop, main]
 concurrency:
  group: acr-cleanup-${{ gitea.ref }}
  cancel-in-progress: false
 jobs:
  cleanup:
    name: ACR Image Cleanup
    runs-on: ci-l2
    timeout-minutes: 20
    permissions:
      contents: read
    env:
      ACR_REGISTRY: xiaoxia-registry.cn-hangzhou.cr.aliyuncs.com
      ACR_NAMESPACE: xiaoxiakeji
      ACR_SERVICE: registry.aliyuncs.com:cn-hangzhou:china:cri-fvec8o9q4mmxrkaa
      GITEA_URL: https://git.xiaoxiajianji.com
      GITEA_REPO: xiaoxia/xiaoxia-saas
    steps:
      - name: Checkout code
        shell: sh
        env:
          GITHUB_TOKEN: ${{ github.token }}
        run: |
          curl -sH "Authorization: token $GITHUB_TOKEN"             "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"             | bash
      # ====== Cron模式：获取staging运行中镜像作为白名单 ======
      - name: Get staging running images (whitelist)
        id: protected_images
        if: gitea.event_name != 'pull_request_target' && !gitea.event.inputs.pr_sha
        env:
          STAGING_SSH_KEY: ${{ secrets.PREVIEW_SSH_KEY }}
        run: |
          set +e
          echo "获取staging服务器运行中镜像作为白名单..."
          mkdir -p ~/.ssh
          echo "$STAGING_SSH_KEY" > ~/.ssh/id_rsa
          chmod 600 ~/.ssh/id_rsa
          staging_host="${STAGING_SSH_HOST:-47.98.113.167}"
          staging_port="${STAGING_SSH_PORT:-22222}"
          ssh-keyscan -p "$staging_port" -H "$staging_host" >> ~/.ssh/known_hosts 2>/dev/null
          # 获取所有运行容器的镜像，提取tag部分
          IMAGES=$(ssh -p "$staging_port" -i ~/.ssh/id_rsa -o StrictHostKeyChecking=no \
            "root@$staging_host" "docker ps --format '{{.Image}}' 2>/dev/null" 2>/dev/null | grep -v "^$" | sort -u)
          PROTECTED_TAGS=""
          if [ -n "$IMAGES" ]; then
            while IFS= read -r img; do
              # 从完整镜像名中提取tag（最后一个冒号后）
              tag=$(echo "$img" | rev | cut -d: -f1 | rev)
              if [ -n "$tag" ] && [ "$tag" != "latest" ] && [ ${#tag} -gt 5 ]; then
                if [ -z "$PROTECTED_TAGS" ]; then
                  PROTECTED_TAGS="$tag"
                else
                  PROTECTED_TAGS="$PROTECTED_TAGS,$tag"
                fi
              fi
            done <<< "$IMAGES"
          fi
          echo "staging运行中镜像tag: ${PROTECTED_TAGS:-（无）}"
          echo "protected_tags=$PROTECTED_TAGS" >> $GITEA_OUTPUT
      # ====== Docker登录 ======
      - name: Docker login to ACR
        env:
          ACR_USERNAME: ${{ secrets.ACR_USERNAME }}
          ACR_PASSWORD: ${{ secrets.ACR_PASSWORD }}
        run: |
          printf '%s' "$ACR_PASSWORD" | docker login "$ACR_REGISTRY" -u "$ACR_USERNAME" --password-stdin
      # ====== 模式1：PR关闭时清理 ======
      - name: Cleanup PR images (PR closed)
        if: gitea.event_name == 'pull_request_target'
        env:
          ACR_USERNAME: ${{ secrets.ACR_USERNAME }}
          ACR_PASSWORD: ${{ secrets.ACR_PASSWORD }}
          GITEA_TOKEN: ${{ secrets.GITEA_TOKEN }}
          PR_SHA: ${{ gitea.event.pull_request.head.sha }}
          PR_NUMBER: ${{ gitea.event.pull_request.number }}
        run: |
          echo "============================================"
          echo "  PR #$PR_NUMBER 已关闭，清理对应镜像"
          echo "  Head SHA: ${PR_SHA::12}"
          echo "============================================"
          echo ""
          python3 scripts/ci/acr_cleanup.py \
            --pr-sha "$PR_SHA" \
            --execute
      # ====== 模式2：Cron全量清理 ======
      - name: Full cleanup (cron / manual)
        if: gitea.event_name != 'pull_request_target' && !gitea.event.inputs.pr_sha
        env:
          ACR_USERNAME: ${{ secrets.ACR_USERNAME }}
          ACR_PASSWORD: ${{ secrets.ACR_PASSWORD }}
          GITEA_TOKEN: ${{ secrets.GITEA_TOKEN }}
          PROTECTED_TAGS: ${{ steps.protected_images.outputs.protected_tags }}
          DRY_RUN_INPUT: ${{ gitea.event.inputs.dry_run }}
        run: |
          echo "============================================"
          echo "  ACR 全量清理（${{ gitea.event_name }}）"
          echo "============================================"
          echo ""
          # 决定是否dry-run
          DRY_RUN_FLAG=""
          if [ "$DRY_RUN_INPUT" = "true" ]; then
            DRY_RUN_FLAG="--dry-run"
            echo "模式: 预览模式 (dry-run)"
          else
            echo "模式: 执行模式"
          fi
          echo ""
          python3 scripts/ci/acr_cleanup.py \
            --keep 20 \
            --protected-tags "$PROTECTED_TAGS" \
            $DRY_RUN_FLAG
      # ====== 模式3：手动指定PR SHA清理 ======
      - name: Cleanup specific PR image (manual)
        if: gitea.event_name == 'workflow_dispatch' && gitea.event.inputs.pr_sha
        env:
          ACR_USERNAME: ${{ secrets.ACR_USERNAME }}
          ACR_PASSWORD: ${{ secrets.ACR_PASSWORD }}
          PR_SHA: ${{ gitea.event.inputs.pr_sha }}
          DRY_RUN_INPUT: ${{ gitea.event.inputs.dry_run }}
        run: |
          echo "手动清理PR镜像: ${PR_SHA::12}"
          echo ""
          DRY_RUN_FLAG=""
          if [ "$DRY_RUN_INPUT" = "true" ]; then
            DRY_RUN_FLAG="--dry-run"
          fi
          python3 scripts/ci/acr_cleanup.py \
            --pr-sha "$PR_SHA" \
            $DRY_RUN_FLAG
@@ -8,7 +8,7 @@ permissions:
 jobs:
  ci-health-report:
    name: CI健康度每日巡检
-    runs-on: ci-l2
+    runs-on: saas
    timeout-minutes: 15
    steps:
    - name: Checkout code
@@ -21,8 +21,20 @@ on:
 permissions:
  contents: read
 concurrency:
-  group: ci-pipeline-${{ gitea.ref }}
+  group: ci-pipeline-${{ gitea.event_name }}-${{ gitea.ref }}
-  cancel-in-progress: true
+  # PR事件取消进行中的旧run，push事件不取消（确保完整CI跑完）
  cancel-in-progress: ${{ gitea.event_name == 'pull_request' }}
 env:
  CI_PG_HOST: host.docker.internal
  CI_LOCAL_PG_PORT: "5432"
  CI_PG_USER: postgres
  CI_PG_PASSWORD: postgres
  CI_PG_DB: xiaoxia_saas
  CI_SHARED_PG_PORT: "5433"
  CI_SHARED_PG_USER: postgres
  CI_SHARED_PG_PASSWORD: ci_pg_2026!
  CI_DEFAULT_DB: xiaoxia_saas
 jobs:
  check-frontend-only:
    name: Check if frontend-only change
@@ -37,7 +49,7 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Check changed files
      id: check
      shell: bash
@@ -90,7 +102,7 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -180,7 +192,7 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -250,7 +262,7 @@ jobs:
    permissions:
      contents: read
    env:
-      DATABASE_URL: postgresql+psycopg://postgres:postgres@host.docker.internal:5432/xiaoxia_saas
+      DATABASE_URL: postgresql+psycopg://${{ env.CI_PG_USER }}:${{ env.CI_PG_PASSWORD }}@${{ env.CI_PG_HOST }}:${{ env.CI_LOCAL_PG_PORT }}/${{ env.CI_PG_DB }}
      USE_IN_MEMORY_DB: 'false'
      CI_USE_SHARED_PG: 'true'
    steps:
@@ -259,7 +271,7 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -334,13 +346,14 @@ jobs:
      OSS_ACCESS_KEY_SECRET: placeholder
      OSS_BUCKET_NAME: xiaoxia-autocut
      OSS_ENDPOINT: oss-cn-hangzhou.aliyuncs.com
      JWT_SECRET_KEY: test-jwt-secret-for-ci-only-2026
    steps:
    - name: Checkout code
      shell: sh
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -398,20 +411,21 @@ jobs:
      - validate-type-check
      - validate-migration
    env:
-      DATABASE_URL: postgresql+psycopg://postgres:postgres@host.docker.internal:5432/xiaoxia_saas
+      DATABASE_URL: postgresql+psycopg://${{ env.CI_PG_USER }}:${{ env.CI_PG_PASSWORD }}@${{ env.CI_PG_HOST }}:${{ env.CI_LOCAL_PG_PORT }}/${{ env.CI_PG_DB }}
      USE_IN_MEMORY_DB: 'false'
      CI_USE_SHARED_PG: 'true'
      OSS_ACCESS_KEY_ID: placeholder
      OSS_ACCESS_KEY_SECRET: placeholder
      OSS_BUCKET_NAME: xiaoxia-autocut
      OSS_ENDPOINT: oss-cn-hangzhou.aliyuncs.com
      JWT_SECRET_KEY: test-jwt-secret-for-ci-only-2026
    steps:
    - name: Checkout code
      shell: sh
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -457,15 +471,13 @@ jobs:
    name: Frontend Lint
    runs-on: ci-l2
    timeout-minutes: 10
    needs: check-frontend-only
    if: needs.check-frontend-only.outputs.skip_frontend != 'true'
    steps:
    - name: Checkout code
      shell: sh
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -527,7 +539,7 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -577,12 +589,7 @@ jobs:
    name: PR Build ${{ matrix.service_display }} Image
    runs-on: runtime-builder
    timeout-minutes: ${{ matrix.timeout }}
-    needs: check-frontend-only
+    if: github.event_name == 'pull_request'
    if: |
      github.event_name == 'pull_request' && (
        (matrix.service == 'web' && needs.check-frontend-only.outputs.skip_frontend != 'true') ||
        (matrix.service != 'web' && needs.check-frontend-only.outputs.skip_backend != 'true')
      )
    strategy:
      fail-fast: false
      matrix:
@@ -611,7 +618,7 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -633,60 +640,83 @@ jobs:
          echo "Docker login failed ($i/3), retrying in 5s..."
          sleep 5
        done
-    - name: Pre-build worker base images (fallback if not exist)
+    - name: Pre-build worker base images (3-level cache)
      if: matrix.service == 'worker'
      id: prebuild
-      shell: sh
+      shell: bash
      run: |
        set -eu
-        REGISTRY="git.xiaoxiajianji.com/xiaoxia-saas"
+        GITEA_REGISTRY="git.xiaoxiajianji.com/xiaoxia-saas"
-        BASE_BUILDER="${REGISTRY}/worker-base-builder:latest"
+        ACR_REGISTRY="xiaoxia-registry.cn-hangzhou.cr.aliyuncs.com/xiaoxiakeji"
-        BASE_RUNTIME="${REGISTRY}/worker-base-runtime:latest"
+        GITEA_BUILDER="${GITEA_REGISTRY}/worker-base-builder:latest"
-        
+        GITEA_RUNTIME="${GITEA_REGISTRY}/worker-base-runtime:latest"
-        # 尝试拉取基础镜像
+        ACR_BUILDER="${ACR_REGISTRY}/worker-base-builder:latest"
-        echo "检查基础镜像..."
+        ACR_RUNTIME="${ACR_REGISTRY}/worker-base-runtime:latest"
        if docker pull "$BASE_BUILDER" 2>/dev/null && docker pull "$BASE_RUNTIME" 2>/dev/null; then
          echo "基础镜像已存在，使用远程镜像"
          echo "fallback=false" >> $GITHUB_OUTPUT
        else
          echo "基础镜像不存在，本地构建（fallback模式）..."
          # 尝试用buildx构建，失败则回退到普通docker build（DooD模式下buildx builder偶发崩溃）
          BUILDER_NAME="ci-pr-builder-${GITHUB_RUN_ID:-local}"
          BUILDX_AVAILABLE=true
          if ! docker buildx create --use --name "$BUILDER_NAME" --driver docker-container > /dev/null 2>&1; then
            BUILDX_AVAILABLE=false
          fi
          if [ "$BUILDX_AVAILABLE" = true ] && ! docker buildx inspect --bootstrap > /dev/null 2>&1; then
            BUILDX_AVAILABLE=false
            docker buildx rm "$BUILDER_NAME" > /dev/null 2>&1 || true
          fi
          build_base() {
            local df="$1"
            local tag="$2"
            local name="$3"
            if [ "$BUILDX_AVAILABLE" = true ]; then
              echo "构建 $name（buildx）..."
              if docker buildx build --load -f "$df" -t "$tag" . > /dev/null 2>&1; then
                echo "$name 构建成功"
                return 0
              fi
              echo "buildx失败，回退到普通docker build"
              BUILDX_AVAILABLE=false
              docker buildx rm "$BUILDER_NAME" > /dev/null 2>&1 || true
            fi
            echo "构建 $name（docker build）..."
            docker build -f "$df" -t "$tag" .
          }
          build_base infra/docker/worker-base-builder.Dockerfile "$BASE_BUILDER" "worker-base-builder"
          build_base infra/docker/worker-base-runtime.Dockerfile "$BASE_RUNTIME" "worker-base-runtime"
          echo "fallback=true" >> $GITHUB_OUTPUT
          echo "基础镜像本地构建完成"
        fi
        # L1: 本地daemon缓存（DooD模式8runner共享宿主机daemon）
        echo "=== L1 本地缓存 ==="
        if docker image inspect "$ACR_BUILDER" > /dev/null 2>&1 \
           && docker image inspect "$ACR_RUNTIME" > /dev/null 2>&1; then
          echo "本地缓存命中"
          echo "has_local_base=true" >> $GITHUB_OUTPUT
          exit 0
        fi
        echo "本地无缓存"
        # L2: Gitea registry缓存（内网快）
        echo "=== L2 Registry拉取 ==="
        if docker pull "$GITEA_BUILDER" 2>/dev/null && docker pull "$GITEA_RUNTIME" 2>/dev/null; then
          echo "Registry拉取成功，重tag供Dockerfile使用"
          docker tag "$GITEA_BUILDER" "$ACR_BUILDER"
          docker tag "$GITEA_RUNTIME" "$ACR_RUNTIME"
          echo "has_local_base=true" >> $GITHUB_OUTPUT
          exit 0
        fi
        echo "Registry无缓存，需本地构建"
        # L3: 本地构建
        echo "=== L3 本地构建 ==="
        BUILDER_NAME="ci-pr-builder-${GITHUB_RUN_ID:-local}"
        if ! docker buildx inspect "$BUILDER_NAME" > /dev/null 2>&1; then
          docker buildx create --use --name "$BUILDER_NAME" --driver docker-container
        else
          docker buildx use "$BUILDER_NAME"
        fi
        docker buildx inspect --bootstrap > /dev/null 2>&1
        echo "构建 worker-base-builder..."
        for attempt in 1 2 3; do
          if docker buildx build --load -f infra/docker/worker-base-builder.Dockerfile -t "$ACR_BUILDER" .; then
            echo "worker-base-builder 构建成功"
            break
          fi
          echo "worker-base-builder 失败，重试 $attempt/3..."
          docker buildx rm "$BUILDER_NAME" 2>/dev/null || true
          docker buildx create --use --name "$BUILDER_NAME" --driver docker-container
          sleep 3
        done
        echo "构建 worker-base-runtime..."
        for attempt in 1 2 3; do
          if docker buildx build --load -f infra/docker/worker-base-runtime.Dockerfile -t "$ACR_RUNTIME" .; then
            echo "worker-base-runtime 构建成功"
            break
          fi
          echo "worker-base-runtime 失败，重试 $attempt/3..."
          docker buildx rm "$BUILDER_NAME" 2>/dev/null || true
          docker buildx create --use --name "$BUILDER_NAME" --driver docker-container
          sleep 3
        done
        # 推送到Gitea registry供后续复用
        echo "=== 推送缓存到Registry ==="
        docker tag "$ACR_BUILDER" "$GITEA_BUILDER"
        docker tag "$ACR_RUNTIME" "$GITEA_RUNTIME"
        docker push "$GITEA_BUILDER" 2>/dev/null || echo "push builder失败（不影响）"
        docker push "$GITEA_RUNTIME" 2>/dev/null || echo "push runtime失败（不影响）"
        echo "has_local_base=true" >> $GITHUB_OUTPUT
        echo "基础镜像构建完成"
    - name: Build PR image (verify only, no push)
      shell: sh
      run: |
@@ -700,15 +730,15 @@ jobs:
          EXTRA_BUILD_ARGS="$EXTRA_BUILD_ARGS NGINX_CONF=infra/docker/nginx-staging.conf"
        fi
-        # Worker fallback模式：基础镜像本地已构建，用普通docker build绕过buildx
+        # Worker有本地base镜像时：用BuildKit直接构建（快，无需起buildx容器）
-        if [ "${{ matrix.service }}" = "worker" ] && [ "${{ steps.prebuild.outputs.fallback }}" = "true" ]; then
+        if [ "${{ matrix.service }}" = "worker" ] && [ "${{ steps.prebuild.outputs.has_local_base }}" = "true" ]; then
-          echo "Fallback模式：用普通docker build（基础镜像本地已构建）"
+          echo "本地base镜像已就绪，BuildKit快速构建"
          BUILD_ARG_STR=""
          for arg in $EXTRA_BUILD_ARGS; do
            BUILD_ARG_STR="$BUILD_ARG_STR --build-arg $arg"
          done
-          docker build -f ${{ matrix.dockerfile }} -t "${IMAGE_TAG}" $BUILD_ARG_STR .
+          DOCKER_BUILDKIT=1 docker build -f ${{ matrix.dockerfile }} -t "${IMAGE_TAG}" $BUILD_ARG_STR .
-          echo "Fallback PR Build successful"
+          echo "快速构建成功"
          exit 0
        fi
@@ -795,7 +825,7 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -927,7 +957,7 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -1079,14 +1109,32 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
    - name: Run Playwright E2E on staging
      shell: bash
      run: |
-        bash scripts/ci/run_staging_tests.sh e2e
+        set -eu
        # DooD模式下不能用-v挂载（宿主机路径与CI容器路径不一致）
        # 改用 docker create + docker cp 方式把代码拷进容器
        CONTAINER_NAME="staging-e2e-${GITHUB_SHA::8}"
        docker rm -f "$CONTAINER_NAME" 2>/dev/null || true
        docker create --name "$CONTAINER_NAME" --ipc=host \
          -e E2E_BASE_URL=https://staging.xiaoxiajianji.com \
          -e E2E_API_BASE=https://staging-api.xiaoxiajianji.com/api/v1 \
          -e E2E_BROWSER_CHANNEL=chromium \
          -e PLAYWRIGHT_HEADLESS=1 \
          -w /workspace/apps/web \
          git.xiaoxiajianji.com/xiaoxia/base/playwright:v1.45.0-jammy \
          sh -lc "npm ci && npx playwright test --reporter=line --project=chromium e2e/auth.spec.ts e2e/auth-guard.spec.ts e2e/core-upload.spec.ts e2e/core-generation.spec.ts"
        docker cp apps "$CONTAINER_NAME:/workspace/"
        docker cp package-lock.json "$CONTAINER_NAME:/workspace/" 2>/dev/null || true
        docker start -a "$CONTAINER_NAME"
        EXIT_CODE=$(docker wait "$CONTAINER_NAME")
        docker rm "$CONTAINER_NAME" 2>/dev/null || true
        exit $EXIT_CODE
    - name: Job duration summary
      if: always()
@@ -1126,14 +1174,30 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
    - name: Run API integration tests on staging
      shell: bash
      run: |
-        bash scripts/ci/run_staging_tests.sh api
+        set -eu
        # DooD模式下不能用-v挂载（宿主机路径与CI容器路径不一致）
        # 改用 docker create + docker cp 方式把代码拷进容器
        CONTAINER_NAME="staging-api-tests-${GITHUB_SHA::8}"
        docker rm -f "$CONTAINER_NAME" 2>/dev/null || true
        docker create --name "$CONTAINER_NAME" \
          -e E2E_BASE_URL=https://staging.xiaoxiajianji.com \
          -e E2E_API_BASE=https://staging-api.xiaoxiajianji.com/api/v1 \
          -w /workspace/apps/web \
          git.xiaoxiajianji.com/xiaoxia/base/playwright:v1.45.0-jammy \
          sh -lc 'npm ci && npx playwright test --reporter=line e2e/test_auth.spec.ts e2e/test_asset.spec.ts e2e/test_project.spec.ts'
        docker cp apps "$CONTAINER_NAME:/workspace/"
        docker cp package-lock.json "$CONTAINER_NAME:/workspace/" 2>/dev/null || true
        docker start -a "$CONTAINER_NAME"
        EXIT_CODE=$(docker wait "$CONTAINER_NAME")
        docker rm "$CONTAINER_NAME" 2>/dev/null || true
        exit $EXIT_CODE
    - name: Job duration summary
      if: always()
@@ -1166,12 +1230,7 @@ jobs:
    runs-on: runtime-builder
    timeout-minutes: ${{ matrix.timeout }}
    needs:
-      - validate-code-quality
+    if: startsWith(github.ref, 'refs/tags/v')
      - validate-type-check
      - unit-tests
      - frontend-lint
      - frontend-unit-test
    if: startsWith(github.ref, 'refs/tags/v') || (github.event_name == 'push' && github.ref_name == 'main')
    strategy:
      fail-fast: false
      matrix:
@@ -1200,7 +1259,7 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -1244,20 +1303,14 @@ jobs:
        docker buildx inspect --bootstrap
    - name: Build and push production ${{ matrix.service_display }} image (with retry)
-      shell: bash
+      shell: sh
      run: |
        set -eu
        REGISTRY="xiaoxia-registry.cn-hangzhou.cr.aliyuncs.com/xiaoxiakeji"
-        # 根据ref类型设置镜像标签：tag用版本号，分支用分支名+sha
+        IMAGE_TAG="${REGISTRY}/${{ matrix.image_name }}:${GITHUB_REF_NAME}"
        if [[ "$GITHUB_REF" == refs/tags/* ]]; then
          TAG_NAME="${GITHUB_REF_NAME}"
        else
          TAG_NAME="${GITHUB_REF_NAME}-${GITHUB_SHA::8}"
        fi
        IMAGE_TAG="${REGISTRY}/${{ matrix.image_name }}:${TAG_NAME}"
        CACHE_REF="${REGISTRY}/${{ matrix.cache_name }}:main"
-        EXTRA_BUILD_ARGS="APP_VERSION=\"${TAG_NAME}\""
+        EXTRA_BUILD_ARGS="APP_VERSION=\"${GITHUB_REF_NAME}\""
        if [ "${{ matrix.service }}" = "web" ]; then
          EXTRA_BUILD_ARGS="$EXTRA_BUILD_ARGS NGINX_CONF=infra/docker/nginx-production.conf"
        fi
@@ -1324,7 +1377,7 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -1455,13 +1508,14 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
    - name: Run production browser E2E
-      shell: bash
+      shell: sh
      run: |
        set -eu
        docker run --rm --ipc=host \
          -e E2E_BASE_URL=https://saas.xiaoxiajianji.com \
          -e E2E_API_BASE=https://api.xiaoxiajianji.com/api/v1 \
@@ -1470,7 +1524,7 @@ jobs:
          -v "$PWD:/workspace" \
          -w /workspace/apps/web \
          git.xiaoxiajianji.com/xiaoxia/base/playwright:v1.45.0-jammy \
-          bash -c 'for i in 1 2 3; do npm ci --registry=https://registry.npmmirror.com && break; echo "npm ci attempt $i failed, retrying..."; sleep 15; done && npx playwright test --reporter=line --project=chromium e2e/auth.spec.ts e2e/auth-guard.spec.ts e2e/core-upload.spec.ts e2e/core-generation.spec.ts e2e/core-titles.spec.ts'
+          sh -lc 'npm ci && npx playwright test --reporter=line --project=chromium e2e/auth.spec.ts e2e/auth-guard.spec.ts e2e/core-upload.spec.ts e2e/core-generation.spec.ts e2e/core-titles.spec.ts'
    - name: Job duration summary
      if: always()
@@ -1514,7 +1568,7 @@ jobs:
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
-        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
+        curl -sH "Authorization: token $GITHUB_TOKEN"           "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"           | bash
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
@@ -1554,276 +1608,4 @@ jobs:
        [ ${{ job.status }} = "success" ] || STATUS="error"
        START_TIME=""
        [ -f /tmp/ci_job_start_time ] && START_TIME=$(cat /tmp/ci_job_start_time)
-        python3 scripts/ci/ci_trace_report.py --service xiaoxia-saas-ci --status $STATUS --start-time "$START_TIME" || true
+        python3 scripts/ci/ci_trace_report.py --service xiaoxia-saas-ci --status $STATUS --start-time "$START_TIME" || true
  canary-release:
    name: Canary Release to Production
    runs-on: runtime-builder
    timeout-minutes: 120
    concurrency:
      group: canary-release-production
      cancel-in-progress: false
    if: github.event_name == 'push' && github.ref_name == 'main'
    needs:
    - build-production
    - staging-api-tests
    steps:
    - name: Checkout code
      shell: sh
      env:
        GITHUB_TOKEN: ${{ github.token }}
      run: |
        curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
    - name: Record job start time
      shell: sh
      run: bash scripts/ci/step_timer_start.sh
    - name: Notify canary release start
      continue-on-error: true
      shell: sh
      env:
        CI_NOTIFY_WEBHOOK: ${{ secrets.CI_NOTIFY_WEBHOOK }}
      run: |
        set +e
        NOTIFY_MODE=start JOB_NAME="Canary Release" python3 scripts/ci_notify.py
    - name: Install SSH client
      shell: sh
      run: |
        set -eu
        apt-get update -qq && apt-get install -y -qq openssh-client curl >/dev/null 2>&1
        echo "openssh-client installed"
    - name: Run canary release
      shell: bash
      env:
        PRODUCTION_SSH_HOST: ${{ secrets.PRODUCTION_SSH_HOST }}
        PRODUCTION_SSH_USER: ${{ secrets.PRODUCTION_SSH_USER }}
        PRODUCTION_SSH_PORT: ${{ secrets.PRODUCTION_SSH_PORT }}
        PRODUCTION_SSH_KEY: ${{ secrets.PRODUCTION_SSH_KEY }}
        ACR_USERNAME: ${{ secrets.ACR_USERNAME }}
        ACR_PASSWORD: ${{ secrets.ACR_PASSWORD }}
        CI_NOTIFY_WEBHOOK: ${{ secrets.CI_NOTIFY_WEBHOOK }}
      run: |
        set -eu
        IMAGE_TAG="main-${GITHUB_SHA::8}"
        export IMAGE_TAG
        echo "Canary release version: $IMAGE_TAG"
        bash scripts/ci/canary_release.sh
    - name: Job duration summary
      if: always()
      shell: sh
      run: bash scripts/ci/step_timer_end.sh
    - name: Notify on success
      continue-on-error: true
      if: success()
      shell: sh
      env:
        CI_NOTIFY_WEBHOOK: ${{ secrets.CI_NOTIFY_WEBHOOK }}
      run: |
        set +e
        NOTIFY_MODE=success JOB_NAME="Canary Release" python3 scripts/ci_notify.py
    - name: Notify on failure
      continue-on-error: true
      if: failure()
      shell: sh
      env:
        CI_NOTIFY_WEBHOOK: ${{ secrets.CI_NOTIFY_WEBHOOK }}
      run: |
        set +e
        NOTIFY_MODE=failure JOB_NAME="Canary Release" python3 scripts/ci_notify.py
    - name: Report CI trace
      if: always()
      shell: sh
      env:
        AGENTLOOP_LICENSE_KEY: ${{ secrets.AGENTLOOP_LICENSE_KEY }}
      run: |
        STATUS="ok"
        [ ${{ job.status }} = "success" ] || STATUS="error"
        START_TIME=""
        [ -f /tmp/ci_job_start_time ] && START_TIME=$(cat /tmp/ci_job_start_time)
        python3 scripts/ci/ci_trace_report.py --service xiaoxia-saas-ci --status $STATUS --start-time "$START_TIME" || true
  ci-gate:
    name: CI Gate
    runs-on: ci-l2
    if: always() && github.event_name == 'pull_request'
    needs:
      - check-frontend-only
      - validate-code-quality
      - validate-type-check
      - validate-migration
      - unit-tests
      - integration-tests
      - frontend-lint
      - frontend-unit-test
      - build-pr
    timeout-minutes: 3
    steps:
      - name: Checkout code
        shell: sh
        env:
          GITHUB_TOKEN: ${{ github.token }}
        run: |
          curl -sfH "Authorization: token $GITHUB_TOKEN" -o /tmp/_ci_checkout.sh \
            "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" && bash /tmp/_ci_checkout.sh
      - name: Evaluate CI Gate
        id: gate
        shell: bash
        env:
          GITHUB_TOKEN: ${{ github.token }}
          PR_NUMBER: ${{ github.event.pull_request.number }}
          PR_HEAD_SHA: ${{ github.event.pull_request.head.sha }}
          RESULT_CHECK_FRONTEND: ${{ needs.check-frontend-only.result }}
          RESULT_CODE_QUALITY: ${{ needs.validate-code-quality.result }}
          RESULT_TYPE_CHECK: ${{ needs.validate-type-check.result }}
          RESULT_MIGRATION: ${{ needs.validate-migration.result }}
          RESULT_UNIT_TESTS: ${{ needs.unit-tests.result }}
          RESULT_INTEGRATION: ${{ needs.integration-tests.result }}
          RESULT_FRONTEND_LINT: ${{ needs.frontend-lint.result }}
          RESULT_FRONTEND_UNIT: ${{ needs.frontend-unit-test.result }}
          RESULT_BUILD_PR: ${{ needs.build-pr.result }}
        run: |
          set -eu
          echo "=== CI Gate 评估 ==="
          echo ""
          echo "各job结果:"
          echo "  check-frontend-only: $RESULT_CHECK_FRONTEND"
          echo "  validate-code-quality: $RESULT_CODE_QUALITY"
          echo "  validate-type-check: $RESULT_TYPE_CHECK"
          echo "  validate-migration: $RESULT_MIGRATION"
          echo "  unit-tests: $RESULT_UNIT_TESTS"
          echo "  integration-tests: $RESULT_INTEGRATION"
          echo "  frontend-lint: $RESULT_FRONTEND_LINT"
          echo "  frontend-unit-test: $RESULT_FRONTEND_UNIT"
          echo "  build-pr: $RESULT_BUILD_PR"
          # 查询 AI Code Review 状态（跨workflow，读commit status）
          AI_REVIEW_STATUS="pending"
          AI_REVIEW_DESC=""
          STATUS_JSON=$(curl -sfH "Authorization: token $GITHUB_TOKEN" \
            "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/commits/${PR_HEAD_SHA}/status" 2>/dev/null || true)
          if [ -n "$STATUS_JSON" ]; then
            AI_STATUS=$(echo "$STATUS_JSON" | python3 -c "
          import json,sys
          try:
              data=json.load(sys.stdin)
              for s in data.get('statuses',[]):
                  if 'AI Code Review' in s.get('context',''):
                      print(s['state']+'|'+s.get('description',''))
                      break
          except: pass
          " 2>/dev/null)
            if [ -n "$AI_STATUS" ]; then
              AI_REVIEW_STATUS="${AI_STATUS%%|*}"
              AI_REVIEW_DESC="${AI_STATUS#*|}"
            fi
          fi
          echo "  ai-code-review: $AI_REVIEW_STATUS ($AI_REVIEW_DESC)"
          echo ""
          # 判断PR类型
          SKIP_BACKEND="${{ needs.check-frontend-only.outputs.skip_backend }}"
          SKIP_FRONTEND="${{ needs.check-frontend-only.outputs.skip_frontend }}"
          echo "PR类型: skip_backend=$SKIP_BACKEND, skip_frontend=$SKIP_FRONTEND"
          # 必填检查项（根据PR类型决定）
          # 通用检查（所有PR都必须过）
          REQUIRED_GENERAL=(
            "validate-code-quality:$RESULT_CODE_QUALITY"
            "validate-type-check:$RESULT_TYPE_CHECK"
            "validate-migration:$RESULT_MIGRATION"
            "frontend-lint:$RESULT_FRONTEND_LINT"
            "build-pr:$RESULT_BUILD_PR"
            "ai-code-review:$AI_REVIEW_STATUS"
          )
          # 后端检查
          REQUIRED_BACKEND=(
            "unit-tests:$RESULT_UNIT_TESTS"
            "integration-tests:$RESULT_INTEGRATION"
          )
          # 前端检查
          REQUIRED_FRONTEND=(
            "frontend-unit-test:$RESULT_FRONTEND_UNIT"
          )
          ALL_PASSED=true
          FAILED_ITEMS=()
          check_job() {
            local name=$1
            local result=$2
            if [ "$result" = "success" ]; then
              echo "  ✅ $name: success"
            elif [ "$result" = "skipped" ]; then
              echo "  ⏭️  $name: skipped（跳过，不影响）"
            else
              echo "  ❌ $name: $result"
              ALL_PASSED=false
              FAILED_ITEMS+=("$name=$result")
            fi
          }
          echo ""
          echo "=== 通用检查（所有PR必填）==="
          for item in "${REQUIRED_GENERAL[@]}"; do
            name="${item%%:*}"
            result="${item##*:}"
            # AI Code Review pending时不阻塞（可能还在跑），等它跑完自然会重跑Gate
            if [ "$name" = "ai-code-review" ] && [ "$result" = "pending" ]; then
              echo "  ⏳ $name: pending（审查中，暂不阻塞）"
              continue
            fi
            check_job "$name" "$result"
          done
          if [ "$SKIP_BACKEND" != "true" ]; then
            echo ""
            echo "=== 后端检查 ==="
            for item in "${REQUIRED_BACKEND[@]}"; do
              name="${item%%:*}"
              result="${item##*:}"
              check_job "$name" "$result"
            done
          else
            echo ""
            echo "=== 后端检查（纯前端PR，跳过）==="
          fi
          if [ "$SKIP_FRONTEND" != "true" ]; then
            echo ""
            echo "=== 前端检查 ==="
            for item in "${REQUIRED_FRONTEND[@]}"; do
              name="${item%%:*}"
              result="${item##*:}"
              check_job "$name" "$result"
            done
          else
            echo ""
            echo "=== 前端检查（纯后端PR，跳过）==="
          fi
          echo ""
          if [ "$ALL_PASSED" = "true" ]; then
            echo "✅ CI Gate: PASSED"
            echo "gate_result=success" >> $GITHUB_OUTPUT
            exit 0
          else
            echo "❌ CI Gate: FAILED"
            echo "失败项: ${FAILED_ITEMS[*]}"
            echo "gate_result=failure" >> $GITHUB_OUTPUT
            exit 1
          fi
      - name: Report CI trace
        if: always()
        shell: sh
        env:
          AGENTLOOP_LICENSE_KEY: ${{ secrets.AGENTLOOP_LICENSE_KEY }}
        run: |
          STATUS="ok"
          [ "${{ steps.gate.outputs.gate_result }}" = "success" ] || STATUS="error"
          START_TIME=""
          [ -f /tmp/ci_job_start_time ] && START_TIME=$(cat /tmp/ci_job_start_time)
          python3 scripts/ci/ci_trace_report.py --service xiaoxia-saas-ci --status $STATUS --start-time "$START_TIME" || true
@@ -16,15 +16,15 @@ permissions:
 jobs:
  monitor:
    name: Monitor CI Trigger Reliability
-    runs-on: ci-l2
+    runs-on: ubuntu-latest
    timeout-minutes: 5
    steps:
      - name: Checkout code
-        shell: sh
+        uses: actions/checkout@v3
-        env:
+        # 网络波动自动重试2次
-          GITHUB_TOKEN: ${{ github.token }}
+        retry:
-        run: |
+          max_attempts: 2
-          curl -sH "Authorization: token $GITHUB_TOKEN"             "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"             | bash
+          retry_on: error
      - name: Check CI trigger status for all open PRs
        env:
@@ -15,18 +15,20 @@ concurrency:
 jobs:
  code-review:
    name: AI Code Review
-    runs-on: ci-l2
+    runs-on: ubuntu-latest
    # 跳过草稿 PR
    if: ${{ !gitea.event.pull_request.draft }}
    steps:
      # actions/checkout 由 runner 在宿主机层面处理，不受容器网络影响
      - name: Checkout code
-        shell: sh
+        uses: actions/checkout@v3
-        env:
+        with:
-          GITHUB_TOKEN: ${{ github.token }}
+          fetch-depth: 0
-        run: |
+        # 网络波动自动重试2次
-          curl -sH "Authorization: token $GITHUB_TOKEN"             "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}"             | bash
+        retry:
          max_attempts: 2
          retry_on: error
      - name: Install dependencies
        run: |
@@ -48,7 +50,6 @@ jobs:
          GITEA_TOKEN: ${{ secrets.REVIEW_GITEA_TOKEN }}
          REPO_NAME: ${{ gitea.repository }}
          PR_NUMBER: ${{ gitea.event.pull_request.number }}
          PR_HEAD_SHA: ${{ gitea.event.pull_request.head.sha }}
          # LLM 提供商: coze (扣子原生Bot) / openai (OpenAI兼容)
          LLM_PROVIDER: "coze"
          # 扣子模式配置（默认国内站 api.coze.cn）
@@ -61,9 +62,8 @@ jobs:
          LLM_TIMEOUT: "120"
        run: |
          python3 scripts/ci_code_review.py
-        # 注意：脚本退出码决定job状态
+        # 审查脚本异常不影响 CI 通过
-        # - 有阻塞级问题 → exit 1 → job失败 → 门禁拦截
+        continue-on-error: true
        # - 无阻塞级问题/LLM异常 → exit 0 → 通过（fail-open）
      - name: Report CI trace
        if: always()
@@ -1,5 +1,4 @@
 name: Daily Health Check
 # 注意：使用 curl step_checkout.sh 方式以兼容 docker runner
 on:
  schedule:
@@ -13,7 +12,7 @@ jobs:
  # ── 1. 生产环境冒烟测试 ─────────────────────────────────────────────
  production-smoke:
    name: Production Smoke Test
-    runs-on: ci-l2
+    runs-on: saas
    timeout-minutes: 8
    outputs:
      report: ${{ steps.smoke.outputs.report }}
@@ -24,9 +23,47 @@ jobs:
        env:
          GITHUB_TOKEN: ${{ github.token }}
        run: |
-          curl -sH "Authorization: token $GITHUB_TOKEN" \
+          set -eu
-            "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" \
+          python3 - <<'PY'
-            | bash
+          import io, os, tarfile, time, urllib.request, urllib.error
          url = f"{os.environ['GITHUB_API_URL']}/repos/{os.environ['GITHUB_REPOSITORY']}/archive/{os.environ['GITHUB_SHA']}.tar.gz"
          request = urllib.request.Request(url, headers={"Authorization": f"token {os.environ['GITHUB_TOKEN']}"})
          last_err = None
          for attempt in range(5):
              try:
                  with urllib.request.urlopen(request, timeout=120) as response:
                      archive = response.read()
                  break
              except urllib.error.HTTPError as e:
                  last_err = e
                  if e.code >= 500 and attempt < 4:
                      wait = 2 ** attempt
                      print(f"Checkout HTTP {e.code}, retrying in {wait}s (attempt {attempt+1}/5)...")
                      time.sleep(wait)
                      continue
                  raise
              except Exception as e:
                  last_err = e
                  if attempt < 4:
                      wait = 2 ** attempt
                      print(f"Checkout error: {e}, retrying in {wait}s (attempt {attempt+1}/5)...")
                      time.sleep(wait)
                      continue
                  raise
          else:
              raise last_err
          with tarfile.open(fileobj=io.BytesIO(archive), mode='r:gz') as tar:
              root_prefix = tar.getmembers()[0].name.split('/', 1)[0] + '/'
              for member in tar.getmembers():
                  name = member.name
                  if name == root_prefix[:-1]:
                      continue
                  if name.startswith(root_prefix):
                      member.name = name[len(root_prefix):]
                  if member.name:
                      tar.extract(member, '.')
          PY
      - name: Production health check & smoke test
        id: smoke
        shell: sh
@@ -84,10 +121,10 @@ jobs:
  # ── 2. Staging API 集成测试 ─────────────────────────────────────────
  staging-api-tests:
    name: Staging API Integration Tests
-    runs-on: ci-l2
+    runs-on: saas
    timeout-minutes: 10
    outputs:
-      report: ${{ steps.report.outputs.report }}
+      report: ${{ steps.smoke.outputs.report }}
    steps:
      - name: Checkout code
@@ -95,15 +132,50 @@ jobs:
        env:
          GITHUB_TOKEN: ${{ github.token }}
        run: |
-          curl -sH "Authorization: token $GITHUB_TOKEN" \
+          set -eu
-            "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" \
+          python3 - <<'PY'
-            | bash
+          import io, os, tarfile, time, urllib.request, urllib.error
          url = f"{os.environ['GITHUB_API_URL']}/repos/{os.environ['GITHUB_REPOSITORY']}/archive/{os.environ['GITHUB_SHA']}.tar.gz"
          request = urllib.request.Request(url, headers={"Authorization": f"token {os.environ['GITHUB_TOKEN']}"})
          last_err = None
          for attempt in range(5):
              try:
                  with urllib.request.urlopen(request, timeout=120) as response:
                      archive = response.read()
                  break
              except urllib.error.HTTPError as e:
                  last_err = e
                  if e.code >= 500 and attempt < 4:
                      wait = 2 ** attempt
                      print(f"Checkout HTTP {e.code}, retrying in {wait}s (attempt {attempt+1}/5)...")
                      time.sleep(wait)
                      continue
                  raise
              except Exception as e:
                  last_err = e
                  if attempt < 4:
                      wait = 2 ** attempt
                      print(f"Checkout error: {e}, retrying in {wait}s (attempt {attempt+1}/5)...")
                      time.sleep(wait)
                      continue
                  raise
          else:
              raise last_err
          with tarfile.open(fileobj=io.BytesIO(archive), mode='r:gz') as tar:
              root_prefix = tar.getmembers()[0].name.split('/', 1)[0] + '/'
              for member in tar.getmembers():
                  name = member.name
                  if name == root_prefix[:-1]:
                      continue
                  if name.startswith(root_prefix):
                      member.name = name[len(root_prefix):]
                  if member.name:
                      tar.extract(member, '.')
          PY
      - name: Run API smoke test on staging
        id: smoke
        shell: sh
        env:
          STAGING_TEST_USER: ${{ secrets.STAGING_TEST_USER }}
          STAGING_TEST_PASSWORD: ${{ secrets.STAGING_TEST_PASSWORD }}
        run: |
          set +e
          START_TIME=$(date +%s)
@@ -111,8 +183,8 @@ jobs:
          docker run --rm \
            -e BASE_URL=https://staging-api.xiaoxiajianji.com \
            -e WEB_URL=https://staging.xiaoxiajianji.com \
-            -e TEST_USER="$STAGING_TEST_USER" \
+            -e TEST_USER=18314979086@163.com \
-            -e TEST_PASSWORD="$STAGING_TEST_PASSWORD" \
+            -e TEST_PASSWORD=Ying1234 \
            -e CLEANUP_ENABLED=1 \
            -e PERF_CHECK_ENABLED=1 \
            -e PERF_WARN_THRESHOLD_MS=500 \
@@ -198,10 +270,10 @@ jobs:
  # ── 3. Staging 浏览器 E2E ──────────────────────────────────────────
  staging-e2e:
    name: Staging Browser E2E
-    runs-on: ci-l2
+    runs-on: saas
    timeout-minutes: 15
    outputs:
-      report: ${{ steps.e2e.outputs.report }}
+      report: ${{ steps.smoke.outputs.report }}
    steps:
      - name: Checkout code
@@ -209,9 +281,47 @@ jobs:
        env:
          GITHUB_TOKEN: ${{ github.token }}
        run: |
-          curl -sH "Authorization: token $GITHUB_TOKEN" \
+          set -eu
-            "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/raw/scripts/ci/step_checkout.sh?ref=${GITHUB_SHA}" \
+          python3 - <<'PY'
-            | bash
+          import io, os, tarfile, time, urllib.request, urllib.error
          url = f"{os.environ['GITHUB_API_URL']}/repos/{os.environ['GITHUB_REPOSITORY']}/archive/{os.environ['GITHUB_SHA']}.tar.gz"
          request = urllib.request.Request(url, headers={"Authorization": f"token {os.environ['GITHUB_TOKEN']}"})
          last_err = None
          for attempt in range(5):
              try:
                  with urllib.request.urlopen(request, timeout=120) as response:
                      archive = response.read()
                  break
              except urllib.error.HTTPError as e:
                  last_err = e
                  if e.code >= 500 and attempt < 4:
                      wait = 2 ** attempt
                      print(f"Checkout HTTP {e.code}, retrying in {wait}s (attempt {attempt+1}/5)...")
                      time.sleep(wait)
                      continue
                  raise
              except Exception as e:
                  last_err = e
                  if attempt < 4:
                      wait = 2 ** attempt
                      print(f"Checkout error: {e}, retrying in {wait}s (attempt {attempt+1}/5)...")
                      time.sleep(wait)
                      continue
                  raise
          else:
              raise last_err
          with tarfile.open(fileobj=io.BytesIO(archive), mode='r:gz') as tar:
              root_prefix = tar.getmembers()[0].name.split('/', 1)[0] + '/'
              for member in tar.getmembers():
                  name = member.name
                  if name == root_prefix[:-1]:
                      continue
                  if name.startswith(root_prefix):
                      member.name = name[len(root_prefix):]
                  if member.name:
                      tar.extract(member, '.')
          PY
      - name: Run Playwright E2E on staging
        id: e2e
        shell: sh
@@ -261,7 +371,7 @@ jobs:
  # ── 4. 性能基线巡检 ────────────────────────────────────────────────
  performance-check:
    name: Performance Baseline Check
-    runs-on: ci-l2
+    runs-on: saas
    timeout-minutes: 8
    outputs:
      report: ${{ steps.report.outputs.report }}
@@ -270,9 +380,6 @@ jobs:
      - name: Run performance baseline checks
        id: perf
        shell: sh
        env:
          STAGING_TEST_USER: ${{ secrets.STAGING_TEST_USER }}
          STAGING_TEST_PASSWORD: ${{ secrets.STAGING_TEST_PASSWORD }}
        run: |
          set +e
          START_TIME=$(date +%s)
@@ -308,10 +415,9 @@ jobs:
          # 先登录获取 token
          echo "--- 准备: 获取测试 Token ---"
          LOGIN_BODY="{\"email\":\"$STAGING_TEST_USER\",\"password\":\"$STAGING_TEST_PASSWORD\"}"
          AUTH_RESP=$(curl -s -w "\n%{http_code}" -X POST \
            -H "Content-Type: application/json" \
-            -d "$LOGIN_BODY" \
+            -d '{"email":"18314979086@163.com","password":"Ying1234"}' \
            "https://staging-api.xiaoxiajianji.com/api/v1/auth/login" \
            --max-time 10 2>&1)
          AUTH_CODE=$(echo "$AUTH_RESP" | tail -1)
@@ -341,7 +447,7 @@ jobs:
            # 构建 curl 命令
            CURL_ARGS="-s -o /dev/null -w '%{http_code} %{time_total}' --max-time 30"
            if [ "$method" = "POST" ]; then
-              CURL_ARGS="$CURL_ARGS -X POST -H 'Content-Type: application/json' -d \"$LOGIN_BODY\""
+              CURL_ARGS="$CURL_ARGS -X POST -H 'Content-Type: application/json' -d '{\"email\":\"18314979086@163.com\",\"password\":\"Ying1234\"}'"
            fi
            if [ -n "$TOKEN" ] && [ "$name" != "健康检查" ]; then
              CURL_ARGS="$CURL_ARGS -H 'Authorization: Bearer $TOKEN'"
@@ -389,9 +495,6 @@ jobs:
      - name: Generate performance report
        id: report
        shell: sh
        env:
          STAGING_TEST_USER: ${{ secrets.STAGING_TEST_USER }}
          STAGING_TEST_PASSWORD: ${{ secrets.STAGING_TEST_PASSWORD }}
        run: |
          set +e
          echo ""
@@ -406,11 +509,10 @@ jobs:
          RESULTS=""
          START_TIME=$(date +%s)
          LOGIN_BODY="{\"email\":\"$STAGING_TEST_USER\",\"password\":\"$STAGING_TEST_PASSWORD\"}"
          # 先登录获取 token
          AUTH_RESP=$(curl -s -w "\n%{http_code}" -X POST \
            -H "Content-Type: application/json" \
-            -d "$LOGIN_BODY" \
+            -d '{"email":"18314979086@163.com","password":"Ying1234"}' \
            "https://staging-api.xiaoxiajianji.com/api/v1/auth/login" \
            --max-time 10 2>&1)
          AUTH_CODE=$(echo "$AUTH_RESP" | tail -1)
@@ -426,7 +528,7 @@ jobs:
            local CURL_ARGS="-s -o /dev/null -w '%{http_code} %{time_total}' --max-time 30"
            if [ "$method" = "POST" ]; then
-              CURL_ARGS="$CURL_ARGS -X POST -H 'Content-Type: application/json' -d \"$LOGIN_BODY\""
+              CURL_ARGS="$CURL_ARGS -X POST -H 'Content-Type: application/json' -d '{\"email\":\"18314979086@163.com\",\"password\":\"Ying1234\"}'"
            fi
            if [ -n "$TOKEN" ] && [ "$name" != "健康检查" ]; then
              CURL_ARGS="$CURL_ARGS -H 'Authorization: Bearer $TOKEN'"
@@ -529,7 +631,7 @@ jobs:
  # ── 5. 每日巡检汇总报告 ────────────────────────────────────────────
  daily-report:
    name: Daily Check Report
-    runs-on: ci-l2
+    runs-on: saas
    timeout-minutes: 2
    if: always()
    needs:
@@ -8,11 +8,6 @@ on:
 permissions:
  contents: read
 concurrency:
  group: pr-automation-${{ gitea.event.pull_request.number }}
  cancel-in-progress: true
 jobs:
  auto-approve:
    name: Auto Approve on CI Green
@@ -61,7 +56,7 @@ jobs:
    name: Auto Merge on CI Green + Approved
    runs-on: ci-check
    if: github.event_name == 'pull_request' && !github.event.pull_request.draft && github.event.pull_request.base.ref == 'develop'
-    timeout-minutes: 3  # 短作业模式：检查一次，不满足就退出，由pr-auto-scan每5分钟定时兜底
+    timeout-minutes: 45  # 长等待模式：等CI全绿后自动合并，不遗漏任何PR
    steps:
      - name: Checkout code
        shell: sh
@@ -120,7 +120,7 @@ jobs:
        PREVIEW_SSH_KEY: ${{ secrets.PREVIEW_SSH_KEY }}
      run: |
        set -eux
-        preview_host="${PREVIEW_SSH_HOST:-47.98.113.167}"
+        preview_host="${PREVIEW_SSH_HOST:-172.30.18.197}"
        preview_user="${PREVIEW_SSH_USER:-deploy}"
        preview_port="${PREVIEW_SSH_PORT:-22222}"
        preview_dir="/var/www/preview/pr-${PR_NUMBER}"
@@ -93,28 +93,44 @@ jobs:
      shell: sh
      run: |
        set -eu
-        cd apps/web
+        NPM_CACHE_VOLUME="xiaoxia-npm-cache"
        if ! docker volume inspect "$NPM_CACHE_VOLUME" >/dev/null 2>&1; then
          docker volume create "$NPM_CACHE_VOLUME" >/dev/null
          echo "Created npm cache volume: $NPM_CACHE_VOLUME"
        fi
-        # Install dependencies with retry
+        docker run --rm \
-        for i in 1 2 3; do
+          -v "$PWD:/workspace" \
-          npm ci --registry=https://registry.npmmirror.com --no-audit --no-fund && break
+          -v "$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules" \
-          echo "npm install failed, retry $i/3..."
+          -w /workspace/apps/web \
-          [ $i -eq 3 ] && exit 1
+          -e VITE_API_URL=https://staging-api.xiaoxiajianji.com \
-          rm -rf node_modules
+          docker.m.daocloud.io/library/node:20 \
-          sleep 5
+          sh -lc '
-        done
+            PACKAGE_LOCK_HASH=$(md5sum package-lock.json 2>/dev/null | cut -d" " -f1)
-
+            CACHE_HASH_FILE="node_modules/.package-lock-hash"
-        # TypeScript check
+            CACHE_VALID=false
-        echo "=== TypeScript check ==="
+            if [ -f "$CACHE_HASH_FILE" ] && [ "$(cat "$CACHE_HASH_FILE")" = "$PACKAGE_LOCK_HASH" ] && [ -x "node_modules/.bin/vite" ] && [ -x "node_modules/.bin/tsc" ]; then
-        ./node_modules/.bin/tsc --noEmit
+              CACHE_VALID=true
-
+              echo "Cache hit: dependencies valid, skipping npm ci"
-        # Vite build
+            fi
-        echo "=== Vite build ==="
+            if [ "$CACHE_VALID" = "false" ]; then
-        export VITE_API_URL=https://staging-api.xiaoxiajianji.com
+              echo "Cache miss or invalid: running npm ci..."
-        ./node_modules/.bin/vite build
+              if ! npm ci; then
-
+                echo "npm ci failed, cleaning node_modules and retrying..."
-        echo "=== Build completed ==="
+                rm -rf node_modules
-        ls -la dist/
+                mkdir -p node_modules
                npm ci
              fi
              echo "$PACKAGE_LOCK_HASH" > "$CACHE_HASH_FILE"
              echo "Dependencies installed, cache updated"
            fi
            echo "Running TypeScript check..."
            npx --no-install tsc
            echo "Running Vite build..."
            npx --no-install vite build
            echo "Build completed successfully"
            ls -la dist/
          '
    - name: Install SSH client and rsync
      shell: sh
@@ -1,222 +0,0 @@
 ---
 AIGC:
    Label: "1"
    ContentProducer: 001191110102MACQD9K64018705
    ProduceID: 15868733686388_0/project_7655981463858544923-files/docs/1197_preview_generation_proposal.md
    ReservedCode1: ""
    ContentPropagator: 001191110102MACQD9K64028705
    PropagateID: 15868733686388#1785468313901
    ReservedCode2: ""
 ---
 # #1197 预览生成接口方案评估
 ## 背景
 智能剪辑「一键生成」流程中，第3步预览生成当前被跳过，直接进入下一步。需要实现真正的预览生成功能，让用户在正式生成前能看到效果预览。
 ## 现状分析
 ### 现有生成链路
 ```
 API 触发生成 → GenerationTask入库 → Celery异步任务 → UnifiedRenderService渲染 → OSS上传 → 更新状态
 ```
 **关键节点：**
 1. **API层**：`POST /generation-tasks` 或 `POST /templates/{id}/generate` 触发生成
 2. **任务调度**：Celery task `worker.generate_video`
 3. **渲染引擎**：`UnifiedRenderService`（统一渲染引擎，已接入9个效果层）
 4. **输出配置**：默认 720p (1280x720)，支持 `resolution` 字段自定义
 5. **产物存储**：`GeneratedVideo` 表记录，OSS 存储视频文件
 ### 已有可复用能力
 | 能力 | 位置 | 是否可复用 |
 |------|------|-----------|
 | 任务创建与状态管理 | `GenerationTask` + `CreateGenerationTaskUseCase` | ✅ 是 |
 | 素材下载与预处理 | `_download_video_assets` / `_download_voice_asset` | ✅ 是 |
 | 统一渲染引擎 | `UnifiedRenderService` | ✅ 是 |
 | 分辨率配置 | `resolution` 字段已支持 | ✅ 是 |
 | 混音与后处理 | `_render_video` 内流程 | ✅ 是 |
 | OSS 上传与查重 | `_upload_and_dedup` | ✅ 是 |
 | 进度追踪 | `append_log` / `progress` 字段 | ✅ 是 |
 ## 方案对比
 ### 方案A：复用现有生成链路 + is_preview 标记（推荐）
 **思路**：在现有 GenerationTask 上加 `is_preview` 标记，预览生成走完整链路但参数降级。
 **改动点：**
 1. **数据模型**：`GenerationTask` 加 `is_preview: bool` 字段（默认 false）；`GeneratedVideo` 加 `is_preview: bool`
 2. **API 层**：生成接口加 `is_preview` 参数，预览任务不计入配额
 3. **渲染参数**：预览模式下自动调整
   - 分辨率：480p (854x480)
   - 时长：限制前 15 秒（或模板第一个片段）
   - 码率：降低至 1.5Mbps（正式 4Mbps）
   - 效果层：跳过高级转场/粒子特效等耗时效果
 4. **任务调度**：预览任务走低优先级队列（或复用现有队列，标记优先级）
 5. **前端对接**：预览生成结果带 `is_preview=true` 标记，前端展示"预览"标签
 **优点：**
 - 代码复用率 90%+，改动最小
 - 与正式生成逻辑一致，预览效果真实可信
 - 进度查询、结果展示等功能直接复用
 - 后续可平滑升级：预览满意后一键转正式生成
 **缺点：**
 - 需要区分预览和正式任务，避免数据混淆
 - 预览任务和正式任务竞争同一队列资源（可后续优化为独立队列）
 **开发量估算**：2-3 天
 - 数据模型 + 迁移：0.5 天
 - API 层改造：0.5 天
 - 渲染参数降级：1 天
 - 测试 + 联调：1 天
 ---
 ### 方案B：新建独立预览接口 + 轻量渲染逻辑
 **思路**：新建独立的预览生成接口，使用简化的渲染逻辑（如只拼接素材+基础配音，跳过大部分效果）。
 **改动点：**
 1. 新增 `PreviewTask` 数据模型
 2. 新增 `POST /api/v1/preview/generate` 接口
 3. 新增独立的 Celery task `worker.generate_preview`
 4. 简化渲染流程：只做素材裁剪+拼接+配音，跳过转场/滤镜/字幕特效等
 **优点：**
 - 完全隔离，不影响正式生成链路
 - 可以做极致优化，预览生成速度快
 - 数据模型清晰，不会混淆
 **缺点：**
 - 代码重复率高，两套生成逻辑维护成本翻倍
 - 预览效果与正式生成可能不一致（效果层差异）
 - 前端需要对接两套接口
 - 无法从预览升级为正式生成（需重新走完整流程）
 **开发量估算**：4-5 天
 - 数据模型 + 接口：1 天
 - 简化渲染逻辑：2 天
 - 测试 + 联调：1-2 天
 ---
 ### 方案C：图片预览（首帧/关键帧截图）
 **思路**：不生成视频，只生成几张关键帧的预览图片。
 **优点：**
 - 生成速度极快（秒级）
 - 资源消耗小
 **缺点：**
 - 预览效果差，用户无法感知动态效果
 - 无法验证配音、转场、节奏等时间维度的效果
 - 用户体验不佳，不如"真预览"有说服力
 **开发量估算**：1-2 天
 ---
 ## 推荐方案：方案A（复用现有生成链路）
 ### 核心理由
 1. **效果保真**：预览和正式生成用同一套渲染引擎，效果一致，用户信任度高
 2. **开发效率**：90% 代码复用，2-3 天可上线
 3. **可扩展性强**：后续可加「预览转正式」「低分辨率快速预览」等增强功能
 4. **维护成本低**：一套生成逻辑，bug 修复和新功能同时生效
 ### 详细设计
 #### 1. 数据模型变更
 ```python
 # GenerationTask 新增字段
 is_preview: bool = False
 """是否为预览生成"""
 preview_of: str = ""
 """预览对应的正式任务 ID（或反向关联）"""
 # GeneratedVideo 新增字段
 is_preview: bool = False
 """是否为预览视频"""
 ```
 **迁移**：alembic 新增 migration，两个表各加 1-2 个字段。
 #### 2. API 层
 ```
 POST /api/v1/generation-tasks
  Body 增加 is_preview: bool = false
 POST /api/v1/templates/{id}/generate
  Query 增加 is_preview: bool = false
 ```
 **配额处理**：预览生成不计入用户配额，不占用生成次数限制。
 #### 3. 渲染参数降级
 | 参数 | 正式生成 | 预览生成 |
 |------|---------|---------|
 | 分辨率 | 720p (1280x720) | 480p (854x480) |
 | 码率 | 4 Mbps | 1.5 Mbps |
 | 时长 | 完整时长 | 前 15 秒（或第一段） |
 | 帧率 | 30 fps | 24 fps |
 | 转场效果 | 完整转场 | 仅淡入淡出（或简单切） |
 | 特效滤镜 | 全部启用 | 跳过粒子/光效等高级效果 |
 | 字幕 | 完整渲染 | 正常渲染（字幕是核心信息） |
 | 配音 | 完整混音 | 正常混音（配音是核心信息） |
 **实现方式**：在 `_render_video` 或 UnifiedRenderService 入口处，根据 `is_preview` 标记调整渲染配置。
 #### 4. 任务调度
 - 初期复用现有队列，预览任务正常排队
 - 后续如需优化，可拆分独立预览队列（低优先级）
 - 预览任务可设置较短超时时间
 #### 5. 前端对接
 - 调用生成接口时传 `is_preview=true`
 - 结果列表中预览视频带「预览」标签
 - 预览满意后可一键「升级为正式生成」（重新触发全分辨率生成，可复用素材下载缓存）
 ### 实施步骤
 **Phase 1（MVP，2天）：**
 1. 数据模型 + 迁移
 2. API 层支持 is_preview 参数
 3. 渲染分辨率降级（480p）
 4. 不计入配额
 5. 基础测试
 **Phase 2（优化，1-2天）：**
 1. 时长限制（前15秒）
 2. 效果层降级（跳高级效果）
 3. 预览任务低优先级队列
 4. 预览转正式生成功能
 ## 与前端对齐点
 1. 预览生成的触发时机（第3步自动生成？用户点击才生成？）
 2. 预览时长是固定15秒还是完整但低清？
 3. 是否需要「预览转正式生成」功能
 4. 预览视频的展示形态（和正式视频一样还是有特殊UI）
 ## 风险与注意事项
 1. **数据混淆**：确保统计、计费、列表展示时正确区分预览和正式任务
 2. **存储成本**：预览视频也占 OSS 空间，可设置自动清理（7天后自动删除）
 3. **用户预期**：要明确告诉用户这是预览，效果和正式生成一致但清晰度低
 4. **并发压力**：如果用户频繁生成预览，可能增加系统负载，需要限流
 ---
 > 本内容由 Coze AI 生成，请遵循相关法律法规及《人工智能生成合成内容标识办法》使用与传播。
@@ -1,382 +0,0 @@
 # #1197 预览生成接口技术方案（v2）
 > 更新说明：v2 新增「多版本预览生成」能力，支持一个模板生成多个不重复的预览视频，左侧列表展示，用户可挑选满意的版本转正式生成。
 ## 1. 背景与目标
 **现状**：智能剪辑「一键生成」第3步预览生成被跳过，用户直接进入正式生成，缺少效果预览环节。
 **目标**：
 1. ✅ 实现真正的预览生成（低分辨率快速出片）
 2. ✅ **支持生成 1~N 个不重复的预览版本**（默认 3 个），左侧列表展示
 3. ✅ 预览满意后可一键转正式生成（复用素材下载缓存）
 4. ✅ 不计入用户配额，不占用正式生成次数
 ---
 ## 2. 现有生成链路分析
 ### 2.1 链路总览
 ```
 API 触发生成 → GenerationTask入库 → Celery异步任务
  → 下载素材 → 构建plan/clips → UnifiedRenderService渲染
  → 混音后处理 → OSS上传 + 查重 → 更新状态
 ```
 ### 2.2 决定视频差异的变量
 要做"多个不重复版本"，先分析哪些环节可以引入变化：
 | 变量 | 当前行为 | 能否引入变化 | 影响程度 |
 |------|---------|------------|---------|
 | 素材选择 | 按 asset_ids 顺序全用 | ✅ 可随机选择子集/不同组合 | 大 |
 | 素材排序 | 按 asset_ids 顺序 | ✅ 可 shuffle 重排 | 大 |
 | 配音选择 | 固定 voice_library_id | ✅ 可选不同音色 | 中 |
 | 标题选择 | 固定 title_ids 或随机选 | ✅ 可选不同标题 | 中 |
 | BGM | 固定 bgm_config | ✅ 可选不同BGM | 小 |
 | 转场效果 | 模板固定 | ✅ 可随机化转场类型 | 小 |
 | 播放速度 | 模板固定 | ✅ 可微调速度 | 小 |
 | 分辨率/码率 | 固定 | ✅ 预览可降级 | 不影响内容 |
 ### 2.3 可复用能力
 - 任务创建与状态管理：`GenerationTask` + `CreateGenerationTaskUseCase`
 - 素材下载与预处理：`_download_all_assets`
 - 统一渲染引擎：`UnifiedRenderService`
 - 分辨率配置：`resolution` 字段已支持
 - 批量任务：`batch_id` 字段已存在（可用于预览组）
 ---
 ## 3. 总体方案：复用现有链路 + 多变体引擎
 **核心思路**：沿用 v1 的"复用现有生成链路 + is_preview 标记"方案，在此基础上增加「多版本生成」能力。
 **架构**：
 ```
 预览生成请求（count=N）
    ↓
 创建预览批次（preview_batch）
    ↓
 变体引擎生成 N 个变体参数（variation seed + 参数组合）
    ↓
 为每个变体创建 1 个 GenerationTask（is_preview=true）
    ↓
 N 个 Celery 任务并行执行（走现有生成链路，参数降级）
    ↓
 N 个结果汇聚，前端左侧列表展示
 ```
 ---
 ## 4. 详细设计
 ### 4.1 数据模型变更
 #### 4.1.1 GenerationTask 新增字段
 ```python
 # 现有字段保留，新增：
 is_preview: bool = False
 """是否为预览生成"""
 preview_batch_id: str = ""
 """预览批次 ID（同批次的 N 个预览共享一个 batch）"""
 variant_seed: int = 0
 """变体种子，用于控制随机化行为（素材选择、排序、转场等）"""
 variant_params: dict = field(default_factory=dict)
 """变体参数快照（记录本次使用了哪些素材、标题、配音等，可追溯）
   {
     "asset_ids": [...],      # 实际选用的素材子集
     "title_id": "",         # 选用的标题
     "voice_id": "",         # 选用的配音
     "transition_style": "", # 转场风格
     "bgm_track": "",        # BGM 音轨
   }
 """
 ```
 #### 4.1.2 GeneratedVideo 新增字段
 ```python
 is_preview: bool = False
 """是否为预览视频"""
 preview_batch_id: str = ""
 """所属预览批次"""
 variant_index: int = 0
 """在批次中的序号（0, 1, 2...）"""
 ```
 #### 4.1.3 迁移方案
 alembic 新增 migration，两个表各加 4 个字段，默认值为空/false，无数据回填成本。
 ---
 ### 4.2 变体引擎（Variant Engine）
 **核心组件**：根据 count 和 seed，生成 N 组互不相同的生成参数。
 #### 4.2.1 变纬度设计
 | 维度 | 策略 | 说明 |
 |------|------|------|
 | **素材子集选择** | 从素材池中随机选 M 个（M=min(素材数, 模板clip数*2)） | 版本差异最大的来源 |
 | **素材排序** | 随机打乱顺序 | 影响叙事节奏 |
 | **标题选择** | 从 title_ids 中随机选 1 个 | 影响文案内容 |
 | **配音选择** | 从 voice_ids 中随机选 1 个（如有多个） | 影响听觉体验 |
 | **转场风格** | 从预设转场池中随机选 1 种 | 影响视觉过渡 |
 | **BGM 选择** | 从 bgm 列表中随机选 1 首（如有配置） | 影响氛围 |
 #### 4.2.2 去重机制
 - 同一批次内，变体参数必须两两不同（至少素材组合或排序不同）
 - 使用 `variant_seed` 保证可复现（相同 seed → 相同变体）
 - 如果素材数量不足导致无法生成 N 个不同版本，按实际能生成的数量返回
 #### 4.2.3 接口设计
 ```python
 def generate_variants(
    count: int,
    seed: int,
    asset_pool: list[str],        # 可用素材 ID 列表
    title_pool: list[str] = [],    # 可用标题 ID 列表
    voice_pool: list[str] = [],    # 可用配音 ID 列表
    template_id: str = "",
 ) -> list[dict]:
    """
    生成 count 组变体参数。
    每组参数包含：asset_ids（选用的素材+排序）、title_id、voice_id、
    transition_style 等，确保两两不同。
    """
 ```
 ---
 ### 4.3 API 层设计
 #### 4.3.1 预览生成接口
 ```
 POST /api/v1/templates/{template_id}/generate-preview
 ```
 **请求体**：
 ```json
 {
  "asset_library_id": "lib_xxx",
  "asset_ids": ["asset_1", "asset_2", ...],
  "title_ids": ["title_1", "title_2"],
  "voice_ids": ["voice_1", "voice_2"],
  "bgm_config": {},
  "count": 3,
  "seed": 0
 }
 ```
 | 参数 | 类型 | 必填 | 默认 | 说明 |
 |------|------|------|------|------|
 | template_id | path | ✅ | - | 模板 ID |
 | asset_library_id | body | ✅ | - | 素材库 ID |
 | asset_ids | body | ✅ | - | 素材池（从中选子集/排序） |
 | title_ids | body | - | [] | 标题池（可选，不传则不用标题） |
 | voice_ids | body | - | [] | 配音池（可选） |
 | bgm_config | body | - | {} | BGM 配置 |
 | count | body | - | 3 | 生成几个预览版本（1~10） |
 | seed | body | - | 0 | 随机种子，0 表示随机 |
 **响应**：
 ```json
 {
  "preview_batch_id": "pb_xxx",
  "count": 3,
  "tasks": [
    {
      "task_id": "gen_xxx_0",
      "variant_index": 0,
      "status": "processing"
    },
    {
      "task_id": "gen_xxx_1",
      "variant_index": 1,
      "status": "processing"
    },
    ...
  ]
 }
 ```
 #### 4.3.2 预览批次查询接口
 ```
 GET /api/v1/preview-batches/{batch_id}
 ```
 返回批次内所有预览任务的状态、结果（已完成的带 video_url）。
 **响应**：
 ```json
 {
  "preview_batch_id": "pb_xxx",
  "count": 3,
  "completed_count": 2,
  "tasks": [
    {
      "task_id": "gen_xxx_0",
      "variant_index": 0,
      "status": "completed",
      "video_url": "https://oss.xxx/preview/xxx.mp4",
      "duration": 15.5,
      "thumbnail_url": "https://oss.xxx/preview/xxx.jpg"
    },
    ...
  ]
 }
 ```
 #### 4.3.3 预览转正式生成
 ```
 POST /api/v1/preview-batches/{batch_id}/tasks/{task_id}/promote
 ```
 将某个预览版本升级为正式生成（复用素材缓存，重新全分辨率渲染）。
 ---
 ### 4.4 渲染参数降级
 预览模式下自动调整以下参数：
 | 参数 | 正式生成 | 预览生成 |
 |------|---------|---------|
 | 分辨率 | 720p (1280x720) | 480p (854x480) |
 | 码率 | 4 Mbps | 1.5 Mbps |
 | 帧率 | 30 fps | 24 fps |
 | 时长 | 完整时长 | 前 15 秒（或第一段完整clip） |
 | 转场效果 | 完整转场 | 仅淡入淡出 |
 | 高级特效 | 全部启用 | 跳过粒子/光效等 |
 | 字幕 | 完整渲染 | 正常渲染 |
 | 配音 | 完整混音 | 正常混音 |
 | 输出质量 | high | medium |
 **实现位置**：`_render_video` 函数入口处，根据 `is_preview` 标记调整渲染配置。
 ---
 ### 4.5 任务调度
 - **并行执行**：N 个预览任务并行提交到 Celery，不排队等待
 - **低优先级**：预览任务走独立队列（`preview_queue`），不抢占正式生成资源
 - **超时控制**：预览任务超时时间 5 分钟（正式 30 分钟）
 - **自动清理**：预览视频 7 天后自动从 OSS 删除，任务记录标记为 archived
 ---
 ## 5. 前端对接要点
 ### 5.1 交互流程
 ```
 第2步选素材 → 第3步点击"生成预览"
  → 显示 loading + 进度
  → 预览陆续完成，左侧列表逐张出现
  → 用户点击左侧不同版本，右侧预览区切换
  → 用户选中满意版本 → 点击"正式生成"
 ```
 ### 5.2 需要对齐的接口
 1. **预览创建**：`POST /templates/{id}/generate-preview`
 2. **批次状态轮询**：`GET /preview-batches/{id}`（建议 2s 轮询，或走 SSE）
 3. **预览转正式**：`POST /preview-batches/{id}/tasks/{task_id}/promote`
 ### 5.3 数据格式对齐
 预览视频条目结构：
 ```json
 {
  "id": "gen_xxx",
  "variant_index": 0,
  "status": "completed",
  "video_url": "https://...",
  "duration": 15.5,
  "file_size": 2850000,
  "thumbnail_url": "https://...",
  "is_preview": true
 }
 ```
 ---
 ## 6. 配额与计费
 - 预览生成**不计入**用户配额
 - 同一模板 + 同一素材池，每天最多生成 3 次多版本预览（防滥用）
 - 单个预览批次最多 10 个版本
 ---
 ## 7. 实施步骤
 ### Phase 1：单版本预览（MVP，2 天）
 1. 数据模型 + 迁移（is_preview 字段）
 2. API 层支持 is_preview 参数
 3. 渲染分辨率降级（480p）
 4. 不计入配额
 5. 基础测试
 ### Phase 2：多版本预览（3 天）
 1. 变体引擎实现（素材随机选择 + 排序 + 去重）
 2. preview_batch 批次管理
 3. 批量创建 N 个预览任务
 4. 批次查询接口
 5. 前端联调
 ### Phase 3：预览转正式 + 优化（2 天）
 1. 预览转正式生成接口（promote）
 2. 素材下载缓存复用
 3. 独立预览队列（低优先级）
 4. 自动清理机制
 5. 完整测试 + 压测
 ---
 ## 8. 风险与注意事项
 | 风险 | 影响 | 应对 |
 |------|------|------|
 | 并发预览任务过多打满 worker | 正式生成被阻塞 | 独立预览队列 + 限流 |
 | 变体生成的视频差异不够大 | 用户觉得"都一样" | 优先素材子集+排序差异，保证视觉差异 |
 | 预览视频占用 OSS 存储 | 存储成本上升 | 7 天自动清理 + 低码率 |
 | N 个版本同时下载重复素材 | 带宽浪费 | 批次内共享一次下载（Phase 3 优化） |
 | 用户预期管理 | 以为预览就是最终效果 | 明确标注"预览版"，说明分辨率差异 |
 ---
 ## 9. 开发量估算
 | 阶段 | 后端 | 前端 | 合计 |
 |------|------|------|------|
 | Phase 1 单版本预览 | 2 天 | 1 天 | 3 天 |
 | Phase 2 多版本预览 | 3 天 | 2 天 | 5 天 |
 | Phase 3 转正式+优化 | 2 天 | 1 天 | 3 天 |
 | **总计** | **7 天** | **4 天** | **~7 天（并行）** |
 ---
 ## 10. 与 v1 方案的差异总结
 1. **新增多版本能力**：从"生成1个预览"升级为"生成N个不重复预览"
 2. **新增变体引擎**：负责素材选择/排序/配音/标题的随机化
 3. **新增批次概念**：preview_batch 管理一组预览任务
 4. **新增 promote 接口**：预览转正式生成
 5. **独立队列**：预览不抢占正式生成资源
 6. **开发量**：从 2-3 天增加到约 7 天（后端）
@@ -1,64 +0,0 @@
 """#642 - 生成任务新增 bgm_config 字段
 Revision ID: 052_generation_task_bgm_config
 Revises: 051_generation_task_resolution
 Create Date: 2026-07-25
 Changes:
 1. generation_tasks 表新增 bgm_config 字段（JSON类型），存储用户自定义BGM配置
 2. 为空时使用默认空字典
 背景：
 #642 一键生成支持自定义BGM 功能在 SQLAlchemy 模型中加了 bgm_config 字段，
 但遗漏了 alembic migration，导致 staging 环境数据库没有该列，
 创建生成任务时直接 500。
 """
 import sqlalchemy as sa
 from alembic import context, op
 revision = "052_generation_task_bgm_config"
 down_revision = "051_generation_task_resolution"
 branch_labels = None
 depends_on = None
 def upgrade() -> None:
    conn = op.get_bind()
    if context.get_context().dialect.name == "postgresql":
        # 检查列是否已存在（幂等）
        result = conn.execute(
            sa.text(
                "SELECT column_name FROM information_schema.columns "
                "WHERE table_name = 'generation_tasks' AND column_name = 'bgm_config'"
            )
        )
        if result.scalar() is not None:
            return
    op.add_column(
        "generation_tasks",
        sa.Column(
            "bgm_config",
            sa.JSON,
            nullable=False,
            server_default=sa.text("'{}'::json"),
        ),
    )
 def downgrade() -> None:
    conn = op.get_bind()
    if context.get_context().dialect.name == "postgresql":
        # 检查列是否存在（幂等）
        result = conn.execute(
            sa.text(
                "SELECT column_name FROM information_schema.columns "
                "WHERE table_name = 'generation_tasks' AND column_name = 'bgm_config'"
            )
        )
        if result.scalar() is None:
            return
    op.drop_column("generation_tasks", "bgm_config")
@@ -1,61 +0,0 @@
 """#1197 - 预览生成：generation_tasks 表新增 is_preview 字段
 Revision ID: 053
 Revises: 052
 Create Date: 2026-08-15
 Changes:
 1. generation_tasks 表新增 is_preview 字段，标记是否为预览生成任务（低清 480p）
 2. 默认 False，与现有正式生成任务兼容
 3. 加索引以支持按预览/正式任务筛选
 """
 import sqlalchemy as sa
 from alembic import context, op
 revision = "053_generation_task_is_preview"
 down_revision = "052_generation_task_bgm_config"
 branch_labels = None
 depends_on = None
 def upgrade() -> None:
    conn = op.get_bind()
    if context.get_context().dialect.name == "postgresql":
        # 检查列是否已存在（幂等）
        result = conn.execute(
            sa.text(
                "SELECT column_name FROM information_schema.columns "
                "WHERE table_name = 'generation_tasks' AND column_name = 'is_preview'"
            )
        )
        if result.scalar() is not None:
            return
    op.add_column(
        "generation_tasks",
        sa.Column("is_preview", sa.Boolean, nullable=False, server_default=sa.text("false")),
    )
    # 加索引
    op.create_index(
        "ix_generation_tasks_is_preview",
        "generation_tasks",
        ["is_preview"],
    )
 def downgrade() -> None:
    conn = op.get_bind()
    if context.get_context().dialect.name == "postgresql":
        result = conn.execute(
            sa.text(
                "SELECT column_name FROM information_schema.columns "
                "WHERE table_name = 'generation_tasks' AND column_name = 'is_preview'"
            )
        )
        if result.scalar() is None:
            return
    op.drop_index("ix_generation_tasks_is_preview", table_name="generation_tasks")
    op.drop_column("generation_tasks", "is_preview")
@@ -1,82 +0,0 @@
 """确认生成 API 改造：为 generation_tasks 表添加 source_task_id、output_width、output_height、cover_url、custom_title 字段
 Revision ID: 054_confirm_gen_fields
 Revises: 053_generation_task_is_preview
 Create Date: 2026-08-16
 Changes:
 1. generation_tasks 表新增 source_task_id（来源预览任务 ID，带索引）
 2. generation_tasks 表新增 output_width / output_height（动态输出分辨率）
 3. generation_tasks 表新增 cover_url / custom_title（自定义封面和标题）
 """
 import sqlalchemy as sa
 from alembic import op
 revision = "054_confirm_gen_fields"
 down_revision = "053_generation_task_is_preview"
 branch_labels = None
 depends_on = None
 def upgrade() -> None:
    conn = op.get_bind()
    is_pg = conn.dialect.name == "postgresql"
    if is_pg:
        # 幂等检查：source_task_id 列是否已存在
        result = conn.execute(
            sa.text(
                "SELECT column_name FROM information_schema.columns "
                "WHERE table_name = 'generation_tasks' AND column_name = 'source_task_id'"
            )
        )
        if result.scalar() is not None:
            return
    # source_task_id
    op.add_column(
        "generation_tasks",
        sa.Column("source_task_id", sa.String(32), nullable=False, server_default=""),
    )
    # output_width
    op.add_column(
        "generation_tasks",
        sa.Column("output_width", sa.Integer, nullable=False, server_default=sa.text("1280")),
    )
    # output_height
    op.add_column(
        "generation_tasks",
        sa.Column("output_height", sa.Integer, nullable=False, server_default=sa.text("720")),
    )
    # cover_url
    op.add_column(
        "generation_tasks",
        sa.Column("cover_url", sa.String(1000), nullable=False, server_default=""),
    )
    # custom_title
    op.add_column(
        "generation_tasks",
        sa.Column("custom_title", sa.String(500), nullable=False, server_default=""),
    )
    # 索引
    op.create_index(
        "ix_generation_tasks_source_task_id",
        "generation_tasks",
        ["source_task_id"],
    )
 def downgrade() -> None:
    op.drop_index("ix_generation_tasks_source_task_id", table_name="generation_tasks")
    op.drop_column("generation_tasks", "custom_title")
    op.drop_column("generation_tasks", "cover_url")
    op.drop_column("generation_tasks", "output_height")
    op.drop_column("generation_tasks", "output_width")
    op.drop_column("generation_tasks", "source_task_id")
@@ -1,84 +0,0 @@
 """封面模板表 cover_templates
 Revision ID: 055_cover_templates
 Revises: 054_confirm_gen_fields
 Create Date: 2026-08-09
 Changes:
 1. 新建 cover_templates 表，支持系统预置和用户自定义封面模板
 2. user_id 为 NULL 表示系统模板，is_system 标记区分
 3. config 为 JSON 字段，存储封面配置信息
 """
 import json
 import sqlalchemy as sa
 from alembic import context, op
 revision = "055_cover_templates"
 down_revision = "054_confirm_gen_fields"
 branch_labels = None
 depends_on = None
 SYSTEM_TEMPLATES = [
    ("a8b0120fd98e44788f5a6590f983d327", "默认模板", {}),
    ("6d8c501b11424432b3df3a45ae89b1a9", "大胆红", {"background_color": "#ef4444"}),
    ("04937fb57fea4bad95e7883e71a6b246", "优雅黑", {"background_color": "#111827"}),
    ("3ff9cc821174437ca53931073e7f536e", "渐变蓝", {"background_color": "#3b82f6"}),
    ("db51b3ea8f1a4f4caa94bf2d51f27d11", "渐变紫", {"background_color": "#8b5cf6"}),
    ("5027d113432a4f798a3b4ee1644d66af", "暖橙", {"background_color": "#f97316"}),
    ("0e10def2b5a148d686416494474726c2", "清新绿", {"background_color": "#22c55e"}),
    ("38ea98ac00c04bada064006d880546f0", "科技蓝", {"background_color": "#06b6d4"}),
 ]
 def upgrade() -> None:
    conn = op.get_bind()
    if context.get_context().dialect.name == "postgresql":
        result = conn.execute(sa.text("SELECT to_regclass('public.cover_templates')"))
        if result.scalar() is not None:
            return
    op.create_table(
        "cover_templates",
        sa.Column("id", sa.String(36), primary_key=True),
        sa.Column("user_id", sa.String(36), nullable=True, index=True),
        sa.Column("name", sa.String(200), nullable=False),
        sa.Column("thumbnail_url", sa.String(1000), nullable=False, server_default=""),
        sa.Column("is_system", sa.Boolean, nullable=False, server_default=sa.false(), index=True),
        sa.Column("config", sa.JSON, nullable=False, server_default="{}"),
        sa.Column("created_at", sa.DateTime(timezone=True), nullable=False, server_default=sa.func.now()),
        sa.Column("updated_at", sa.DateTime(timezone=True), nullable=False, server_default=sa.func.now()),
    )
    # 预置系统模板 seed 数据
    cover_templates = sa.table(
        "cover_templates",
        sa.column("id", sa.String),
        sa.column("user_id", sa.String),
        sa.column("name", sa.String),
        sa.column("thumbnail_url", sa.String),
        sa.column("is_system", sa.Boolean),
        sa.column("config", sa.JSON),
        sa.column("created_at", sa.DateTime),
        sa.column("updated_at", sa.DateTime),
    )
    for tid, name, config in SYSTEM_TEMPLATES:
        conn.execute(
            cover_templates.insert().values(
                id=tid,
                user_id=None,
                name=name,
                thumbnail_url="",
                is_system=True,
                config=json.dumps(config),
                created_at=sa.func.now(),
                updated_at=sa.func.now(),
            )
        )
 def downgrade() -> None:
    op.drop_table("cover_templates")
@@ -5,11 +5,8 @@ from app.api.routes.assets import router as assets_router
 from app.api.routes.auth import router as auth_router
 from app.api.routes.chunked_upload import router as chunked_upload_router
 from app.api.routes.classification_jobs import router as classification_jobs_router
 from app.api.routes.cover_templates import router as cover_templates_router
 from app.api.routes.duplication import router as duplication_router
 from app.api.routes.feature_flags import router as feature_flags_router
 from app.api.routes.generation_cover import router as generation_cover_router
 from app.api.routes.generation_preview import router as generation_preview_router
 from app.api.routes.generation_tasks import router as generation_tasks_router
 from app.api.routes.health import router as health_check_router
 from app.api.routes.ingest_jobs import router as ingest_jobs_router
@@ -47,10 +44,6 @@ api_router.include_router(
    prefix="/tags",
    tags=["Tag"],
 )
 api_router.include_router(
    cover_templates_router,
    tags=["CoverTemplate"],
 )
 api_router.include_router(
    task_center_router,
    tags=["TaskCenter"],
@@ -94,16 +87,6 @@ api_router.include_router(
    prefix="/generation",
    tags=["Generation"],
 )
 api_router.include_router(
    generation_preview_router,
    prefix="/generation",
    tags=["Generation"],
 )
 api_router.include_router(
    generation_cover_router,
    prefix="/generation",
    tags=["Generation"],
 )
 api_router.include_router(
    titles_router,
    prefix="/titles",
@@ -19,9 +19,6 @@ from app.schemas.asset import (
    BatchTagRequest,
    CreateAssetRequest,
    ListAssetsResponse,
    SmartMatchItem,
    SmartMatchRequest,
    SmartMatchResponse,
    UpdateAssetRequest,
    UpdateAssetReviewRequest,
 )
@@ -33,7 +30,6 @@ from packages.application import (
    CreateAssetUseCase,
 )
 from packages.domain import AssetStatus, ClassificationStatus
 from packages.domain.smart_match import smart_select_assets
 logger = logging.getLogger(__name__)
@@ -523,51 +519,6 @@ def batch_mark_assets(
    )
@router.post("/smart-match", response_model=SmartMatchResponse)
 def smart_match_assets(
    request: SmartMatchRequest,
    authenticated_user: AuthenticatedUser = Depends(get_current_user),
    asset_repository: Any = Depends(get_asset_repository),
    asset_library_repository: Any = Depends(get_asset_library_repository),
    project_repository: Any = Depends(get_project_repository),
 ) -> SmartMatchResponse:
    """智能选素材：根据素材库内容，按质量分+时长均衡+新鲜度+未使用偏好综合评分，返回 Top N 素材。"""
    library = asset_library_repository.get(request.library_id)
    if library is None:
        raise HTTPException(status_code=404, detail=f"AssetLibrary {request.library_id} not found")
    check_project_access(library.project_id, authenticated_user.user.id, project_repository)
    # 获取素材库中所有 ready 素材（DB 层按 kind 过滤，避免加载不必要的数据到内存）
    # kind → file_type 映射：schema 已校验只允许 video/image/audio，与 file_type 一致
    if request.kind:
        filtered_assets = asset_repository.find_by_library_and_file_type(
            request.library_id, request.kind, status=["ready"], limit=10000
        )
    else:
        filtered_assets = asset_repository.find_by_library(
            request.library_id, status=["ready"], limit=10000
        )
    total_candidates = len(filtered_assets)
    # 调用统一智能选素材算法（kind 已在 DB 层过滤，无需重复过滤）
    results = smart_select_assets(
        filtered_assets,
        limit=request.limit,
        kind=None,
    )
    items = [
        SmartMatchItem(
            asset=_to_asset_response(r.asset),
            score=r.score,
            breakdown=r.breakdown,
        )
        for r in results
    ]
    return SmartMatchResponse(items=items, total_candidates=total_candidates)
@router.get("/{asset_id}", response_model=AssetResponse)
 def get_asset(
    asset_id: str,
@@ -671,28 +622,20 @@ def create_asset(
    asset_library_repository: Any = Depends(get_asset_library_repository),
    project_repository: Any = Depends(get_project_repository),
 ) -> AssetResponse:
-    # 先获取素材库，用于推导 project_id（前端可能不传）
+    project = project_repository.find_by_id(request.project_id)
    library = asset_library_repository.get(request.library_id)
    if library is None:
        raise HTTPException(status_code=404, detail=f"AssetLibrary {request.library_id} not found")
    # project_id 自动推导：优先用请求值，否则从 library 关联的项目获取
    project_id = request.project_id or library.project_id
    project = project_repository.find_by_id(project_id)
    if project is None:
-        raise HTTPException(status_code=404, detail=f"Project {project_id} not found")
+        raise HTTPException(status_code=404, detail=f"Project {request.project_id} not found")
    if not project.can_access(authenticated_user.user.id):
        raise HTTPException(status_code=403, detail="Access denied to project")
-    # 确保 library 和 project 归属一致
+    library = asset_library_repository.get(request.library_id)
-    if library.project_id != project_id:
+    if library is None or library.project_id != request.project_id:
-        raise HTTPException(status_code=400, detail="AssetLibrary does not belong to the specified project")
+        raise HTTPException(status_code=404, detail=f"AssetLibrary {request.library_id} not found")
    use_case = CreateAssetUseCase(asset_repository)
    item = use_case.execute(
        CreateAssetCommand(
-            project_id=project_id,
+            project_id=request.project_id,
            library_id=request.library_id,
            name=request.name,
            storage_key=request.storage_key,
@@ -1,153 +0,0 @@
 """封面模板 CRUD 路由。
 API:
  GET    /api/v1/cover-templates       - 列出当前用户可见的模板
  POST   /api/v1/cover-templates       - 创建自定义模板
  PUT    /api/v1/cover-templates/{id}  - 更新模板
  DELETE /api/v1/cover-templates/{id}  - 删除自定义模板（系统模板不可删）
 """
 import logging
 from typing import Any
 from app.auth import AuthenticatedUser, get_current_user
 from app.dependencies import get_cover_template_repository
 from app.schemas.cover_template import (
    CoverTemplateResponse,
    CreateCoverTemplateRequest,
    ListCoverTemplatesResponse,
    UpdateCoverTemplateRequest,
 )
 from fastapi import APIRouter, Depends, HTTPException, Response
 from sqlalchemy.exc import OperationalError, ProgrammingError
 from packages.domain.cover_template import CoverTemplate
 logger = logging.getLogger(__name__)
 router = APIRouter(prefix="/cover-templates", tags=["CoverTemplate"])
@router.get("", response_model=ListCoverTemplatesResponse)
 def list_cover_templates(
    skip: int = 0,
    limit: int = 100,
    authenticated_user: AuthenticatedUser = Depends(get_current_user),
    repo: Any = Depends(get_cover_template_repository),
 ) -> ListCoverTemplatesResponse:
    """列出当前用户可见的封面模板（系统模板 + 用户自定义模板）。
    当数据库表不存在时（迁移未执行），降级返回空列表而非 500。
    """
    user_id = authenticated_user.user.id
    try:
        items = repo.list_for_user(user_id, skip=skip, limit=limit)
        total = repo.count_for_user(user_id)
    except (OperationalError, ProgrammingError) as exc:
        logger.warning("cover_templates 表查询失败（可能未迁移），返回空列表: %s", exc)
        return ListCoverTemplatesResponse(items=[], total=0)
    return ListCoverTemplatesResponse(
        items=[
            CoverTemplateResponse(
                id=t.id,
                name=t.name,
                thumbnail_url=t.thumbnail_url,
                is_system=t.is_system,
                created_at=t.created_at,
                config=t.config or {},
            )
            for t in items
        ],
        total=total,
    )
@router.post("", response_model=CoverTemplateResponse, status_code=201)
 def create_cover_template(
    request: CreateCoverTemplateRequest,
    authenticated_user: AuthenticatedUser = Depends(get_current_user),
    repo: Any = Depends(get_cover_template_repository),
 ) -> CoverTemplateResponse:
    """创建用户自定义封面模板。"""
    user_id = authenticated_user.user.id
    config_dict = request.config.model_dump() if request.config else {}
    template = CoverTemplate.create_user(
        user_id=user_id,
        name=request.name,
        config=config_dict,
        thumbnail_url=request.thumbnail_url,
    )
    try:
        created = repo.create(template)
    except (OperationalError, ProgrammingError) as exc:
        logger.warning("cover_templates 表不可用（可能未迁移）: %s", exc)
        raise HTTPException(status_code=503, detail="封面模板服务暂不可用，请稍后重试") from None
    return CoverTemplateResponse(
        id=created.id,
        name=created.name,
        thumbnail_url=created.thumbnail_url,
        is_system=created.is_system,
        created_at=created.created_at,
        config=created.config,
    )
@router.put("/{template_id}", response_model=CoverTemplateResponse)
 def update_cover_template(
    template_id: str,
    request: UpdateCoverTemplateRequest,
    authenticated_user: AuthenticatedUser = Depends(get_current_user),
    repo: Any = Depends(get_cover_template_repository),
 ) -> CoverTemplateResponse:
    """更新封面模板（仅允许更新自己的模板）。"""
    user_id = authenticated_user.user.id
    try:
        template = repo.get(template_id)
    except (OperationalError, ProgrammingError) as exc:
        logger.warning("cover_templates 表不可用: %s", exc)
        raise HTTPException(status_code=503, detail="封面模板服务暂不可用，请稍后重试") from None
    if template is None:
        raise HTTPException(status_code=404, detail="模板不存在")
    if template.is_system:
        raise HTTPException(status_code=403, detail="系统模板不可修改")
    if template.user_id != user_id:
        raise HTTPException(status_code=403, detail="无权修改该模板")
    if request.name is not None:
        template.update(name=request.name)
    if request.config is not None:
        template.update(config=request.config.model_dump())
    if request.thumbnail_url is not None:
        template.update(thumbnail_url=request.thumbnail_url)
    updated = repo.update(template)
    return CoverTemplateResponse(
        id=updated.id,
        name=updated.name,
        thumbnail_url=updated.thumbnail_url,
        is_system=updated.is_system,
        created_at=updated.created_at,
        config=updated.config,
    )
@router.delete("/{template_id}", status_code=204, response_class=Response)
 def delete_cover_template(
    template_id: str,
    authenticated_user: AuthenticatedUser = Depends(get_current_user),
    repo: Any = Depends(get_cover_template_repository),
 ) -> None:
    """删除用户自定义封面模板（系统模板不可删除）。"""
    user_id = authenticated_user.user.id
    try:
        template = repo.get(template_id)
    except (OperationalError, ProgrammingError) as exc:
        logger.warning("cover_templates 表不可用: %s", exc)
        raise HTTPException(status_code=503, detail="封面模板服务暂不可用，请稍后重试") from None
    if template is None:
        raise HTTPException(status_code=404, detail="模板不存在")
    if template.is_system:
        raise HTTPException(status_code=403, detail="系统模板不可删除")
    if template.user_id != user_id:
        raise HTTPException(status_code=403, detail="无权删除该模板")
    repo.delete(template_id)
@@ -1,237 +0,0 @@
 """封面生成路由 — Generation 模块.
 端点：
 - POST /generate-cover     AI 生成封面（从预览视频中抽帧）
 挂载路径: /api/v1/generation/generate-cover
 """
 from __future__ import annotations
 import logging
 from typing import Any, List, Optional
 from app.auth import AuthenticatedUser, get_current_user
 from app.dependencies import get_db_session, get_generated_video_repository
 from app.services.edit_plan_service import EditPlanService
 from app.services.edit_template_service import EditTemplateService
 from fastapi import APIRouter, Depends, HTTPException, Query
 from pydantic import BaseModel, Field
 from sqlalchemy.orm import Session
 from packages.adapters.sqlalchemy_impl.generation_task_repository import (
    SQLAlchemyGenerationTaskRepository,
 )
 from packages.application import ListGeneratedVideosByTaskUseCase
 from packages.domain.config_schemas import normalize_plan_config
 from .templates_editor.dependencies import get_draft_plan_id, get_editor_services
 logger = logging.getLogger(__name__)
 router = APIRouter(tags=["Generation"])
 # ── Schemas ──────────────────────────────────────────────────────────────
 class GenerateCoverRequest(BaseModel):
    """AI 封面生成请求体"""
    asset_ids: List[str] = Field(default_factory=list, description="素材 ID 列表（确定视频来源）")
    cover_type: str = Field(
        default="ai_frame",
        description="封面类型: ai_frame / manual / upload / ai_regenerate",
    )
    frame_time: Optional[float] = Field(
        default=None,
        ge=0.0,
        description="手动选帧时间点（秒），仅 cover_type=manual 时有效",
    )
 class GenerateCoverResponse(BaseModel):
    """AI 封面生成响应体"""
    plan_id: str = Field(..., description="剪辑计划 ID")
    cover: dict[str, Any] = Field(..., description="封面数据（type / image_url / frame_time 等）")
 # ── Route ────────────────────────────────────────────────────────────────
@router.post("/generate-cover", response_model=GenerateCoverResponse)
 def generate_cover(
    body: GenerateCoverRequest,
    template_id: str = Query(..., description="模板 ID"),
    plan_id: str = Depends(get_draft_plan_id),
    services: tuple[EditTemplateService, EditPlanService] = Depends(get_editor_services),
    db: Session = Depends(get_db_session),
    current_user: AuthenticatedUser = Depends(get_current_user),
 ) -> GenerateCoverResponse:
    """AI 生成封面 — 从预览视频中抽帧.
    流程（串行）：
    1. 预览视频已渲染完成（通过 3 步查找获取 URL）
    2. 用裸 URL 让 MediaKit 下载视频并抽帧
    3. 帧图下载后上传到 OSS covers/ 路径
    """
    _, plan_svc = services
    plan = plan_svc.get_plan_or_raise(plan_id)
    # ── 3 步查找预览视频 URL ──────────────────────────────────────────
    # 第一步：从 plan.config 读取
    logger.info("[封面生成] 步骤1: 从 plan.config 查找 rendered_storage_key: plan_id=%s", plan_id)
    rendered_storage_key = (plan.config or {}).get("rendered_storage_key", "")
    # 第二步：如果还没有，通过 generation_task_id 查找预览任务的产物
    if not rendered_storage_key:
        generation_task_id = (plan.config or {}).get("generation_task_id", "")
        logger.info(
            "[封面生成] 步骤2: 通过 generation_task_id 查找: plan_id=%s task_id=%s", plan_id, generation_task_id
        )
        if generation_task_id:
            try:
                gen_task_repo = SQLAlchemyGenerationTaskRepository(db)
                task = gen_task_repo.get(generation_task_id)
                if task:
                    video_repo = get_generated_video_repository(db)
                    use_case = ListGeneratedVideosByTaskUseCase(video_repo)
                    videos = use_case.execute(task.id)
                    if videos:
                        rendered_storage_key = getattr(videos[0], "file_url", "") or ""
                        logger.info(
                            "[封面生成] ✅ 步骤2找到视频: plan_id=%s task_id=%s url=%s",
                            plan_id,
                            generation_task_id,
                            rendered_storage_key[:80],
                        )
                        logger.info(
                            plan_id,
                            generation_task_id,
                        )
            except Exception:
                logger.warning(
                    "封面生成: 通过 generation_task_id 查找视频失败: plan_id=%s",
                    plan_id,
                    exc_info=True,
                )
    # 第三步：按 user + template 查找最近的已完成预览任务（兜底）
    if not rendered_storage_key:
        try:
            gen_task_repo = SQLAlchemyGenerationTaskRepository(db)
            logger.info("[封面生成] 步骤3: 通过 user+template 查找: plan_id=%s template_id=%s", plan_id, template_id)
            preview_tasks = gen_task_repo.list_latest_completed_preview(
                user_id=str(current_user.user.id),
                template_id=template_id,
            )
            if preview_tasks:
                completed_preview = preview_tasks[0]
                video_repo = get_generated_video_repository(db)
                use_case = ListGeneratedVideosByTaskUseCase(video_repo)
                videos = use_case.execute(completed_preview.id)
                if videos:
                    rendered_storage_key = getattr(videos[0], "file_url", "") or ""
                    logger.info(
                        "封面视频: 通过 user+template 找到预览任务: plan_id=%s template_id=%s task_id=%s",
                        plan_id,
                        template_id,
                        completed_preview.id,
                    )
        except Exception:
            logger.warning(
                "封面警告: user+template 查找预览任务失败: plan_id=%s template_id=%s",
                plan_id,
                template_id,
                exc_info=True,
            )
    # 仍然找不到才报 400
    if not rendered_storage_key:
        logger.error("[封面生成] ❌ 找不到预览视频: plan_id=%s", plan_id)
        raise HTTPException(
            status_code=400,
            detail="请先生成预览视频，再生成封面",
        )
    # 回写到 plan.config
    plan_svc.update_plan_config(plan_id, {"rendered_storage_key": rendered_storage_key})
    # 使用裸 URL（rendered/* 已配置公开读）
    primary_video_url = None
    try:
        if rendered_storage_key.startswith("http"):
            primary_video_url = rendered_storage_key
        else:
            from packages.shared.storage import get_shared_storage_service
            storage_svc = get_shared_storage_service()
            primary_video_url = storage_svc.get_url(rendered_storage_key)
        # 防御性规范化：合并路径中的双斜杠（// -> /），但保留协议头的 ://
        # 历史数据中 project_id 为空时会产生 projects//tasks/ 路径，
        # MediaKit 的 HTTP 客户端会规范化 URL 导致 404
        if primary_video_url:
            import re as _re
            primary_video_url = _re.sub(r"(?<!:)//", "/", primary_video_url)
        logger.info(
            "获取预览视频URL用于封面生成: plan_id=%s url=%s",
            plan_id,
            primary_video_url[:80] if primary_video_url else "",
        )
    except Exception as e:
        raise HTTPException(
            status_code=500,
            detail=f"获取预览视频URL失败: {e}",
        ) from e
    # 优先使用渲染时预抽的封面候选帧（跳过 MediaKit，秒级返回）
    cover_candidates = (plan.config or {}).get("cover_candidates", [])
    if cover_candidates and body.cover_type in ("ai_frame", "ai_regenerate"):
        logger.info(
            "[封面生成] 使用预存封面候选帧: plan_id=%s count=%d",
            plan_id,
            len(cover_candidates),
        )
        first_frame = cover_candidates[0]
        cover_data = {
            "type": "ai_frame",
            "image_url": first_frame.get("image_url", ""),
            "frame_time": first_frame.get("frame_time", 0.0),
            "confidence": 0.9,
        }
        if cover_data["image_url"]:
            current_config = dict(plan.config) if plan.config else {}
            current_config["cover"] = cover_data
            normalized = normalize_plan_config(current_config)
            plan_svc.update_plan_config(plan_id, {"cover": normalized["cover"]})
            return GenerateCoverResponse(plan_id=plan_id, cover=cover_data)
    from packages.shared.ai_service import run_generate_cover
    try:
        logger.info("[封面生成] 开始调用 AI 封面生成服务: plan_id=%s", plan_id)
        cover_data = run_generate_cover(
            plan_id=plan_id,
            asset_ids=body.asset_ids,
            cover_type=body.cover_type,
            frame_time=body.frame_time,
            primary_video_url=primary_video_url,
        )
    except RuntimeError as e:
        raise HTTPException(status_code=500, detail=str(e)) from e
    current_config = dict(plan.config) if plan.config else {}
    current_config["cover"] = cover_data
    normalized = normalize_plan_config(current_config)
    plan_svc.update_plan_config(plan_id, {"cover": normalized["cover"]})
    logger.info(
        "封面生成完成: template_id=%s plan_id=%s type=%s by user=%s",
        template_id,
        plan_id,
        body.cover_type,
        current_user.user.id,
    )
    return GenerateCoverResponse(plan_id=plan_id, cover=cover_data)
@@ -1,366 +0,0 @@
 """预览生成路由 — Phase 1：单版本预览接口（创建 + 查询）。
 路径前缀：/api/v1/generation/preview（与 /generation/tasks 同体系）
 """
 from __future__ import annotations
 import logging
 from app.auth import AuthenticatedUser, get_current_user
 from app.core.storage import get_storage_service
 from app.core.task_enqueue import (
    GLOBAL_PENDING_LIMIT,
    USER_PENDING_LIMIT,
    GlobalQueueFull,
    UserPendingLimitExceeded,
    safe_enqueue_generation_task,
 )
 from app.dependencies import (
    get_asset_repository,
    get_db_session,
    get_generated_video_repository,
    get_generation_task_repository,
 )
 from app.schemas.generation_task import (
    CreatePreviewGenerationTaskRequest,
    PreviewGenerationTaskResponse,
 )
 from fastapi import APIRouter, Depends, HTTPException
 from sqlalchemy.orm import Session
 from packages.adapters.sqlalchemy_impl.edit_template_repository import (
    SQLAlchemyEditTemplateRepository,
 )
 from packages.adapters.sqlalchemy_impl.template_repository import (
    SQLAlchemyTemplateRepository,
 )
 from packages.application import (
    CreateGenerationTaskCommand,
    CreateGenerationTaskUseCase,
    GetGenerationTaskUseCase,
    ListGeneratedVideosByTaskUseCase,
 )
 logger = logging.getLogger(__name__)
 router = APIRouter()
 # 模板 mode → 视频比例映射
 _TEMPLATE_MODE_TO_RATIO = {
    "pip": "9:16",
    "standard": "16:9",
    "square": "1:1",
 }
 def _infer_video_ratio_from_template(template_id: str, db: Session, user_id: str = "") -> str:
    """从模板 mode 推断视频比例，前端未传 video_ratio 时使用。
    Returns:
        视频比例字符串（如 "9:16"），查询失败返回空字符串。
    """
    if not template_id:
        return ""
    try:
        repo = SQLAlchemyTemplateRepository(db)
        template = repo.get(template_id, user_id)
        if template:
            mode = getattr(template, "mode", "") or ""
            ratio = _TEMPLATE_MODE_TO_RATIO.get(mode.strip(), "")
            if ratio:
                logger.info(
                    "[预览生成] 从模板 mode=%s 推断 video_ratio=%s",
                    mode,
                    ratio,
                )
            return ratio
    except Exception:
        logger.warning(
            "[预览生成] 查询模板失败，跳过 video_ratio 推断: template_id=%s",
            template_id,
            exc_info=True,
        )
    return ""
 def _resolve_strategy_id_from_template(template_id: str, db: Session, user_id: str = "") -> str:
    """从模板读取 editing_mode / mode 作为 strategy_id。
    优先查新模板系统（EditTemplate.editing_mode），fallback 旧模板（Template.mode）。
    Worker 端使用 strategy_id 作为渲染 mode，为空则默认 one_take。
    """
    if not template_id:
        return ""
    # 优先查新模板系统
    try:
        new_repo = SQLAlchemyEditTemplateRepository(db)
        new_template = new_repo.get(template_id)
        if new_template and getattr(new_template, "editing_mode", ""):
            mode = new_template.editing_mode.strip()
            if mode:
                logger.info(
                    "[预览生成] 从新模板 editing_mode=%s (template_id=%s)",
                    mode,
                    template_id,
                )
                # 画中画已下线，pip/voice_pip 统一映射为 one_take
                if mode in ("pip", "voice_pip"):
                    logger.info("[预览生成] %s → one_take (画中画已下线)", mode)
                    mode = "one_take"
                return mode
    except Exception:
        logger.debug(
            "[预览生成] 新模板查询失败，尝试旧模板: template_id=%s",
            template_id,
            exc_info=True,
        )
    # fallback 旧模板系统
    try:
        old_repo = SQLAlchemyTemplateRepository(db)
        old_template = old_repo.get(template_id, user_id)
        if old_template:
            mode = getattr(old_template, "mode", "") or ""
            mode = mode.strip()
            if mode:
                logger.info(
                    "[预览生成] 从旧模板 mode=%s (template_id=%s)",
                    mode,
                    template_id,
                )
                # 画中画已下线，pip/voice_pip 统一映射为 one_take
                if mode in ("pip", "voice_pip"):
                    logger.info("[预览生成] %s → one_take (画中画已下线)", mode)
                    mode = "one_take"
                return mode
    except Exception:
        logger.warning(
            "[预览生成] 旧模板查询也失败，strategy_id 留空: template_id=%s",
            template_id,
            exc_info=True,
        )
    return ""
 def _mark_task_failed(repo, task, reason: str) -> None:
    """入队失败时将任务标记为 failed，避免产生僵尸 pending 数据。"""
    try:
        task.mark_failed(error_message=f"入队失败：{reason}")
        repo.update(task)
    except Exception:
        logger.exception("[预览生成] 标记任务失败时异常: task_id=%s", task.id)
 def _to_preview_response(task, generated_videos: list | None = None) -> PreviewGenerationTaskResponse:
    """将领域任务对象转换为预览响应 DTO。
    Args:
        task: GenerationTask 领域对象
        generated_videos: 生成的视频列表（可选），取第一个作为 video_url
    Returns:
        PreviewGenerationTaskResponse
    """
    video_url = ""
    duration = 0.0
    file_size = 0
    if generated_videos:
        first_video = generated_videos[0]
        raw_url = getattr(first_video, "file_url", "") or ""
        # rendered/* 已配置公开读，直接用裸 URL
        if raw_url.startswith("http"):
            video_url = raw_url
        else:
            storage = get_storage_service()
            video_url = storage.get_url(raw_url)
        duration = float(getattr(first_video, "duration", 0.0) or 0.0)
        file_size = int(getattr(first_video, "file_size", 0) or 0)
    # 从 extra_meta / metadata 中提取统计信息（如果有）
    extra_meta = getattr(task, "extra_meta", {}) or {}
    clip_count = int(extra_meta.get("clip_count", len(getattr(task, "asset_ids", [])) or 0))
    transition_count = int(extra_meta.get("transition_count", max(0, clip_count - 1)))
    material_usage = extra_meta.get("material_usage", {}) or {}
    # 计算生成耗时
    generate_duration = 0.0
    started_at = getattr(task, "started_at", None)
    completed_at = getattr(task, "completed_at", None)
    if started_at and completed_at:
        generate_duration = (completed_at - started_at).total_seconds()
    return PreviewGenerationTaskResponse(
        task_id=task.id,
        status=task.status.value if hasattr(task.status, "value") else str(task.status),
        progress=float(task.progress or 0.0),
        is_preview=bool(getattr(task, "is_preview", True)),
        resolution=getattr(task, "resolution", "") or "",
        video_url=video_url,
        duration=duration,
        file_size=file_size,
        clip_count=clip_count,
        transition_count=transition_count,
        material_usage=material_usage,
        error_message=task.error_message or "",
        created_at=task.created_at,
        started_at=started_at,
        finished_at=completed_at,
        generate_duration=generate_duration,
    )
@router.post("/preview", response_model=PreviewGenerationTaskResponse, status_code=201)
 def create_preview_generation_task(
    request: CreatePreviewGenerationTaskRequest,
    authenticated_user: AuthenticatedUser = Depends(get_current_user),
    generation_task_repository=Depends(get_generation_task_repository),
    db: Session = Depends(get_db_session),
    asset_repo=Depends(get_asset_repository),
 ) -> PreviewGenerationTaskResponse:
    """创建预览生成任务。
    预览渲染品质与正式生成一致（1080p, CRF 23, medium preset），确认生成时可直接复用预览产物。
    Args:
        request: 预览任务创建请求（template_id + asset_ids 等）
    Returns:
        201 + 预览任务详情
    """
    user_id = authenticated_user.user.id
    logger.info(
        "[预览生成] 接收请求: user_id=%s, template_id=%s, asset_count=%d, preview_count=%d",
        user_id,
        request.template_id,
        len(request.asset_ids),
        request.preview_count,
    )
    # 预检查队列限流
    try:
        user_pending = generation_task_repository.count_pending_by_user(user_id)
        global_pending = generation_task_repository.count_pending_total()
        if user_pending + 1 > USER_PENDING_LIMIT:
            raise UserPendingLimitExceeded(user_id=user_id, pending_count=user_pending + 1, limit=USER_PENDING_LIMIT)
        if global_pending + 1 > GLOBAL_PENDING_LIMIT:
            raise GlobalQueueFull(pending_count=global_pending + 1, limit=GLOBAL_PENDING_LIMIT)
    except UserPendingLimitExceeded as e:
        raise HTTPException(
            status_code=429,
            detail=f"您的待处理任务过多（当前 {e.pending_count - 1}/{e.limit}），请等待后再提交",
        ) from e
    except GlobalQueueFull as e:
        raise HTTPException(
            status_code=503,
            detail="系统繁忙，请稍后再试",
        ) from e
    # 确定视频比例：优先前端传入，否则从模板 mode 推断
    video_ratio = request.video_ratio or ""
    if not video_ratio and request.template_id:
        video_ratio = _infer_video_ratio_from_template(request.template_id, db, user_id)
    # 从模板读取 editing_mode / mode 作为 strategy_id（渲染 pipeline 的 mode 参数）
    strategy_id = _resolve_strategy_id_from_template(request.template_id, db, user_id)
    use_case = CreateGenerationTaskUseCase(generation_task_repository)
    try:
        task = use_case.execute(
            CreateGenerationTaskCommand(
                project_id="",
                asset_library_id="",
                strategy_id=strategy_id,
                voice_library_id=request.voice_library_id,
                template_id=request.template_id,
                asset_ids=list(request.asset_ids),
                title_ids=list(request.title_ids),
                voice_ids=list(request.voice_ids),
                created_by_user_id=user_id,
                source_edit_plan_id=request.source_edit_plan_id,
                asset_select_mode="",
                batch_id="",
                video_title=request.video_title,
                resolution="",
                bgm_config=request.bgm_config or {},
                auto_retry_enabled=False,
                auto_retry_max=0,
                is_preview=True,
            )
        )
    except ValueError as e:
        logger.warning("[预览生成] 创建失败: %s", e)
        raise HTTPException(status_code=400, detail=str(e)) from e
    except Exception as e:
        logger.error("[预览生成] 创建失败: %s", e, exc_info=True)
        raise HTTPException(status_code=500, detail="创建预览生成任务失败，请稍后再试") from e
    # 入队执行；若入队失败则标记任务为 failed 避免僵尸数据
    try:
        if not safe_enqueue_generation_task(
            task,
            generation_task_repository,
            user_id=user_id,
            log_prefix="[预览生成]",
            log_task_status=True,
        ):
            logger.warning("[预览生成] 任务入队失败: task_id=%s", task.id)
            _mark_task_failed(generation_task_repository, task, "任务入队失败")
            raise HTTPException(status_code=500, detail="任务入队失败，请稍后重试")
    except UserPendingLimitExceeded as e:
        _mark_task_failed(generation_task_repository, task, "待处理任务超限")
        raise HTTPException(
            status_code=429,
            detail=f"您的待处理任务过多（当前 {e.pending_count - 1}/{e.limit}），请等待后再提交",
        ) from None
    except GlobalQueueFull:
        _mark_task_failed(generation_task_repository, task, "系统队列已满")
        raise HTTPException(
            status_code=503,
            detail="系统繁忙，请稍后再试",
        ) from None
    return _to_preview_response(task)
@router.get("/preview/{task_id}", response_model=PreviewGenerationTaskResponse)
 def get_preview_generation_task(
    task_id: str,
    authenticated_user: AuthenticatedUser = Depends(get_current_user),
    generation_task_repository=Depends(get_generation_task_repository),
    generated_video_repository=Depends(get_generated_video_repository),
 ) -> PreviewGenerationTaskResponse:
    """查询预览生成任务状态。
    Args:
        task_id: 任务 ID
    Returns:
        预览任务详情（含状态、进度、结果 URL 等）
    """
    use_case = GetGenerationTaskUseCase(generation_task_repository)
    task = use_case.execute(task_id)
    if task is None:
        raise HTTPException(status_code=404, detail=f"预览任务 {task_id} 不存在")
    # 权限校验：任务必须属于当前用户（统一转 str 比较，避免 UUID/str 类型差异）
    task_user_id = str(getattr(task, "created_by_user_id", "") or "")
    if not task_user_id or task_user_id != str(authenticated_user.user.id):
        raise HTTPException(status_code=403, detail="无权访问该任务")
    # 校验是否为预览任务
    if not getattr(task, "is_preview", False):
        raise HTTPException(status_code=404, detail=f"预览任务 {task_id} 不存在")
    # 查询生成的视频（取第一个）
    generated_videos = []
    status_val = task.status.value if hasattr(task.status, "value") else str(task.status)
    if status_val == "completed":
        list_use_case = ListGeneratedVideosByTaskUseCase(generated_video_repository)
        generated_videos = list_use_case.execute(task_id)
    return _to_preview_response(task, generated_videos=generated_videos)
@@ -26,7 +26,6 @@ from app.schemas.generated_video import (
 )
 from app.schemas.generation_task import (
    BatchGenerationTaskResponse,
    ConfirmGenerationRequest,
    CreateGenerationTaskRequest,
    GenerationTaskResponse,
    ListGenerationTasksResponse,
@@ -39,7 +38,6 @@ from packages.application import (
    GetGenerationTaskUseCase,
    ListGeneratedVideosByTaskUseCase,
 )
 from packages.domain.smart_match import smart_select_assets
 logger = logging.getLogger(__name__)
@@ -63,12 +61,6 @@ def _to_generation_task_response(task) -> GenerationTaskResponse:
        video_title=getattr(task, "video_title", ""),
        resolution=getattr(task, "resolution", ""),
        bgm_config=getattr(task, "bgm_config", {}) or {},
        is_preview=getattr(task, "is_preview", False),
        source_task_id=getattr(task, "source_task_id", ""),
        output_width=getattr(task, "output_width", 1280),
        output_height=getattr(task, "output_height", 720),
        cover_url=getattr(task, "cover_url", ""),
        custom_title=getattr(task, "custom_title", ""),
        logs=getattr(task, "logs", "[]"),
        status=task.status,
        progress=task.progress,
@@ -132,11 +124,19 @@ def _select_assets_from_library(
        return [a.id for a in selected]
    if mode == "smart":
-        # 智能匹配：统一使用 packages/domain/smart_match.py 的多维评分+多样性选取
+        # 智能匹配：按质量分降序 + 时长降序作为tiebreaker
-        # 评分维度：质量分(40%) + 时长适配(30%) + 新鲜度(20%) + 未使用加分(10%)
+        # 注意：这里使用简单的 quality_score 排序保持向后兼容
-        limit = count if count > 0 else None
+        # 更复杂的4维评分+多样性策略由 SmartAssetSelector 服务提供（用于 AI 精选等场景）
-        results = smart_select_assets(ready_video_assets, limit=limit, kind="video")
+        scored_assets = sorted(
-        return [r.asset.id for r in results]
+            ready_video_assets,
            key=lambda a: (
                -(a.quality_score if a.quality_score is not None else 0.0),
                -(getattr(a, "duration", 0.0) or 0.0),
            ),
        )
        if count > 0:
            scored_assets = scored_assets[:count]
        return [a.id for a in scored_assets]
    # 默认 all 模式：返回全部 ready 视频素材
    return [a.id for a in ready_video_assets]
@@ -271,19 +271,13 @@ def create_generation_task(
            detail="系统繁忙，请稍后再试",
        ) from e
    # 画中画已下线：strategy_id 中的 pip/voice_pip 统一映射为 one_take
    effective_strategy_id = request.strategy_id
    if effective_strategy_id in ("pip", "voice_pip"):
        logger.info("画中画已下线，strategy_id %s → one_take", effective_strategy_id)
        effective_strategy_id = "one_take"
    try:
        for _ in range(count):
            task = use_case.execute(
                CreateGenerationTaskCommand(
                    project_id=project_id,
                    asset_library_id=asset_library_id,
-                    strategy_id=effective_strategy_id,
+                    strategy_id=request.strategy_id,
                    voice_library_id=request.voice_library_id,
                    template_id=request.template_id,
                    asset_ids=resolved_asset_ids,
@@ -298,12 +292,6 @@ def create_generation_task(
                    bgm_config=request.bgm_config,
                    auto_retry_enabled=request.auto_retry_enabled,
                    auto_retry_max=request.auto_retry_max,
                    is_preview=request.is_preview,
                    source_task_id=request.source_task_id,
                    output_width=request.output_width,
                    output_height=request.output_height,
                    cover_url=request.cover_url,
                    custom_title=request.custom_title,
                )
            )
            try:
@@ -344,120 +332,6 @@ def create_generation_task(
    return BatchGenerationTaskResponse(items=items, total=len(items))
@router.post("/tasks/{task_id}/confirm", response_model=BatchGenerationTaskResponse)
 def confirm_generation(
    task_id: str,
    request: ConfirmGenerationRequest,
    authenticated_user: AuthenticatedUser = Depends(get_current_user),
    generation_task_repository: Any = Depends(get_generation_task_repository),
    project_repository: Any = Depends(get_project_repository),
 ) -> BatchGenerationTaskResponse:
    """确认生成 -- 复用预览渲染产物（预览与正式品质一致）。
    预览已使用 1080p / CRF 23 / medium 渲染，品质与正式生成一致。
    确认时直接将预览任务标记为正式产出，无需重新渲染，实现秒出。
    仅当预览任务未完成时，才创建新的正式任务走渲染流程。
    """
    # 1. 查找源预览任务
    source_task = generation_task_repository.get(task_id)
    if source_task is None:
        raise HTTPException(status_code=404, detail=f"Preview task {task_id} not found")
    # 2. 权限检查
    if source_task.created_by_user_id and source_task.created_by_user_id != authenticated_user.user.id:
        raise HTTPException(status_code=403, detail="Access denied to this task")
    if source_task.project_id:
        check_project_access(source_task.project_id, authenticated_user.user.id, project_repository)
    # 3. 如果预览任务已完成，检查分辨率一致性后复用产物（秒出）
    if source_task.is_completed and getattr(source_task, "is_preview", False):
        # 校验请求的分辨率是否与预览实际渲染的分辨率一致
        req_w = request.output_width or 0
        req_h = request.output_height or 0
        src_w = getattr(source_task, "output_width", 0) or 0
        src_h = getattr(source_task, "output_height", 0) or 0
        resolution_match = (req_w == 0 or req_w == src_w) and (req_h == 0 or req_h == src_h)
        if resolution_match:
            source_task.mark_confirmed(
                cover_url=request.cover_url,
                custom_title=request.custom_title,
                output_width=request.output_width,
                output_height=request.output_height,
            )
            generation_task_repository.update(source_task)
            logger.info(
                "[确认生成] 复用预览产物: task_id=%s, user_id=%s",
                task_id,
                authenticated_user.user.id,
            )
            return BatchGenerationTaskResponse(
                items=[_to_generation_task_response(source_task)],
                total=1,
            )
        # 分辨率不一致，跳过复用，走新建任务流程
        logger.info(
            "[确认生成] 分辨率不一致，跳过复用: task_id=%s, src=%sx%s, req=%sx%s",
            task_id,
            src_w,
            src_h,
            req_w,
            req_h,
        )
    # 4. 预览任务未完成，创建新的正式任务走渲染流程
    use_case = CreateGenerationTaskUseCase(generation_task_repository)
    new_task = use_case.execute(
        CreateGenerationTaskCommand(
            project_id=source_task.project_id,
            asset_library_id=source_task.asset_library_id,
            strategy_id=source_task.strategy_id,
            voice_library_id=source_task.voice_library_id,
            template_id=source_task.template_id,
            asset_ids=source_task.asset_ids,
            title_ids=source_task.title_ids,
            voice_ids=source_task.voice_ids,
            created_by_user_id=authenticated_user.user.id,
            source_edit_plan_id=source_task.source_edit_plan_id or "",
            asset_select_mode=source_task.asset_select_mode,
            video_title=getattr(source_task, "video_title", ""),
            resolution=getattr(source_task, "resolution", ""),
            is_preview=False,
            source_task_id=task_id,
            output_width=request.output_width,
            output_height=request.output_height,
            cover_url=request.cover_url,
            custom_title=request.custom_title,
        )
    )
    # 5. 调度 worker
    try:
        if not safe_enqueue_generation_task(
            new_task,
            generation_task_repository,
            user_id=authenticated_user.user.id,
            log_prefix="[确认生成]",
            log_task_status=True,
        ):
            logger.warning("[确认生成] 入队失败: task_id=%s", new_task.id)
    except UserPendingLimitExceeded:
        raise HTTPException(
            status_code=429,
            detail="您的待处理任务过多，请等待完成后再提交",
        ) from None
    except GlobalQueueFull:
        raise HTTPException(
            status_code=503,
            detail="系统繁忙，请稍后再试",
        ) from None
    return BatchGenerationTaskResponse(
        items=[_to_generation_task_response(new_task)],
        total=1,
    )
@router.get("/tasks", response_model=ListGenerationTasksResponse)
 def list_generation_tasks(
    authenticated_user: AuthenticatedUser = Depends(get_current_user),
@@ -555,12 +429,6 @@ def retry_generation_task(
            asset_select_mode=getattr(task, "asset_select_mode", ""),
            video_title=getattr(task, "video_title", ""),
            resolution=getattr(task, "resolution", ""),
            is_preview=getattr(task, "is_preview", False),
            source_task_id=getattr(task, "source_task_id", ""),
            output_width=getattr(task, "output_width", 1280),
            output_height=getattr(task, "output_height", 720),
            cover_url=getattr(task, "cover_url", ""),
            custom_title=getattr(task, "custom_title", ""),
        )
    )
    try:
@@ -11,6 +11,7 @@
 - bgm.py:            BGM 管理
 - effects.py:        转场 + 滤镜
 - export.py:         导出配置
 - cover.py:          封面管理 + AI 生成封面
 - subtitles.py:      字幕管理
 - ai_features.py:    AI 推荐
 - generation.py:     生成（触发/进度/记录）
@@ -30,6 +31,7 @@ from .adjustments import router as adjustments_router
 from .ai_features import router as ai_features_router
 from .bgm import router as bgm_router
 from .clips import router as clips_router
 from .cover import router as cover_router
 from .dependencies import get_draft_plan_id, get_editor_services  # noqa: F401
 from .draft import router as draft_router
 from .effects import router as effects_router
@@ -49,6 +51,7 @@ _sub_routers = [
    bgm_router,
    effects_router,
    export_router,
    cover_router,
    subtitles_router,
    ai_features_router,
    generation_router,
@@ -27,14 +27,18 @@ from packages.domain.edit_plan import EditPlanStatus
 logger = logging.getLogger(__name__)
-def _auto_fallback_draft_to_editing(svc: EditPlanService, plan_id: str, plan_check) -> None:
+def _auto_fallback_draft_to_editing(
    svc: EditPlanService, plan_id: str, plan_check
 ) -> None:
    """自动兜底 1: draft → editing"""
    if plan_check.status == EditPlanStatus.DRAFT:
        logger.info("模板编辑器自动兜底: plan=%s draft→editing", plan_id)
        svc.transition_status(plan_id, EditPlanStatus.EDITING)
-def _auto_fallback_copy_template_clips(svc: EditPlanService, plan_id: str, plan_check, db: Session) -> None:
+def _auto_fallback_copy_template_clips(
    svc: EditPlanService, plan_id: str, plan_check, db: Session
 ) -> None:
    """自动兜底 2: 无片段 + 有 template_id → 从模板复制片段配置"""
    existing_clips = svc.count_clips(plan_id)
    if existing_clips == 0 and plan_check.template_id:
@@ -49,15 +53,15 @@ def _auto_fallback_copy_template_clips(svc: EditPlanService, plan_id: str, plan_
            for cfg in configs:
                svc.create_clip(
                    plan_id=plan_id,
-                    clip_type=cfg.clip_type.value if hasattr(cfg.clip_type, "value") else cfg.clip_type,
+                    clip_type=cfg.clip_type.value
                    if hasattr(cfg.clip_type, "value")
                    else cfg.clip_type,
                    order=cfg.order,
                    template_clip_config_id=cfg.id,
                    duration=cfg.default_duration,
-                    transition_effect=(
+                    transition_effect=cfg.transition_effect.value
-                        cfg.transition_effect.value
+                    if hasattr(cfg.transition_effect, "value")
-                        if hasattr(cfg.transition_effect, "value")
+                    else cfg.transition_effect,
                        else cfg.transition_effect
                    ),
                )
            logger.info(
                "模板编辑器自动兜底: plan=%s 从 template_clip_configs 复制了 %d 个片段",
@@ -86,20 +90,14 @@ def _auto_fallback_copy_template_clips(svc: EditPlanService, plan_id: str, plan_
            )
-def _auto_fallback_assign_assets(svc: EditPlanService, plan_id: str, plan_check) -> list:
+def _auto_fallback_assign_assets(
    svc: EditPlanService, plan_id: str, plan_check
 ) -> list:
    """自动兜底 3: 为没有素材的片段分配素材。返回剩余无素材片段列表。"""
    all_clips = svc.list_clips(plan_id)
    clips_without_asset = [c for c in all_clips if not c.asset_id]
    config_asset_ids = (plan_check.config or {}).get("asset_ids", [])
    logger.info(
        "模板编辑器自动兜底3 诊断: plan=%s total_clips=%d " "clips_without_asset=%d config_asset_ids=%r",
        plan_id,
        len(all_clips),
        len(clips_without_asset),
        config_asset_ids[:5] if config_asset_ids else [],
    )
    if clips_without_asset and config_asset_ids:
        logger.info(
            "模板编辑器自动兜底3: plan=%s 为 %d 个无素材片段分配 %d 个指定素材",
@@ -107,42 +105,11 @@ def _auto_fallback_assign_assets(svc: EditPlanService, plan_id: str, plan_check)
            len(clips_without_asset),
            len(config_asset_ids),
        )
        assigned = 0
        for i, clip in enumerate(clips_without_asset):
            asset_idx = i % len(config_asset_ids)
-            try:
+            svc.assign_asset(clip.id, config_asset_ids[asset_idx])
-                svc.assign_asset(clip.id, config_asset_ids[asset_idx])
+        logger.info("模板编辑器自动兜底3: plan=%s 素材分配完成", plan_id)
-                assigned += 1
+        clips_without_asset = []
            except Exception as exc:
                logger.error(
                    "模板编辑器自动兜底3: plan=%s clip=%s 分配素材 %s 失败: %s",
                    plan_id,
                    clip.id,
                    config_asset_ids[asset_idx],
                    exc,
                )
        logger.info(
            "模板编辑器自动兜底3: plan=%s 素材分配完成 assigned=%d/%d",
            plan_id,
            assigned,
            len(clips_without_asset),
        )
        # 重新检查剩余无素材片段
        all_clips_after = svc.list_clips(plan_id)
        clips_without_asset = [c for c in all_clips_after if not c.asset_id]
        if clips_without_asset:
            logger.warning(
                "模板编辑器自动兜底3: plan=%s 仍有 %d 个片段无素材",
                plan_id,
                len(clips_without_asset),
            )
    elif not clips_without_asset:
        logger.info("模板编辑器自动兜底3: plan=%s 所有片段已有素材，跳过", plan_id)
    elif not config_asset_ids:
        logger.info(
            "模板编辑器自动兜底3: plan=%s config.asset_ids 为空，跳过分配",
            plan_id,
        )
    return clips_without_asset
@@ -154,74 +121,43 @@ def _auto_fallback_auto_material_mode(
    clips_without_asset: list,
    asset_library_repo: Any,
    asset_repo: Any,
    user_id: str = "",
 ) -> None:
-    """自动兜底 4: 自动选素材分配给无素材片段
+    """自动兜底 4: 项目有视频素材库时自动选素材"""
    查找策略（按优先级）：
    1. plan 有 project_id → 从项目素材库查找
    2. plan 无 project_id 但有 user_id → 从用户上传的素材中查找
    """
    if not clips_without_asset:
        return
-
+    if not plan_check.project_id:
    ready_videos: list = []
    source_desc = ""
    # 策略 1: 通过 project_id 查找项目素材库
    if plan_check.project_id:
        libs = asset_library_repo.find_by_project(plan_check.project_id)
        video_lib = None
        for lib in libs:
            lib_kind = lib.kind.value if hasattr(lib.kind, "value") else lib.kind
            if lib_kind == "video":
                video_lib = lib
                break
        if video_lib:
            assets = asset_repo.find_by_library(video_lib.id)
            ready_videos = [
                a
                for a in assets
                if (a.status.value if hasattr(a.status, "value") else a.status) == "ready"
                and a.mime_type
                and a.mime_type.startswith("video")
            ]
            source_desc = f"素材库 {video_lib.name}"
    # 策略 2: 通过 user_id 查找用户上传的素材
    if not ready_videos and user_id and hasattr(asset_repo, "find_ready_videos_by_user"):
        logger.info(
            "模板编辑器自动兜底4: plan=%s project_id 为空，尝试通过 user_id=%s 查找素材",
            plan_id,
            user_id,
        )
        ready_videos = asset_repo.find_ready_videos_by_user(user_id)
        source_desc = f"用户上传 (user_id={user_id[:8]}...)"
    if not ready_videos:
        logger.warning(
            "模板编辑器自动兜底4: plan=%s 未找到可用素材 (project_id=%s, user_id=%s)",
            plan_id,
            plan_check.project_id or "(empty)",
            user_id[:8] + "..." if user_id else "(empty)",
        )
        return
    logger.info(
-        "模板编辑器自动兜底4: plan=%s 自动选素材分配给 %d 个无素材片段 (来源: %s, 共 %d 个)",
+        "模板编辑器自动兜底4: plan=%s 自动选素材分配给 %d 个无素材片段",
        plan_id,
        len(clips_without_asset),
        source_desc,
        len(ready_videos),
    )
    random.shuffle(ready_videos)
    for i, clip in enumerate(clips_without_asset):
        asset = ready_videos[i % len(ready_videos)]
        svc.assign_asset(clip.id, asset.id)
    logger.info(
        "模板编辑器自动兜底4: plan=%s 从 %s 分配了 %d 个素材给 %d 个片段",
        plan_id,
        source_desc,
        len(ready_videos),
        len(clips_without_asset),
    )
    libs = asset_library_repo.find_by_project(plan_check.project_id)
    video_lib = None
    for lib in libs:
        lib_kind = lib.kind.value if hasattr(lib.kind, "value") else lib.kind
        if lib_kind == "video":
            video_lib = lib
            break
    if video_lib:
        assets = asset_repo.find_by_library(video_lib.id)
        ready_videos = [
            a
            for a in assets
            if (a.status.value if hasattr(a.status, "value") else a.status) == "ready"
            and a.mime_type
            and a.mime_type.startswith("video")
        ]
        if ready_videos:
            random.shuffle(ready_videos)
            for i, clip in enumerate(clips_without_asset):
                asset = ready_videos[i % len(ready_videos)]
                svc.assign_asset(clip.id, asset.id)
            logger.info(
                "模板编辑器自动兜底4: plan=%s 从素材库 %s 分配了 %d 个素材",
                plan_id,
                video_lib.name,
                len(ready_videos),
            )
@@ -24,94 +24,6 @@ logger = logging.getLogger(__name__)
 router = APIRouter(tags=["Template Editor"])
 def _build_asset_analyses(
    asset_ids: list[str],
    db: Session,
 ) -> dict[str, str]:
    """调用 MediaKit 视频理解，返回 {asset_id: 分析文本}.
    如果 MediaKit 不可用或分析失败，返回空 dict（调用方降级处理）。
    """
    if not asset_ids:
        return {}
    try:
        from packages.adapters.sqlalchemy_impl.asset_repository import SQLAlchemyAssetRepository
        from packages.shared.mediakit_client import get_mediakit_client
        from packages.shared.storage import get_shared_storage_service
        client = get_mediakit_client()
        if not client.is_available:
            logger.info("MediaKit 未配置，跳过视频理解分析")
            return {}
        asset_repo = SQLAlchemyAssetRepository(db)
        storage_svc = get_shared_storage_service()
        # 查找素材并获取下载 URL（使用并行列表保持索引对应，避免 URL 重复导致映射覆盖）
        video_urls: list[str] = []
        valid_asset_ids: list[str] = []
        for aid in asset_ids[:10]:  # MediaKit 单次最多 10 个视频
            asset = asset_repo.get(aid)
            if not asset or not asset.storage_key:
                continue
            # 只处理视频素材
            mime = getattr(asset, "mime_type", "")
            if not mime.startswith("video/"):
                continue
            try:
                url = storage_svc.get_download_url(asset.storage_key)
                if url:
                    video_urls.append(url)
                    valid_asset_ids.append(aid)
            except Exception as e:
                logger.warning("获取素材URL失败: asset_id=%s error=%s", aid, str(e))
        if not video_urls:
            logger.info("无可用视频素材，跳过视频理解分析")
            return {}
        # 调用 MediaKit 视频理解
        prompt = (
            "请简要描述这段视频的主要内容，包括：场景（室内/室外/具体场所）、"
            "主体（人物/物体/动物）、动作/活动、氛围/情绪、主要色调。"
            "控制在100字以内。"
        )
        # 限制轮询参数以适配 API 网关超时（nginx 60s）
        # 视频理解最多 30s（poll_interval=2s * max_poll_attempts=15）
        # 剩余 30s 留给 LLM 调用
        contents = client.analyze_videos(
            video_urls=video_urls,
            prompt=prompt,
            level="Economy",
            poll_interval=2.0,
            max_poll_attempts=15,
        )
        if not contents:
            logger.warning("MediaKit 视频理解未返回结果")
            return {}
        # 将结果映射回 asset_id（通过索引对应）
        analyses: dict[str, str] = {}
        for i, content in enumerate(contents):
            if i < len(valid_asset_ids) and content:
                analyses[valid_asset_ids[i]] = content
        logger.info(
            "MediaKit 视频理解完成: total=%d analyzed=%d",
            len(video_urls),
            len(analyses),
        )
        return analyses
    except Exception as e:
        logger.warning("MediaKit 视频理解异常，将降级到无分析模式: %s", str(e))
        return {}
@router.post("/ai-recommend", response_model=AIRecommendResponse)
 def editor_ai_recommend(
    template_id: str,
@@ -134,16 +46,12 @@ def editor_ai_recommend(
    from packages.shared.ai_service import run_ai_recommend
    # 调用 MediaKit 视频理解，获取素材内容分析
    asset_analyses = _build_asset_analyses(body.asset_ids, db)
    result = run_ai_recommend(
        plan_id=plan_id,
        template_id=plan.template_id,
        asset_ids=body.asset_ids,
        editing_mode=body.editing_mode,
        target_duration=body.target_duration,
        asset_analyses=asset_analyses,
    )
    try:
@@ -0,0 +1,209 @@
 """封面管理路由.
 端点：
 - GET  /cover              封面配置
 - PUT  /cover              更新封面
 - POST /cover/extract      抽帧生成封面
 - POST /cover/smart        智能选帧
 - POST /generate-cover     AI 生成封面
 """
 from __future__ import annotations
 import logging
 from app.auth import AuthenticatedUser, get_current_user
 from app.services.edit_plan_service import EditPlanService
 from app.services.edit_template_service import EditTemplateService
 from fastapi import APIRouter, Depends, HTTPException
 from packages.domain.config_schemas import normalize_plan_config
 from .dependencies import get_draft_plan_id, get_editor_services
 from .schemas import (
    CoverConfigResponse,
    CoverExtractRequest,
    CoverGenerateResponse,
    CoverSmartRequest,
    CoverUpdateRequest,
    GenerateCoverRequest,
    GenerateCoverResponse,
 )
 logger = logging.getLogger(__name__)
 router = APIRouter(tags=["Template Editor"])
@router.get("/cover", response_model=CoverConfigResponse)
 def get_editor_cover(
    template_id: str,
    plan_id: str = Depends(get_draft_plan_id),
    services: tuple[EditTemplateService, EditPlanService] = Depends(get_editor_services),
    _: AuthenticatedUser = Depends(get_current_user),
 ) -> CoverConfigResponse:
    """获取草稿封面配置"""
    _, plan_svc = services
    plan = plan_svc.get_plan_or_raise(plan_id)
    config = plan.config or {}
    cover_config = config.get("cover", {})
    return CoverConfigResponse(
        type=cover_config.get("cover_type", "auto"),
        image_url=cover_config.get("cover_image_url", ""),
        frame_time=cover_config.get("frame_time", 0.0),
    )
@router.put("/cover", response_model=CoverConfigResponse)
 def update_editor_cover(
    template_id: str,
    body: CoverUpdateRequest,
    plan_id: str = Depends(get_draft_plan_id),
    services: tuple[EditTemplateService, EditPlanService] = Depends(get_editor_services),
    _: AuthenticatedUser = Depends(get_current_user),
 ) -> CoverConfigResponse:
    """更新草稿封面配置"""
    _, plan_svc = services
    plan = plan_svc.get_plan_or_raise(plan_id)
    config = dict(plan.config) if plan.config else {}
    current_cover = dict(config.get("cover", {}))
    update_data = body.model_dump(exclude_none=True)
    current_cover.update(update_data)
    config["cover"] = current_cover
    normalized = normalize_plan_config(config)
    plan_svc.update_plan_config(plan_id, {"cover": normalized["cover"]})
    return CoverConfigResponse(
        type=current_cover.get("cover_type", "auto"),
        image_url=current_cover.get("cover_image_url", ""),
        frame_time=current_cover.get("frame_time", 0.0),
    )
@router.post("/cover/extract", response_model=CoverGenerateResponse)
 def extract_editor_cover(
    template_id: str,
    body: CoverExtractRequest,
    plan_id: str = Depends(get_draft_plan_id),
    services: tuple[EditTemplateService, EditPlanService] = Depends(get_editor_services),
    current_user: AuthenticatedUser = Depends(get_current_user),
 ) -> CoverGenerateResponse:
    """从指定片段抽帧生成封面"""
    _, plan_svc = services
    plan = plan_svc.get_plan_or_raise(plan_id)
    clip = plan_svc.get_clip(body.clip_id)
    if not clip or clip.plan_id != plan_id:
        raise HTTPException(status_code=400, detail="片段不存在或不属于当前草稿")
    cover_url = f"cover/extract/{plan_id}_{body.clip_id}_{body.frame_time}.jpg"
    config = dict(plan.config) if plan.config else {}
    cover_config = dict(config.get("cover", {}))
    cover_config.update(
        {
            "cover_type": "extract",
            "cover_image_url": cover_url,
            "clip_id": body.clip_id,
            "frame_time": body.frame_time,
        }
    )
    config["cover"] = cover_config
    normalized = normalize_plan_config(config)
    plan_svc.update_plan_config(plan_id, {"cover": normalized["cover"]})
    logger.info(
        "模板编辑器封面抽帧: template_id=%s plan_id=%s clip_id=%s by user=%s",
        template_id,
        plan_id,
        body.clip_id,
        current_user.user.id,
    )
    return CoverGenerateResponse(
        type="extract",
        image_url=cover_url,
        frame_time=body.frame_time,
    )
@router.post("/cover/smart", response_model=CoverGenerateResponse)
 def smart_editor_cover(
    template_id: str,
    body: CoverSmartRequest,
    plan_id: str = Depends(get_draft_plan_id),
    services: tuple[EditTemplateService, EditPlanService] = Depends(get_editor_services),
    current_user: AuthenticatedUser = Depends(get_current_user),
 ) -> CoverGenerateResponse:
    """智能选帧生成封面"""
    _, plan_svc = services
    plan = plan_svc.get_plan_or_raise(plan_id)
    cover_url = f"cover/smart/{plan_id}_smart.jpg"
    strategy = getattr(body, "strategy", "auto")
    config = dict(plan.config) if plan.config else {}
    cover_config = dict(config.get("cover", {}))
    cover_config.update(
        {
            "cover_type": "smart",
            "cover_image_url": cover_url,
            "strategy": strategy,
        }
    )
    config["cover"] = cover_config
    normalized = normalize_plan_config(config)
    plan_svc.update_plan_config(plan_id, {"cover": normalized["cover"]})
    logger.info(
        "模板编辑器智能封面: template_id=%s plan_id=%s strategy=%s by user=%s",
        template_id,
        plan_id,
        strategy,
        current_user.user.id,
    )
    return CoverGenerateResponse(
        type="smart",
        image_url=cover_url,
        frame_time=None,
    )
@router.post("/generate-cover", response_model=GenerateCoverResponse)
 def editor_generate_cover(
    template_id: str,
    body: GenerateCoverRequest,
    plan_id: str = Depends(get_draft_plan_id),
    services: tuple[EditTemplateService, EditPlanService] = Depends(get_editor_services),
    current_user: AuthenticatedUser = Depends(get_current_user),
 ) -> GenerateCoverResponse:
    """AI 生成封面"""
    _, plan_svc = services
    plan = plan_svc.get_plan_or_raise(plan_id)
    from packages.shared.ai_service import run_generate_cover
    cover_data = run_generate_cover(
        plan_id=plan_id,
        asset_ids=body.asset_ids,
        cover_type=body.cover_type,
        frame_time=body.frame_time,
    )
    current_config = dict(plan.config) if plan.config else {}
    current_config["cover"] = cover_data
    normalized = normalize_plan_config(current_config)
    plan_svc.update_plan_config(plan_id, {"cover": normalized["cover"]})
    logger.info(
        "模板编辑器封面生成: template_id=%s plan_id=%s type=%s by user=%s",
        template_id,
        plan_id,
        body.cover_type,
        current_user.user.id,
    )
    return GenerateCoverResponse(plan_id=plan_id, cover=cover_data)
@@ -18,7 +18,6 @@ from app.dependencies import (
    get_asset_library_repository,
    get_asset_repository,
    get_db_session,
    get_generated_video_repository,
 )
 from app.schemas.generation_task import GenerationTaskResponse
 from app.services.edit_plan_service import EditPlanService
@@ -29,7 +28,6 @@ from sqlalchemy.orm import Session
 from packages.adapters.sqlalchemy_impl.generation_task_repository import (
    SQLAlchemyGenerationTaskRepository,
 )
 from packages.application.generated_videos import ListGeneratedVideosByTaskUseCase
 from packages.application.generation_tasks import (
    CreateGenerationTaskCommand,
    CreateGenerationTaskUseCase,
@@ -73,60 +71,25 @@ def generate_editor_draft(
    _auto_fallback_copy_template_clips(plan_svc, plan_id, plan_check, db)
    clips_without_asset = _auto_fallback_assign_assets(plan_svc, plan_id, plan_check)
    _auto_fallback_auto_material_mode(
-        plan_svc,
+        plan_svc, plan_id, plan_check, clips_without_asset, asset_library_repo, asset_repo
        plan_id,
        plan_check,
        clips_without_asset,
        asset_library_repo,
        asset_repo,
        user_id=str(current_user.user.id),
    )
-    # 检查是否可复用已完成的预览产物（预览品质已与正式一致）
+    # 检查是否可生成
    gen_task_repo = SQLAlchemyGenerationTaskRepository(db)
    reusable_task = _find_reusable_preview_task(gen_task_repo, plan_id, plan_check)
    if reusable_task:
        # 复用预览产物：标记为正式产出，跳过渲染
        reusable_task.mark_confirmed()
        gen_task_repo.update(reusable_task)
        # 将产物 URL 写入 plan config
        rendered_url = _get_task_output_url(reusable_task, gen_task_repo, db)
        plan_svc.update_plan_config(
            plan_id,
            {
                "generation_task_id": reusable_task.id,
                "rendered_storage_key": rendered_url,  # 统一用 rendered_storage_key
            },
        )
        plan_svc.transition_status(plan_id, EditPlanStatus.COMPLETED)
        updated_plan = plan_svc.get_plan_or_raise(plan_id)
        logger.info(
            "模板编辑器复用预览产物: template_id=%s plan_id=%s task_id=%s by user=%s",
            template_id,
            plan_id,
            reusable_task.id,
            current_user.user.id,
        )
        return EditPlanGenerateResponse(
            plan_id=plan_id,
            plan_status=updated_plan.status.value if hasattr(updated_plan.status, "value") else updated_plan.status,
            generation_task_id=reusable_task.id,
            clip_count=len((plan_check.config or {}).get("clips", [])),
        )
    # 检查是否可生成（含最后防线自动修复 + 诊断日志）
    try:
        can_gen, reason = plan_svc.can_generate(plan_id)
    except ValueError as exc:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail=str(exc)) from exc
+        raise HTTPException(
            status_code=status.HTTP_404_NOT_FOUND, detail=str(exc)
        ) from exc
    if not can_gen:
-        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=reason)
+        raise HTTPException(
            status_code=status.HTTP_400_BAD_REQUEST, detail=reason
        )
    try:
        clip_count = plan_svc.mark_clips_ready(plan_id)
        gen_task_repo = SQLAlchemyGenerationTaskRepository(db)
        user_id = current_user.user.id
        _check_queue_limits(gen_task_repo, user_id)
@@ -160,7 +123,9 @@ def generate_editor_draft(
        return EditPlanGenerateResponse(
            plan_id=plan_id,
-            plan_status=updated_plan.status.value if hasattr(updated_plan.status, "value") else updated_plan.status,
+            plan_status=updated_plan.status.value
            if hasattr(updated_plan.status, "value")
            else updated_plan.status,
            generation_task_id=gen_task.id,
            clip_count=clip_count,
        )
@@ -182,58 +147,6 @@ def generate_editor_draft(
        ) from _e
 def _find_reusable_preview_task(gen_task_repo, plan_id: str, plan) -> "object | None":
    """查找该 plan 关联的已完成预览任务，判断是否可复用。
    复用条件：
    1. 存在 source_edit_plan_id == plan_id 的已完成预览任务
    2. plan 在预览完成后未被修改（updated_at <= 预览完成时间）
    Returns:
        可复用的 GenerationTask，或 None
    """
    try:
        tasks = gen_task_repo.list_by_source_edit_plan(plan_id)
    except Exception:
        return None
    for task in tasks:
        if not getattr(task, "is_preview", False):
            continue
        if not task.is_completed:
            continue
        # 检查 plan 是否在预览完成后被修改
        completed_at = getattr(task, "completed_at", None)
        if completed_at and hasattr(plan, "updated_at"):
            plan_updated = plan.updated_at
            # 如果 plan.updated_at 为空，无法判断是否修改过，跳过
            if plan_updated is None:
                continue
            # 如果 plan 在预览完成后又被修改了，不能复用
            if plan_updated > completed_at:
                continue
        return task
    return None
 def _get_task_output_url(task, gen_task_repo, db) -> str:
    """获取任务的输出视频 URL。"""
    try:
        video_repo = get_generated_video_repository(db)
        use_case = ListGeneratedVideosByTaskUseCase(video_repo)
        videos = use_case.execute(task.id)
        if videos:
            url = getattr(videos[0], "file_url", "") or ""
            # 规范化：合并路径中的双斜杠（保留协议头 ://）
            if url:
                import re as _re
                url = _re.sub(r"(?<!:)//", "/", url)
            return url
    except Exception:
        pass
    return ""
@router.get("/generation-status", response_model=EditPlanGenerationStatusResponse)
 def get_editor_generation_status(
    template_id: str,
@@ -247,7 +160,9 @@ def get_editor_generation_status(
    try:
        gen_status = plan_svc.get_generation_status(plan_id)
    except ValueError as exc:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail=str(exc)) from exc
+        raise HTTPException(
            status_code=status.HTTP_404_NOT_FOUND, detail=str(exc)
        ) from exc
    plan = gen_status["plan"]
    clips = gen_status["clips"]
@@ -265,22 +180,25 @@ def get_editor_generation_status(
        for c in clips
    ]
-    raw_video_url = (plan.config or {}).get("rendered_storage_key", "") or (plan.config or {}).get("rendered_url", "")
+    raw_video_url = (plan.config or {}).get("rendered_url", "")
    video_url = ""
    if raw_video_url:
-        if raw_video_url.startswith("http"):
+        try:
-            video_url = raw_video_url  # 已经是完整 URL
+            video_url = storage_service.get_download_url(
-        else:
+                raw_video_url, expires_seconds=86400
-            try:
+            )
-                video_url = storage_service.get_url(raw_video_url)  # storage_key -> 完整 URL
+        except Exception as e:
-            except Exception as e:
+            logger.warning(
-                logger.warning("生成视频URL获取失败: template_id=%s error=%s", template_id, e)
+                "生成视频签名URL失败: template_id=%s error=%s", template_id, e
-                video_url = raw_video_url
+            )
            video_url = raw_video_url
    progress = gen_status.get("progress", 0.0)
    error_message = gen_status.get("error_message", "")
    gen_task_status = gen_status.get("generation_task_status")
-    plan_status_val = plan.status.value if hasattr(plan.status, "value") else plan.status
+    plan_status_val = (
        plan.status.value if hasattr(plan.status, "value") else plan.status
    )
    if plan_status_val == "completed" and progress < 100:
        progress = 100.0
@@ -99,6 +99,29 @@ class AIRecommendResponse(BaseModel):
    confidence: float = Field(..., ge=0.0, le=1.0, description="AI 推荐置信度 (0~1)")
 # ── 封面生成 ────────────────────────────────────────────────────────────────
 class GenerateCoverRequest(BaseModel):
    """AI 封面生成请求体"""
    asset_ids: List[str] = Field(default_factory=list, description="素材 ID 列表（确定视频来源）")
    cover_type: str = Field(
        default="ai_frame",
        description="封面类型: ai_frame / manual / upload / ai_regenerate",
    )
    frame_time: Optional[float] = Field(
        default=None,
        ge=0.0,
        description="手动选帧时间点（秒），仅 cover_type=manual 时有效",
    )
 class GenerateCoverResponse(BaseModel):
    """AI 封面生成响应体"""
    plan_id: str = Field(..., description="剪辑计划 ID")
    cover: dict[str, Any] = Field(..., description="封面数据（type / image_url / frame_time 等）")
 # ── BGM ────────────────────────────────────────────────────────────────────
@@ -234,6 +257,43 @@ class ClipsFromAssetsResponse(BaseModel):
 # ── 封面配置 ────────────────────────────────────────────────────────────────
 class CoverConfigResponse(BaseModel):
    """封面配置响应"""
    type: str = Field(..., description="封面类型: ai_frame / manual / upload")
    image_url: str = Field(default="", description="封面图片 URL")
    frame_time: Optional[float] = Field(default=None, description="抽帧时间点（秒）")
 class CoverUpdateRequest(BaseModel):
    """更新封面配置请求"""
    type: Optional[str] = Field(default=None, description="封面类型")
    image_url: Optional[str] = Field(default=None, description="封面图片 URL")
    frame_time: Optional[float] = Field(default=None, ge=0.0, description="抽帧时间点（秒）")
 class CoverExtractRequest(BaseModel):
    """从片段抽帧生成封面请求"""
    clip_id: str = Field(..., description="片段 ID")
    frame_time: float = Field(1.0, ge=0.0, description="抽帧时间点（秒）")
 class CoverSmartRequest(BaseModel):
    """智能选帧请求"""
    clip_id: Optional[str] = Field(default=None, description="指定片段 ID（不传则用第一个视频片段）")
 class CoverGenerateResponse(BaseModel):
    """封面生成响应"""
    type: str = Field(..., description="封面类型")
    image_url: str = Field(..., description="封面图片 URL")
    frame_time: Optional[float] = Field(default=None, description="抽帧时间点（秒）")
 # ── 导出配置 ────────────────────────────────────────────────────────────────
@@ -11,7 +11,6 @@ from app.dependencies import get_cosyvoice_service, get_voice_clone_profile_repo
 from app.schemas.voice_clone import (
    CreateVoiceCloneRequest,
    ListVoiceCloneResponse,
    VoiceClonePreviewResponse,
    VoiceCloneProfileResponse,
    VoiceCloneStatusResponse,
 )
@@ -20,7 +19,7 @@ from fastapi import APIRouter, Depends, HTTPException, Query, Response, status
 from packages.adapters.sqlalchemy_impl.voice_clone_profile_repository import (
    SQLAlchemyVoiceCloneProfileRepository,
 )
-from packages.application.cosyvoice_service import CosyVoiceError, CosyVoiceService
+from packages.application.cosyvoice_service import CosyVoiceService
 from packages.application.voice_clone.use_cases import (
    DeleteVoiceCloneUseCase,
    GetVoiceCloneStatusUseCase,
@@ -37,13 +36,6 @@ logger = logging.getLogger(__name__)
 router = APIRouter()
 # 克隆音色试听缓存（减少重复TTS调用）
 # key: clone_id, value: (audio_url, duration, file_size, text, timestamp)
 _clone_preview_cache: dict[str, tuple[str, float, int, str, float]] = {}
 CLONE_PREVIEW_CACHE_TTL = 7 * 24 * 3600  # 7天TTL
 # 默认试听文本
 CLONE_PREVIEW_TEMPLATE = "你好，这是我的克隆音色，很高兴能为你配音。"
 def _to_response(profile) -> VoiceCloneProfileResponse:
    # source_audio_url 是用户传入的原始 URL（可能是外部地址），不做预签名转换
@@ -231,78 +223,3 @@ def retry_voice_clone(
                logger.error(f"Failed to mark profile as failed after dispatch error: {inner_e}")
    return _to_response(profile)
@router.get("/{clone_id}/preview", response_model=VoiceClonePreviewResponse)
 def get_voice_clone_preview(
    clone_id: str,
    text: str = Query("", description="自定义试听文本，为空则使用默认示例"),
    authenticated_user: AuthenticatedUser = Depends(get_current_user),
    repository: SQLAlchemyVoiceCloneProfileRepository = Depends(get_voice_clone_profile_repository),
    cosyvoice: CosyVoiceService = Depends(get_cosyvoice_service),
 ) -> VoiceClonePreviewResponse:
    """获取克隆音色试听音频（实时 TTS 合成）。
    - 克隆音色必须处于 ready 状态
    - 使用默认试听文本时，结果缓存 7 天
    - 可传入自定义 text 参数试听不同文本
    """
    import time
    use_case = GetVoiceCloneUseCase(repository)
    try:
        profile = use_case.execute(clone_id, authenticated_user.user.id)
    except VoiceCloneNotFoundError as _e:
        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="Voice clone not found") from _e
    if not profile.is_ready:
        raise HTTPException(
            status_code=status.HTTP_400_BAD_REQUEST,
            detail=f"Voice clone is not ready (current status: {profile.status})",
        )
    # 有自定义文本时不缓存
    use_cache = not text.strip()
    if use_cache and clone_id in _clone_preview_cache:
        audio_url, duration, file_size, cached_text, cached_at = _clone_preview_cache[clone_id]
        if time.time() - cached_at < CLONE_PREVIEW_CACHE_TTL:
            return VoiceClonePreviewResponse(
                clone_id=clone_id,
                voice_id=profile.voice_id,
                audio_url=audio_url,
                text=cached_text,
                duration=duration,
                file_size=file_size,
            )
    # 合成试听音频
    preview_text = text.strip() or CLONE_PREVIEW_TEMPLATE
    try:
        result = cosyvoice.synthesize_speech(
            text=preview_text,
            voice_id=profile.voice_id,
            format="mp3",
            speed=1.0,
        )
    except CosyVoiceError as e:
        raise HTTPException(status_code=502, detail=f"TTS 合成失败: {e}") from e
    # 缓存（仅默认试听文本）
    if use_cache:
        _clone_preview_cache[clone_id] = (
            result.audio_url,
            result.duration,
            result.file_size,
            preview_text,
            time.time(),
        )
    return VoiceClonePreviewResponse(
        clone_id=clone_id,
        voice_id=profile.voice_id,
        audio_url=result.audio_url,
        text=preview_text,
        duration=result.duration,
        file_size=result.file_size,
    )
@@ -5,8 +5,6 @@
 from __future__ import annotations
 import logging
 import time
 from typing import Literal, Optional
 from app.api.routes._helpers import get_user_plan
@@ -44,7 +42,6 @@ from packages.domain.preset_voices import PRESET_VOICES, get_preset_voice_by_id
 from packages.ports.user_repository import UserRepository
 router = APIRouter()
 logger = logging.getLogger(__name__)
 # 预置音色试听音频缓存（内存缓存，减少重复TTS调用）
 # key: voice_id, value: (audio_url, timestamp)
@@ -54,65 +51,6 @@ PREVIEW_CACHE_TTL = 7 * 24 * 3600  # 7天TTL
 PREVIEW_TEMPLATE = "你好，我是{name}，很高兴认识你。"
 def _resolve_preset_preview_url(
    voice_id: str,
    fallback_url: str,
    cosyvoice: CosyVoiceService,
 ) -> str:
    """为预置音色获取有效的 preview_url.
    优先从内存缓存读取；缓存失效时调用 CosyVoice 重新合成；
    合成失败时降级返回硬编码 URL（可能已过期，但不会报错）。
    """
    # 检查缓存
    if voice_id in _preset_preview_cache:
        audio_url, cached_at = _preset_preview_cache[voice_id]
        if time.time() - cached_at < PREVIEW_CACHE_TTL:
            return audio_url
    # 缓存失效，调用 CosyVoice 合成
    preset = get_preset_voice_by_id(voice_id)
    if preset is None:
        return fallback_url
    preview_text = PREVIEW_TEMPLATE.format(name=preset.name)
    try:
        result = cosyvoice.synthesize_speech(
            text=preview_text,
            voice_id=voice_id,
            format="mp3",
            speed=1.0,
        )
        audio_url = result.audio_url
        _preset_preview_cache[voice_id] = (audio_url, time.time())
        logger.info("Preset voice preview generated: %s", voice_id)
        return audio_url
    except Exception as e:
        logger.warning("Failed to generate preview for %s, using fallback: %s", voice_id, e)
        return fallback_url
 def _resolve_all_preset_preview_urls(
    presets: list,
    cosyvoice: CosyVoiceService,
 ) -> dict[str, str]:
    """顺序解析所有预置音色的 preview_url.
    采用顺序调用（而非并行）以避免触发 DashScope API 速率限制。
    首次调用后结果缓存 7 天，后续请求直接命中缓存。
    Returns:
        voice_id -> preview_url 映射
    """
    result_map: dict[str, str] = {}
    for p in presets:
        try:
            result_map[p.voice_id] = _resolve_preset_preview_url(p.voice_id, p.preview_url, cosyvoice)
        except Exception:
            result_map[p.voice_id] = p.preview_url
    return result_map
 def _get_voice_repository(session: Session = Depends(get_db_session)) -> SQLAlchemyVoiceLibraryRepository:
    return SQLAlchemyVoiceLibraryRepository(session)
@@ -180,16 +118,8 @@ def _to_unified_response(item, profile_id_map: dict | None = None, sign_url=None
    )
-def _preset_to_unified_response(preset, preview_url_map: dict[str, str] | None = None) -> UnifiedVoiceItemResponse:
+def _preset_to_unified_response(preset) -> UnifiedVoiceItemResponse:
-    """将预置音色转换为统一响应格式。
+    """将预置音色转换为统一响应格式。"""
    Args:
        preset: 预置音色对象
        preview_url_map: voice_id -> preview_url 动态映射，优先使用
    """
    preview_url = preset.preview_url
    if preview_url_map and preset.voice_id in preview_url_map:
        preview_url = preview_url_map[preset.voice_id]
    return UnifiedVoiceItemResponse(
        id=preset.voice_id,
        type="preset",
@@ -199,40 +129,11 @@ def _preset_to_unified_response(preset, preview_url_map: dict[str, str] | None =
        language=preset.language,
        voice_id=preset.voice_id,
        voice_provider="cosyvoice",
-        preview_url=preview_url,
+        preview_url=preset.preview_url,
        tags=preset.tags or [],
    )
 def _clone_profile_to_unified_response(profile) -> UnifiedVoiceItemResponse:
    """将克隆音色档案转换为统一响应格式。
    注意：克隆音色是「音色模型」（可用于 TTS 合成任意文本），
    不同于配音库条目（具体的配音作品）。
    """
    return UnifiedVoiceItemResponse(
        id=profile.id,
        type="clone",
        name=profile.name,
        description=profile.description or "",
        gender=profile.gender or "unknown",
        language=profile.language or "zh-CN",
        voice_id=profile.voice_id or "",
        voice_provider=profile.voice_model or "cosyvoice",
        audio_url="",  # 克隆音色没有预合成音频，需通过 /voice-clones/{id}/preview 试听
        preview_url="",  # 试听需实时合成，前端调用 preview 接口
        duration=0,
        file_size=0,
        status=profile.status.value if hasattr(profile.status, "value") else str(profile.status),
        tags=[],
        user_id=profile.user_id,
        project_id=None,
        voice_clone_profile_id=profile.id,
        created_at=profile.created_at,
        updated_at=profile.updated_at,
    )
 # ==================== 统一配音列表（预置 + 克隆）====================
@@ -249,7 +150,6 @@ def list_voices_unified(
    voice_repository: SQLAlchemyVoiceLibraryRepository = Depends(_get_voice_repository),
    clone_profile_repository: SQLAlchemyVoiceCloneProfileRepository = Depends(_get_clone_profile_repository),
    sign_url=Depends(get_audio_url_signer),
    cosyvoice: CosyVoiceService = Depends(get_cosyvoice_service),
 ) -> UnifiedVoiceListResponse:
    """获取配音列表（预置音色 + 用户克隆音色）。
@@ -265,30 +165,19 @@ def list_voices_unified(
    has_preset = type is None or type == "preset"
    has_clone = type is None or type == "clone"
-    # 获取预置音色（动态生成 preview_url）
+    # 获取预置音色
    if has_preset:
-        preview_url_map = _resolve_all_preset_preview_urls(PRESET_VOICES, cosyvoice)
+        preset_items = [_preset_to_unified_response(p) for p in PRESET_VOICES]
        preset_items = [_preset_to_unified_response(p, preview_url_map) for p in PRESET_VOICES]
        preset_count = len(preset_items)
-    # 获取克隆音色（从 voice_clone_profile 读取，ready 状态的克隆音色）
+    # 获取克隆音色
    if has_clone:
-        # status_filter 映射：不传则默认只返回 ready 状态（可用的克隆音色）
+        use_case = ListVoiceLibraryUseCase(voice_repository)
-        # 前端可以传 status=all 获取所有状态，或传具体状态过滤
+        clone_items_raw, clone_count = use_case.execute(user_id, status=status_filter, skip=skip, limit=limit)
-        filter_status = None
+        # 批量查询 voice_id → profile_id 映射，填充 voice_clone_profile_id
-        if status_filter and status_filter != "all":
+        voice_ids = [i.voice_id for i in clone_items_raw if i.voice_id]
-            filter_status = status_filter
+        profile_id_map = clone_profile_repository.find_profile_ids_by_voice_ids(voice_ids) if voice_ids else {}
-        elif not status_filter:
+        clone_items = [_to_unified_response(i, profile_id_map, sign_url) for i in clone_items_raw]
            filter_status = "ready"
        clone_profiles = clone_profile_repository.list_by_user(
            user_id,
            status=filter_status,
            limit=limit,
            offset=skip,
        )
        clone_count = clone_profile_repository.count_by_user(user_id, status=filter_status)
        clone_items = [_clone_profile_to_unified_response(p) for p in clone_profiles]
    # 组装结果
    if type == "preset":
@@ -315,15 +204,11 @@ def list_voices_unified(
@router.get("/presets", response_model=PresetVoiceListResponse)
-def list_preset_voices(
+def list_preset_voices() -> PresetVoiceListResponse:
    cosyvoice: CosyVoiceService = Depends(get_cosyvoice_service),
 ) -> PresetVoiceListResponse:
    """获取预置音色列表。
    不需要认证，返回所有系统预置的 CosyVoice 音色。
    preview_url 通过 CosyVoice 动态生成，不依赖硬编码的过期 URL。
    """
    preview_url_map = _resolve_all_preset_preview_urls(PRESET_VOICES, cosyvoice)
    items = [
        PresetVoiceItemResponse(
            voice_id=p.voice_id,
@@ -331,7 +216,7 @@ def list_preset_voices(
            description=p.description,
            gender=p.gender,
            language=p.language,
-            preview_url=preview_url_map.get(p.voice_id, p.preview_url),
+            preview_url=p.preview_url,
            tags=p.tags or [],
        )
        for p in PRESET_VOICES
@@ -351,6 +236,8 @@ def get_preset_voice_preview(
    - 相同 voice_id 重复调用直接返回缓存的音频URL
    - 可传入自定义 text 参数试听不同文本
    """
    import time
    preset = get_preset_voice_by_id(voice_id)
    if preset is None:
        raise HTTPException(status_code=404, detail=f"预置音色不存在: {voice_id}")
@@ -22,9 +22,6 @@ from packages.adapters.sqlalchemy_impl.asset_repository import SQLAlchemyAssetRe
 from packages.adapters.sqlalchemy_impl.classification_job_repository import (
    SQLAlchemyClassificationJobRepository,
 )
 from packages.adapters.sqlalchemy_impl.cover_template_repository import (
    SQLAlchemyCoverTemplateRepository,
 )
 from packages.adapters.sqlalchemy_impl.duplication_repository import (
    SQLAlchemyDuplicationRecordRepository,
 )
@@ -131,13 +128,6 @@ def get_project_repository(
    return SQLAlchemyProjectRepository(session)
 def get_cover_template_repository(
    session: Session = Depends(get_db_session),
 ) -> SQLAlchemyCoverTemplateRepository:
    """Provide the SQLAlchemy cover template repository implementation."""
    return SQLAlchemyCoverTemplateRepository(session)
 def get_tag_repository(
    session: Session = Depends(get_db_session),
 ) -> TagRepository:
@@ -2,7 +2,7 @@ from pydantic import BaseModel, Field
 class CreateAssetRequest(BaseModel):
-    project_id: str | None = Field(default=None, description="可选，不传时从 library.project_id 自动推导")
+    project_id: str = Field(..., min_length=1)
    library_id: str = Field(..., min_length=1)
    name: str = Field(..., min_length=1, max_length=100)
    storage_key: str = Field(..., min_length=1, max_length=255)
@@ -101,30 +101,3 @@ class ListAssetsResponse(BaseModel):
    total: int = Field(default=0, ge=0)
    skip: int = Field(default=0, ge=0)
    limit: int = Field(default=100, ge=1)
 class SmartMatchRequest(BaseModel):
    """智能选素材请求。"""
    library_id: str = Field(..., min_length=1, description="素材库 ID")
    limit: int | None = Field(default=None, ge=1, le=200, description="最大返回数量，不传则返回全部匹配素材")
    kind: str | None = Field(
        default=None,
        pattern="^(video|image|audio)$",
        description="按文件类型过滤，不传则返回所有类型",
    )
 class SmartMatchItem(BaseModel):
    """智能选素材结果条目。"""
    asset: AssetResponse
    score: float = Field(..., ge=0, le=100, description="综合得分 0-100")
    breakdown: dict[str, float] = Field(default_factory=dict, description="各维度得分明细")
 class SmartMatchResponse(BaseModel):
    """智能选素材响应。"""
    items: list[SmartMatchItem]
    total_candidates: int = Field(default=0, ge=0, description="参与评分的候选素材总数")
@@ -1,54 +0,0 @@
 """封面模板 Schema。"""
 from datetime import datetime
 from typing import Any
 from pydantic import BaseModel, Field
 class CoverTemplateConfig(BaseModel):
    """封面模板配置。"""
    background_enabled: bool = Field(default=True, description="是否启用背景")
    background_color: str = Field(default="#000000", description="背景颜色")
    portrait_enabled: bool = Field(default=True, description="是否显示人像")
    title_text: str = Field(default="", description="主标题文字")
    subtitle_text: str = Field(default="", description="副标题文字")
    mask_enabled: bool = Field(default=False, description="是否启用蒙版")
 class CreateCoverTemplateRequest(BaseModel):
    """创建封面模板请求。"""
    name: str = Field(..., min_length=1, max_length=200, description="模板名称")
    thumbnail_url: str = Field(default="", description="缩略图 URL")
    config: CoverTemplateConfig | None = Field(default=None, description="模板配置")
 class UpdateCoverTemplateRequest(BaseModel):
    """更新封面模板请求。"""
    name: str | None = Field(default=None, min_length=1, max_length=200, description="模板名称")
    thumbnail_url: str | None = Field(default=None, description="缩略图 URL")
    config: CoverTemplateConfig | None = Field(default=None, description="模板配置")
 class CoverTemplateResponse(BaseModel):
    """封面模板响应。"""
    id: str
    name: str
    thumbnail_url: str
    is_system: bool
    created_at: datetime
    config: dict[str, Any] = Field(default_factory=dict)
    class Config:
        from_attributes = True
 class ListCoverTemplatesResponse(BaseModel):
    """封面模板列表响应。"""
    items: list[CoverTemplateResponse]
    total: int = Field(default=0, ge=0)
@@ -1,18 +1,8 @@
 import json
 from datetime import datetime
 from pydantic import BaseModel, Field, field_validator, model_validator
 class ConfirmGenerationRequest(BaseModel):
    """确认生成请求体 — 基于预览任务创建正式生成任务"""
    output_width: int = Field(default=1080, description="输出视频宽度")
    output_height: int = Field(default=1920, description="输出视频高度")
    cover_url: str = Field(default="", description="自定义封面图片 URL")
    custom_title: str = Field(default="", description="自定义视频标题")
 class CreateGenerationTaskRequest(BaseModel):
    """创建生成任务请求。
@@ -66,13 +56,6 @@ class CreateGenerationTaskRequest(BaseModel):
        default_factory=dict,
        description="自定义BGM配置，覆盖模板BGM设置。支持 enabled/source/asset_id/preset_id/audio_url/volume 等字段",
    )
    # ── 预览 / 确认生成 ──
    is_preview: bool = Field(default=False, description="是否为预览任务")
    source_task_id: str = Field(default="", description="来源预览任务 ID（确认生成时传入）")
    output_width: int = Field(default=1280, description="输出视频宽度")
    output_height: int = Field(default=720, description="输出视频高度")
    cover_url: str = Field(default="", description="封面图片 URL")
    custom_title: str = Field(default="", description="自定义视频标题")
    @model_validator(mode="after")
    def _check_at_least_one_mode(self) -> "CreateGenerationTaskRequest":
@@ -103,12 +86,6 @@ class GenerationTaskResponse(BaseModel):
    video_title: str = ""
    resolution: str = ""
    bgm_config: dict = Field(default_factory=dict)
    is_preview: bool = False
    source_task_id: str = ""
    output_width: int = 1280
    output_height: int = 720
    cover_url: str = ""
    custom_title: str = ""
    status: str
    progress: float
    result_count: int
@@ -145,75 +122,3 @@ class ListGenerationTasksResponse(BaseModel):
    """用户级生成任务列表响应（跨 project）。"""
    items: list[GenerationTaskResponse]
 # ── 预览生成（Phase 1） ───────────────────────────────────────────────────────
 class CreatePreviewGenerationTaskRequest(BaseModel):
    """创建预览生成任务请求。
    仅支持模板模式：template_id + asset_ids 等素材 ID 列表。
    预览渲染品质与正式生成一致（1080p, CRF 23, medium preset）。
    """
    template_id: str
    asset_ids: list[str] = Field(default_factory=list)
    title_ids: list[str] = Field(default_factory=list)
    voice_ids: list[str] = Field(default_factory=list)
    voice_library_id: str = Field(
        default="", description="配音素材库ID（用户上传的音频或AI配音），对应配音选择页面选择的配音素材"
    )
    video_title: str = Field(default="", description="生成视频的标题/名称，为空则使用默认命名")
    duration: float = Field(default=0.0, ge=0, description="期望视频时长（秒），0 表示由模板决定")
    video_ratio: str = Field(default="", description="视频比例，如 16:9 / 9:16，为空使用模板默认")
    bgm_config: dict = Field(
        default_factory=dict,
        description="自定义BGM配置，覆盖模板BGM设置。支持 enabled/source/asset_id/preset_id/audio_url/volume 等字段",
    )
    preview_count: int = Field(
        default=1,
        ge=1,
        le=10,
        description="预览视频生成数量，范围 1-10，默认 1",
    )
    source_edit_plan_id: str = Field(
        default="",
        description="关联的编辑计划ID（可选），用于确认生成时复用预览产物",
    )
    @model_validator(mode="after")
    def _check_template_id(self) -> "CreatePreviewGenerationTaskRequest":
        if not self.template_id.strip():
            raise ValueError("template_id 不能为空")
        return self
    @model_validator(mode="after")
    def _check_asset_ids(self) -> "CreatePreviewGenerationTaskRequest":
        if not self.asset_ids and not self.title_ids and not self.voice_ids:
            raise ValueError("asset_ids/title_ids/voice_ids 至少需要提供一个")
        return self
 class PreviewGenerationTaskResponse(BaseModel):
    """预览生成任务响应。
    包含任务状态、进度、分辨率、生成结果 URL 等关键字段。
    """
    task_id: str
    status: str
    progress: float
    is_preview: bool = True
    resolution: str = ""
    video_url: str = ""
    duration: float = 0.0
    file_size: int = 0
    clip_count: int = 0
    transition_count: int = 0
    material_usage: dict = Field(default_factory=dict)
    error_message: str = ""
    created_at: datetime | None = None
    started_at: datetime | None = None
    finished_at: datetime | None = None
    generate_duration: float = 0.0
@@ -63,25 +63,3 @@ class ListVoiceCloneResponse(BaseModel):
    items: List[VoiceCloneProfileResponse]
    total: int
 class VoiceClonePreviewResponse(BaseModel):
    """克隆音色试听响应。"""
    clone_id: str
    """音色克隆档案 ID"""
    voice_id: str
    """CosyVoice 音色 ID"""
    audio_url: str
    """试听音频 URL"""
    text: str
    """试听文本"""
    duration: float = 0.0
    """音频时长（秒）"""
    file_size: int = 0
    """文件大小（字节）"""
@@ -12,13 +12,12 @@
 from __future__ import annotations
 import json
 import logging
 import math
 import random
 from typing import Any, Dict, List, Optional
 from packages.domain.ai_parsing import generate_titles_fallback as _generate_titles_fallback_base
 from packages.domain.ai_parsing import keyword_match_fallback as _semantic_match_fallback_base
 from packages.domain.ai_parsing import parse_semantic_match_response as _parse_semantic_match_base
 from packages.domain.ai_parsing import parse_titles_from_response as _parse_titles_from_response
 from packages.shared.ai_client import get_doubao_client
 logger = logging.getLogger(__name__)
@@ -65,9 +64,85 @@ def _generate_titles_fallback(
    style: str = "viral",
    count: int = 5,
 ) -> List[str]:
-    """本地降级：基于模板规则生成标题（薄包装，转发到 ai_parsing 模块）."""
+    """本地降级：基于模板规则生成标题.
    当豆包 API 不可用或调用失败时使用，保证接口始终有返回。
    """
    style_info = TITLE_STYLES.get(style, TITLE_STYLES["viral"])
-    return _generate_titles_fallback_base(description, style_info, count)
+    examples = style_info["examples"]
    # 从描述中提取关键词（取前几个词）
    keywords = [w for w in description.strip().split() if len(w) > 1][:3]
    keyword = keywords[0] if keywords else "精彩内容"
    # 基于模板生成
    templates = [
        f"「{keyword}」{examples[0][:10]}...",
        f"{keyword}：{examples[1]}",
        f"关于{keyword}，你不知道的3件事",
        f"{keyword}入门指南，新手必看",
        f"深度解析：{keyword}背后的秘密",
        f"{keyword}怎么做？手把手教你",
        f"干货分享 | {keyword}全攻略",
        f"建议收藏：{keyword}实用技巧",
        f"{keyword}避坑指南，别再踩雷了",
        f"一分钟搞懂{keyword}",
    ]
    random.shuffle(templates)
    return templates[: min(count, len(templates))]
 def _parse_titles_from_response(content: str) -> List[str]:
    """从模型返回中解析标题列表.
    支持多种返回格式：
    - JSON 数组: ["标题1", "标题2"]
    - 编号列表: 1. 标题1 / 2. 标题2
    - 换行分隔: 标题1\n标题2
    - 带破折号: - 标题1
    """
    if not content:
        return []
    # 尝试解析 JSON
    try:
        # 清理可能的 markdown 代码块标记
        cleaned = content.strip()
        if cleaned.startswith("```"):
            cleaned = cleaned.strip("`")
            if cleaned.lower().startswith("json"):
                cleaned = cleaned[4:]
            cleaned = cleaned.strip()
        data = json.loads(cleaned)
        if isinstance(data, list):
            return [str(item).strip() for item in data if str(item).strip()]
        if isinstance(data, dict) and "titles" in data:
            titles = data["titles"]
            if isinstance(titles, list):
                return [str(t).strip() for t in titles if str(t).strip()]
    except (json.JSONDecodeError, ValueError):
        pass
    # 尝试按行解析
    titles: List[str] = []
    for line in content.strip().split("\n"):
        line = line.strip()
        if not line:
            continue
        # 去掉编号前缀 "1. " "1、" "（1）"
        import re
        line = re.sub(r"^[\d]+[\.、\)）]\s*", "", line)
        # 去掉破折号前缀 "- " "• "
        line = re.sub(r"^[-•·]\s*", "", line)
        # 去掉引号
        line = line.strip('"').strip("'").strip("「」")
        if line and len(line) < 100:  # 过滤过长的行
            titles.append(line)
    return titles
 def generate_smart_titles(
@@ -166,19 +241,132 @@ def _semantic_match_fallback(
    description: str,
    assets: List[Dict[str, Any]],
 ) -> List[Dict[str, Any]]:
-    """本地降级：基于关键词的简单匹配（薄包装，转发到 ai_parsing 模块）."""
+    """本地降级：基于关键词的简单匹配.
-    return _semantic_match_fallback_base(description, assets)
+
    计算描述中的关键词与素材名称/标签/描述的重叠度，
    作为匹配度评分。0-1分。
    """
    import re
    # 提取关键词（中文按2字以上片段，英文按单词）
    desc = description.lower()
    # 简单分词：提取2字以上的中文字符串和英文单词
    keywords = set()
    # 英文单词
    for word in re.findall(r"[a-zA-Z]{3,}", desc):
        keywords.add(word)
    # 中文2-4字片段
    for i in range(len(desc)):
        for j in range(i + 2, min(i + 5, len(desc) + 1)):
            fragment = desc[i:j]
            if all("\u4e00" <= c <= "\u9fff" for c in fragment):
                keywords.add(fragment)
    if not keywords:
        # 没有关键词时给所有素材中等分数
        for asset in assets:
            asset["match_score"] = 0.5
            asset["match_reason"] = "fallback_default"
        return assets
    results = []
    for asset in assets:
        # 组合素材的文本信息：名称 + 标签 + 描述
        asset_text_parts = [
            str(asset.get("name", "")).lower(),
            " ".join(str(t) for t in asset.get("tags", [])).lower(),
            str(asset.get("description", "")).lower(),
        ]
        asset_text = " | ".join(asset_text_parts)
        # 计算匹配度：命中关键词占比 + 稀有关键词加权
        hit_count = 0
        hit_keywords = []
        for kw in keywords:
            if kw in asset_text:
                hit_count += 1
                hit_keywords.append(kw)
        # 基础匹配度 = 命中关键词数 / 总关键词数（开根号平滑）
        base_score = math.sqrt(hit_count / len(keywords)) if keywords else 0.5
        # 名称命中加分（名称匹配更重要）
        name = str(asset.get("name", "")).lower()
        name_hits = sum(1 for kw in hit_keywords if kw in name)
        name_bonus = min(0.2, name_hits * 0.05)
        score = min(1.0, base_score * 0.8 + name_bonus)
        score = round(score, 3)
        results.append(
            {
                **asset,
                "match_score": score,
                "match_reason": "fallback_keyword",
            }
        )
    # 按匹配度降序
    results.sort(key=lambda x: x["match_score"], reverse=True)
    return results
 def _parse_semantic_match_response(
    content: str,
    asset_ids: List[str],
 ) -> Optional[Dict[str, float]]:
-    """从模型返回中解析素材匹配度（薄包装，转发到 ai_parsing 模块）."""
+    """从模型返回中解析素材匹配度.
-    result = _parse_semantic_match_base(content, asset_ids)
+
-    if result is None:
+    期望格式：JSON 对象 {asset_id: score} 或 {"matches": [{asset_id, score}]}
    score 范围 0-1。
    """
    if not content:
        return None
-    return dict(result)
+
    # 尝试解析 JSON
    try:
        cleaned = content.strip()
        if cleaned.startswith("```"):
            cleaned = cleaned.strip("`")
            if cleaned.lower().startswith("json"):
                cleaned = cleaned[4:]
            cleaned = cleaned.strip()
        data = json.loads(cleaned)
        result: Dict[str, float] = {}
        # 格式1: {"asset_id1": 0.8, "asset_id2": 0.6}
        if isinstance(data, dict):
            if "matches" in data and isinstance(data["matches"], list):
                # 格式2: {"matches": [{"asset_id": "...", "score": 0.8}]}
                for item in data["matches"]:
                    if isinstance(item, dict):
                        aid = item.get("asset_id") or item.get("id")
                        score = item.get("score", 0)
                        if aid and isinstance(score, (int, float)):
                            result[str(aid)] = max(0.0, min(1.0, float(score)))
            else:
                for key, value in data.items():
                    if isinstance(value, (int, float)):
                        result[str(key)] = max(0.0, min(1.0, float(value)))
        # 格式3: [{"asset_id": "...", "score": 0.8}]
        elif isinstance(data, list):
            for item in data:
                if isinstance(item, dict):
                    aid = item.get("asset_id") or item.get("id")
                    score = item.get("score", 0)
                    if aid and isinstance(score, (int, float)):
                        result[str(aid)] = max(0.0, min(1.0, float(score)))
        if len(result) >= max(1, len(asset_ids) // 2):  # 至少一半素材有评分才算成功
            return result
    except (json.JSONDecodeError, ValueError):
        pass
    return None
 def semantic_match_assets(
@@ -0,0 +1,276 @@
 """封面管理服务.
 提供封面配置管理和从视频抽帧生成封面的能力。
 抽帧使用 FFmpeg，上传使用共享存储服务。
 """
 from __future__ import annotations
 import logging
 import tempfile
 from pathlib import Path
 from typing import Any, Dict
 logger = logging.getLogger(__name__)
 # ── 常量 ──────────────────────────────────────────────────────────────────────
 DEFAULT_COVER_WIDTH = 1080
 DEFAULT_COVER_HEIGHT = 1920
 DEFAULT_COVER_QUALITY = 5  # JPEG quality (1-31, 越小越好)
 COVER_STORAGE_PREFIX = "covers"
 class CoverService:
    """封面管理服务."""
    def __init__(self, storage_service: Any, asset_repository: Any) -> None:
        self._storage = storage_service
        self._asset_repo = asset_repository
    # ── 配置读写 ──────────────────────────────────────────────────────────
    @staticmethod
    def get_cover_config(plan_config: Dict[str, Any]) -> Dict[str, Any]:
        """从 plan.config 中提取封面配置.
        Args:
            plan_config: 剪辑计划的 config 字段
        Returns:
            封面配置 dict
        """
        cover = plan_config.get("cover", {})
        if not isinstance(cover, dict):
            cover = {}
        # 确保默认字段存在
        return {
            "type": cover.get("type", "ai_frame"),
            "image_url": cover.get("image_url", ""),
            "frame_time": cover.get("frame_time"),
        }
    # ── 抽帧生成封面 ──────────────────────────────────────────────────────
    def extract_cover_from_clip(
        self,
        plan_id: str,
        asset_id: str,
        frame_time: float = 1.0,
        *,
        width: int = DEFAULT_COVER_WIDTH,
        height: int = DEFAULT_COVER_HEIGHT,
        quality: int = DEFAULT_COVER_QUALITY,
    ) -> Dict[str, Any]:
        """从指定素材的指定时间点抽取一帧作为封面.
        Args:
            plan_id: 剪辑计划 ID（用于生成存储路径）
            asset_id: 素材 ID
            frame_time: 抽帧时间点（秒）
            width: 输出宽度
            height: 输出高度
            quality: JPEG 质量
        Returns:
            封面数据 dict，包含 type / image_url / frame_time
        Raises:
            ValueError: 素材不存在或不是视频
            RuntimeError: 抽帧或上传失败
        """
        # 1. 获取素材
        asset = self._asset_repo.get(asset_id) if self._asset_repo else None
        if not asset:
            raise ValueError(f"素材不存在: {asset_id}")
        storage_key = getattr(asset, "storage_key", "")
        if not storage_key:
            raise ValueError(f"素材没有文件: {asset_id}")
        mime_type = getattr(asset, "mime_type", "")
        if mime_type and not mime_type.startswith("video"):
            raise ValueError(f"素材不是视频类型: {mime_type}")
        # 2. 下载视频到临时目录
        with tempfile.TemporaryDirectory(prefix="cover_extract_") as tmp_dir:
            tmp_path = Path(tmp_dir)
            video_path = tmp_path / f"source_{asset_id[:8]}"
            logger.info("下载素材用于封面抽帧: asset_id=%s", asset_id)
            try:
                self._storage.download_file(storage_key, str(video_path))
            except Exception as e:
                raise RuntimeError(f"下载素材失败: {e}") from e
            if not video_path.exists() or video_path.stat().st_size == 0:
                raise RuntimeError("下载的素材文件为空")
            # 3. FFmpeg 抽帧
            output_path = tmp_path / "cover.jpg"
            self._extract_frame(
                video_path=video_path,
                output_path=output_path,
                time_sec=frame_time,
                width=width,
                height=height,
                quality=quality,
            )
            if not output_path.exists() or output_path.stat().st_size == 0:
                raise RuntimeError("封面抽帧失败")
            # 4. 上传到 OSS
            cover_key = f"{COVER_STORAGE_PREFIX}/{plan_id}/cover_{int(frame_time * 1000)}.jpg"
            logger.info("上传封面到存储: key=%s", cover_key)
            try:
                self._storage.upload_file(
                    file_or_path=str(output_path),
                    storage_key=cover_key,
                    content_type="image/jpeg",
                )
            except Exception as e:
                raise RuntimeError(f"上传封面失败: {e}") from e
            # 5. 获取访问 URL
            try:
                image_url = self._storage.get_url(cover_key)
            except Exception:
                image_url = cover_key  # 降级为 storage_key
        logger.info(
            "封面抽帧完成: plan_id=%s asset_id=%s time=%.2fs size=%d",
            plan_id,
            asset_id,
            frame_time,
            output_path.stat().st_size if output_path.exists() else 0,
        )
        return {
            "type": "manual",
            "image_url": image_url,
            "frame_time": frame_time,
        }
    def generate_smart_cover(
        self,
        plan_id: str,
        asset_id: str,
        *,
        width: int = DEFAULT_COVER_WIDTH,
        height: int = DEFAULT_COVER_HEIGHT,
        quality: int = DEFAULT_COVER_QUALITY,
    ) -> Dict[str, Any]:
        """智能选帧：从视频中选取多帧，选最清晰的一帧.
        Args:
            plan_id: 剪辑计划 ID
            asset_id: 素材 ID
            width: 输出宽度
            height: 输出高度
            quality: JPEG 质量
        Returns:
            封面数据 dict
        """
        # 简单实现：取视频 1/3 处的帧作为智能封面
        # 更复杂的多帧选清晰帧可以后续优化
        frame_time = 3.0  # 默认第3秒，后续可以根据视频时长动态计算
        result = self.extract_cover_from_clip(
            plan_id=plan_id,
            asset_id=asset_id,
            frame_time=frame_time,
            width=width,
            height=height,
            quality=quality,
        )
        result["type"] = "ai_frame"
        return result
    # ── 内部方法 ──────────────────────────────────────────────────────────
    @staticmethod
    def _extract_frame(
        video_path: Path,
        output_path: Path,
        *,
        time_sec: float,
        width: int,
        height: int,
        quality: int,
    ) -> None:
        """使用 FFmpeg 从视频中抽取一帧.
        Args:
            video_path: 视频文件路径
            output_path: 输出图片路径
            time_sec: 抽帧时间点（秒）
            width: 输出宽度
            height: 输出高度
            quality: JPEG 质量
        """
        import subprocess
        # scale + crop 实现 cover 裁剪
        vf = f"scale={width}:{height}:force_original_aspect_ratio=increase," f"crop={width}:{height}"
        command = [
            "ffmpeg",
            "-y",
            "-ss",
            f"{time_sec:.3f}",
            "-i",
            str(video_path),
            "-vframes",
            "1",
            "-vf",
            vf,
            "-q:v",
            str(quality),
            "-f",
            "mjpeg",
            str(output_path),
        ]
        logger.debug("FFmpeg 抽帧命令: %s", " ".join(command))
        try:
            result = subprocess.run(
                command,
                capture_output=True,
                text=True,
                timeout=60,
            )
            if result.returncode != 0:
                logger.warning("FFmpeg 抽帧返回非零: %s\nstderr: %s", result.returncode, result.stderr[-500:])
                # 尝试不使用 scale+crop 的简化命令
                simple_command = [
                    "ffmpeg",
                    "-y",
                    "-ss",
                    f"{time_sec:.3f}",
                    "-i",
                    str(video_path),
                    "-vframes",
                    "1",
                    "-q:v",
                    str(quality),
                    "-f",
                    "mjpeg",
                    str(output_path),
                ]
                result2 = subprocess.run(
                    simple_command,
                    capture_output=True,
                    text=True,
                    timeout=60,
                )
                if result2.returncode != 0:
                    raise RuntimeError(f"FFmpeg 抽帧失败: {result2.stderr[-300:]}")
        except subprocess.TimeoutExpired as e:
            raise RuntimeError("FFmpeg 抽帧超时") from e
        except FileNotFoundError as e:
            raise RuntimeError("FFmpeg 不可用") from e
@@ -16,11 +16,6 @@ from packages.adapters.sqlalchemy_impl import (
    SQLAlchemyEditPlanRepository,
    SQLAlchemyGenerationTaskRepository,
 )
 from packages.domain.clip_operations import calculate_merge as _calc_merge
 from packages.domain.clip_operations import calculate_shift_orders as _calc_shift_orders
 from packages.domain.clip_operations import calculate_split as _calc_split
 from packages.domain.clip_operations import validate_merge_clips as _validate_merge
 from packages.domain.clip_operations import validate_split_time as _validate_split
 from packages.domain.edit_plan import EditPlan, EditPlanStatus
 from packages.domain.edit_plan_clip import EditPlanClip, EditPlanClipStatus
@@ -389,45 +384,36 @@ class EditPlanService:
        clip = self.get_clip_or_raise(clip_id)
        plan_id = clip.plan_id
-        # 纯逻辑：校验 + 计算
+        if split_time <= 0 or split_time >= clip.duration:
-        _validate_split(split_time, clip.duration)
+            raise ValueError(f"分割时间必须在 (0, {clip.duration:.3f}) 范围内，当前: {split_time}")
        split = _calc_split(
            duration=clip.duration,
            split_time=split_time,
            start_time=clip.start_time,
        )
        self._auto_resume_editing(plan_id)
        original_duration = clip.duration
        left_duration = round(split_time, 3)
        right_duration = round(original_duration - split_time, 3)
        original_order = clip.order
        # 更新左半部分（原片段）
-        clip.duration = split.left_duration
+        clip.duration = left_duration
        left_clip = self._clip_repo.update(clip)
        # 后面片段的 order 全部 +1（给右半部分腾位置）
        all_clips = self._clip_repo.list_by_plan(plan_id)
-        shifts = _calc_shift_orders(
+        for c in all_clips:
-            all_clips,
+            if c.order > original_order and c.id != clip_id:
-            threshold_order=original_order,
+                c.order += 1
-            shift=1,
+                self._clip_repo.update(c)
            excluded_ids={clip_id},
            id_attr="id",
            order_attr="order",
        )
        for c, new_order in shifts:
            c.order = new_order
            self._clip_repo.update(c)
        # 创建右半部分新片段（继承原片段的大部分属性）
        right_config = dict(clip.config) if clip.config else {}
        # 素材裁剪信息
        if clip.asset_id:
            # 右半部分从 split_time 开始播放
-            right_config["trim_start"] = split.right_trim_start
+            right_config["trim_start"] = left_duration
            # 左半部分在 split_time 处结束
            left_config = dict(left_clip.config) if left_clip.config else {}
-            left_config["trim_end"] = split.left_trim_end
+            left_config["trim_end"] = right_duration
            left_clip.config = left_config
            left_clip = self._clip_repo.update(left_clip)
@@ -438,8 +424,8 @@ class EditPlanService:
            template_clip_config_id=clip.template_clip_config_id,
            asset_id=clip.asset_id,
            text_content=clip.text_content,
-            start_time=split.right_start_time,
+            start_time=clip.start_time + left_duration,
-            duration=split.right_duration,
+            duration=right_duration,
            transition_effect=clip.transition_effect,
            transition_duration=clip.transition_duration,
            playback_speed=clip.playback_speed,
@@ -452,8 +438,8 @@ class EditPlanService:
            clip_id,
            plan_id,
            split_time,
-            split.left_duration,
+            left_duration,
-            split.right_duration,
+            right_duration,
        )
        return {
@@ -482,45 +468,70 @@ class EditPlanService:
            clip = self.get_clip_or_raise(cid)
            clips.append(clip)
-        # 纯逻辑：校验 + 计算
+        # 校验：同一计划
-        plan_id, first_order = _validate_merge(clips)
+        plan_id = clips[0].plan_id
-        merge = _calc_merge(clips)
+        for c in clips[1:]:
            if c.plan_id != plan_id:
                raise ValueError("只能合并同一计划下的片段")
        # 按 order 排序
        clips.sort(key=lambda c: c.order)
        # 校验：order 连续
        for i in range(1, len(clips)):
            if clips[i].order != clips[i - 1].order + 1:
                raise ValueError(f"片段不连续：order {clips[i-1].order} → {clips[i].order}")
        # 校验：类型一致
        clip_type = clips[0].clip_type
        for c in clips[1:]:
            if c.clip_type != clip_type:
                raise ValueError("只能合并相同类型的片段")
        self._auto_resume_editing(plan_id)
        # 计算合并后的属性
        first_clip = clips[0]
        total_duration = round(sum(c.duration for c in clips), 3)
        first_order = first_clip.order
        # 合并文案（用换行连接）
        merged_text = "\n".join(c.text_content for c in clips if c.text_content.strip())
        # 合并 config（后面的覆盖前面的）
        merged_config: Dict[str, Any] = {}
        for c in clips:
            if c.config:
                merged_config.update(c.config)
        # 清理 trim 相关字段（合并后就是完整片段了）
        merged_config.pop("trim_start", None)
        merged_config.pop("trim_end", None)
        # 更新第一个片段（保留它作为合并结果）
-        first_clip = sorted(clips, key=lambda c: c.order)[0]
+        first_clip.duration = total_duration
-        first_clip.duration = merge.total_duration
+        first_clip.text_content = merged_text
-        first_clip.text_content = merge.merged_text
+        first_clip.config = merged_config
        first_clip.config = merge.merged_config
        # 转场保留第一个的（合并后的入点转场）
        # playback_speed 取第一个的
        merged_clip = self._clip_repo.update(first_clip)
        # 删除其余片段
-        rest_ids = [c.id for c in clips if c.id != merged_clip.id]
+        for c in clips[1:]:
-        for cid in rest_ids:
+            self._clip_repo.delete(c.id)
            self._clip_repo.delete(cid)
        # 后面的片段 order 前移 (len - 1) 位
        shift = len(clips) - 1
        all_clips = self._clip_repo.list_by_plan(plan_id)
-        shifts = _calc_shift_orders(
+        for c in all_clips:
-            all_clips,
+            if c.order > first_order and c.id != merged_clip.id:
-            threshold_order=first_order,
+                c.order -= shift
-            shift=-merge.shift_amount,
+                self._clip_repo.update(c)
            excluded_ids={merged_clip.id},
            id_attr="id",
            order_attr="order",
        )
        for c, new_order in shifts:
            c.order = new_order
            self._clip_repo.update(c)
        logger.info(
            "合并片段: plan_id=%s count=%d total_duration=%.3fs",
            plan_id,
            len(clips),
-            merge.total_duration,
+            total_duration,
        )
        return merged_clip
@@ -571,10 +582,6 @@ class EditPlanService:
    def can_generate(self, plan_id: str) -> tuple[bool, str]:
        """检查是否可以触发渲染
        包含最后一道防线的自动修复：
        - 如果 clips 存在但都没有 asset_id，且 config.asset_ids 非空，
          直接在内部执行素材分配，不再依赖前置 fallback 链路。
        Returns:
            tuple: (can_generate, reason)
        """
@@ -589,69 +596,10 @@ class EditPlanService:
        if not clips:
            return False, "请先添加片段后再生成视频"
        # 检查是否至少有一个片段分配了素材
        has_asset = any(c.asset_id for c in clips)
        config_asset_ids_count = len((plan.config or {}).get("asset_ids", []))
        clips_with_asset_count = sum(1 for c in clips if c.asset_id)
        logger.info(
            "can_generate 诊断: plan=%s status=%s total_clips=%d " "clips_with_asset=%d config_asset_ids_count=%d",
            plan_id,
            plan.status,
            len(clips),
            clips_with_asset_count,
            config_asset_ids_count,
        )
        if not has_asset:
            # ── 最后防线：自动从 config.asset_ids 分配素材 ──
            config_asset_ids = (plan.config or {}).get("asset_ids", [])
            if config_asset_ids:
                logger.warning(
                    "can_generate 最后防线触发: plan=%s clips=%d 均无素材，" "从 config.asset_ids(%d个) 自动分配",
                    plan_id,
                    len(clips),
                    len(config_asset_ids),
                )
                clips_without_asset = [c for c in clips if not c.asset_id]
                assigned_count = 0
                for i, clip in enumerate(clips_without_asset):
                    asset_idx = i % len(config_asset_ids)
                    try:
                        self.assign_asset(clip.id, config_asset_ids[asset_idx])
                        assigned_count += 1
                    except Exception as exc:
                        logger.warning(
                            "can_generate 最后防线: plan=%s clip=%s 分配素材 %s 失败: %s",
                            plan_id,
                            clip.id,
                            config_asset_ids[asset_idx],
                            exc,
                        )
                logger.info(
                    "can_generate 最后防线: plan=%s 已为 %d/%d 个片段分配素材",
                    plan_id,
                    assigned_count,
                    len(clips_without_asset),
                )
                # 重新加载 clips 验证分配结果
                clips = self._clip_repo.list_by_plan(plan_id)
                if not any(c.asset_id for c in clips):
                    return False, "没有可渲染的就绪片段，自动修复后仍未分配素材"
            else:
                logger.warning(
                    "can_generate 失败: plan=%s clips=%d 均无素材，" "且 config.asset_ids 为空，无法自动修复",
                    plan_id,
                    len(clips),
                )
                return False, "没有可渲染的就绪片段，请确保已选择素材"
        return True, ""
    def mark_clips_ready(self, plan_id: str) -> int:
-        """将已分配素材的 pending 片段标记为 ready
+        """将所有 pending 状态的片段标记为 ready
        只标记同时满足以下条件的片段：
        - status == PENDING
        - asset_id 非空（已分配素材）
        Returns:
            int: 标记的片段数量
@@ -662,16 +610,10 @@ class EditPlanService:
        )
        count = 0
        for clip in clips:
-            if clip.asset_id:
+            clip.mark_ready()
-                clip.mark_ready()
+            self._clip_repo.update(clip)
-                self._clip_repo.update(clip)
+            count += 1
-                count += 1
+        logger.info("标记片段就绪: plan_id=%s count=%d", plan_id, count)
        logger.info(
            "标记片段就绪: plan_id=%s marked=%d total_pending=%d",
            plan_id,
            count,
            len(clips),
        )
        return count
    def update_plan_config(self, plan_id: str, config_updates: Dict[str, Any]) -> EditPlan:
@@ -23,13 +23,6 @@ from packages.domain.template_clip_config import (
    TemplateClipConfig,
    TransitionEffect,
 )
 from packages.domain.template_clip_converter import (
    clip_configs_to_snapshots,
    clips_to_template_clip_configs,
    filter_plan_config_to_template,
    snapshots_to_template_clip_configs,
    validate_template_name,
 )
 logger = logging.getLogger(__name__)
@@ -128,7 +121,9 @@ class EditTemplateService:
            ValueError: 名称为空或重复
        """
        # 名称校验
-        clean_name = validate_template_name(name)
+        clean_name = name.strip()
        if not clean_name:
            raise ValueError("模板名称不能为空")
        # 名称重复检查
        existing = self._template_repo.list_all(skip=0, limit=1000)
@@ -476,7 +471,12 @@ class EditTemplateService:
                raise ValueError(f"模板名称已存在: {clean_name}")
        # 从计划 config 中提取模板级配置，去掉运行时/素材相关字段
-        template_config = filter_plan_config_to_template(plan.config)
+        plan_config = plan.config or {}
        template_config: dict[str, Any] = {}
        for key, value in plan_config.items():
            # 跳过明显的运行时/实例字段，保留风格/模式类配置
            if key not in {"asset_ids", "source_edit_plan_id", "generation_task_id"}:
                template_config[key] = value
        template = EditTemplate.create(
            name=clean_name,
@@ -497,7 +497,40 @@ class EditTemplateService:
        # 5. 转换每个片段为模板片段配置
        created_configs: List[TemplateClipConfig] = []
-        for clip_config_obj in clips_to_template_clip_configs(created_template.id, clips):
+        for clip in clips:
            clip_config: dict[str, Any] = {}
            # 播放速度存入 config
            if clip.playback_speed and clip.playback_speed != 1.0:
                clip_config["playback_speed"] = clip.playback_speed
            # 片段自有 config 合并（优先级：clip.config 覆盖上面的）
            if clip.config:
                clip_config.update(clip.config)
            # 去掉素材相关字段
            clip_config.pop("asset_info", None)
            clip_config.pop("source_asset_id", None)
            # 转场效果兼容校验
            try:
                transition = TransitionEffect(clip.transition_effect)
            except ValueError:
                transition = TransitionEffect.CUT
            # 片段类型兼容校验
            try:
                clip_type = ClipType(clip.clip_type)
            except ValueError:
                clip_type = ClipType.MAIN
            clip_config_obj = TemplateClipConfig.create(
                template_id=created_template.id,
                clip_type=clip_type,
                order=clip.order,
                min_duration=clip.duration,
                max_duration=clip.duration,
                text_template=clip.text_content or "",
                transition_effect=transition,
                config=clip_config,
            )
            created = self._clip_config_repo.create(clip_config_obj)
            created_configs.append(created)
@@ -646,6 +679,8 @@ class EditTemplateService:
        Raises:
            ValueError: 模板/草稿不存在，或草稿不属于该模板
        """
        from packages.domain.template_clip_config import TemplateClipConfig
        # 1. 校验模板和草稿
        template = self.get_template_or_raise(template_id)
        draft = self._plan_repo.get(draft_plan_id)
@@ -665,14 +700,39 @@ class EditTemplateService:
        editing_mode = config.get("editing_mode", "one_take")
        # 4. 提取模板配置（去掉草稿/运行时字段）
-        template_config = filter_plan_config_to_template(draft.config)
+        draft_config = draft.config or {}
        template_config: dict[str, Any] = {}
        skip_keys = {
            "is_template_draft",
            "asset_ids",
            "source_edit_plan_id",
            "generation_task_id",
        }
        for key, value in draft_config.items():
            if key not in skip_keys:
                template_config[key] = value
        # 5. 事务更新
        try:
            # 5.0 先保存旧版快照（发布前的状态），用于回滚
            old_version = template.version or 1
            old_clip_configs = self._clip_config_repo.list_by_template(template_id)
-            old_clip_snapshots = clip_configs_to_snapshots(old_clip_configs)
+            old_clip_snapshots = [
                {
                    "clip_type": cfg.clip_type.value if hasattr(cfg.clip_type, "value") else cfg.clip_type,
                    "order": cfg.order,
                    "min_duration": cfg.min_duration,
                    "max_duration": cfg.max_duration,
                    "text_template": cfg.text_template or "",
                    "transition_effect": (
                        cfg.transition_effect.value
                        if hasattr(cfg.transition_effect, "value")
                        else cfg.transition_effect
                    ),
                    "config": cfg.config or {},
                }
                for cfg in old_clip_configs
            ]
            from packages.domain.template_version import EditTemplateVersion
@@ -699,7 +759,46 @@ class EditTemplateService:
            # 创建新的片段配置
            created_configs: list[TemplateClipConfig] = []
-            for config_obj in clips_to_template_clip_configs(template_id, draft_clips):
+            for clip in draft_clips:
                clip_config: dict[str, Any] = {}
                # 播放速度存入 config
                if clip.playback_speed and clip.playback_speed != 1.0:
                    clip_config["playback_speed"] = clip.playback_speed
                # 片段自有 config 合并
                if clip.config:
                    clip_config.update(clip.config)
                # 去掉素材相关字段
                clip_config.pop("asset_info", None)
                clip_config.pop("source_asset_id", None)
                # 转场效果兼容校验
                try:
                    from packages.domain.template_clip_config import (
                        TransitionEffect,
                    )
                    transition = TransitionEffect(clip.transition_effect)
                except (ValueError, ImportError):
                    transition = TransitionEffect.CUT  # type: ignore
                # 片段类型兼容校验
                try:
                    from packages.domain.template_clip_config import ClipType
                    clip_type = ClipType(clip.clip_type)
                except (ValueError, ImportError):
                    clip_type = ClipType.MAIN  # type: ignore
                config_obj = TemplateClipConfig.create(
                    template_id=template_id,
                    clip_type=clip_type,
                    order=clip.order,
                    min_duration=clip.duration,
                    max_duration=clip.duration,
                    text_template=clip.text_content or "",
                    transition_effect=transition,
                    config=clip_config,
                )
                created = self._clip_config_repo.create(config_obj)
                created_configs.append(created)
@@ -744,6 +843,8 @@ class EditTemplateService:
        Raises:
            ValueError: 模板/版本不存在
        """
        from packages.domain.template_clip_config import TemplateClipConfig
        template = self.get_template_or_raise(template_id)
        # 1. 读取目标版本快照
@@ -756,7 +857,22 @@ class EditTemplateService:
        try:
            # 2. 先保存当前状态快照（当前版本号），确保回滚可撤销
            old_clip_configs = self._clip_config_repo.list_by_template(template_id)
-            old_clip_snapshots = clip_configs_to_snapshots(old_clip_configs)
+            old_clip_snapshots = [
                {
                    "clip_type": cfg.clip_type.value if hasattr(cfg.clip_type, "value") else cfg.clip_type,
                    "order": cfg.order,
                    "min_duration": cfg.min_duration,
                    "max_duration": cfg.max_duration,
                    "text_template": cfg.text_template or "",
                    "transition_effect": (
                        cfg.transition_effect.value
                        if hasattr(cfg.transition_effect, "value")
                        else cfg.transition_effect
                    ),
                    "config": cfg.config or {},
                }
                for cfg in old_clip_configs
            ]
            from packages.domain.template_version import EditTemplateVersion
@@ -789,7 +905,37 @@ class EditTemplateService:
                synchronize_session=False
            )
-            for config_obj in snapshots_to_template_clip_configs(template_id, target_version.clip_configs):
+            for clip_snap in target_version.clip_configs:
                # 转场效果兼容校验
                try:
                    from packages.domain.template_clip_config import TransitionEffect
                    transition = TransitionEffect(clip_snap.get("transition_effect", "cut"))
                except (ValueError, ImportError):
                    from packages.domain.template_clip_config import TransitionEffect
                    transition = TransitionEffect.CUT
                # 片段类型兼容校验
                try:
                    from packages.domain.template_clip_config import ClipType
                    clip_type = ClipType(clip_snap.get("clip_type", "main"))
                except (ValueError, ImportError):
                    from packages.domain.template_clip_config import ClipType
                    clip_type = ClipType.MAIN
                config_obj = TemplateClipConfig.create(
                    template_id=template_id,
                    clip_type=clip_type,
                    order=clip_snap.get("order", 0),
                    min_duration=clip_snap.get("min_duration", 0.0),
                    max_duration=clip_snap.get("max_duration", 0.0),
                    text_template=clip_snap.get("text_template", ""),
                    transition_effect=transition,
                    config=clip_snap.get("config", {}) or {},
                )
                self._clip_config_repo.create(config_obj)
            self._db.commit()
@@ -26,13 +26,7 @@ from packages.domain.edit_plan import EditPlan
 from packages.domain.edit_plan_clip import EditPlanClip
 from packages.domain.edit_template import EditTemplate
 from packages.domain.editing_mode import EditingMode
-from packages.domain.plan_generator_utils import (
+from packages.domain.template_clip_config import ClipType, TemplateClipConfig
    create_clips_from_configs,
    distribute_assets,
    generate_default_clips,
    map_clip_types_for_mode,
 )
 from packages.domain.template_clip_config import TemplateClipConfig
 logger = logging.getLogger(__name__)
@@ -49,10 +43,9 @@ class PlanGeneratorService:
    基于模板 + 素材，自动生成 EditPlan 及 EditPlanClip 列表。
    """
-    def __init__(self, db: Session, asset_repo=None) -> None:
+    def __init__(self, db: Session) -> None:
        self._plan_repo = SQLAlchemyEditPlanRepository(db)
        self._clip_repo = SQLAlchemyEditPlanClipRepository(db)
        self._asset_repo = asset_repo
    # ── 公开接口 ─────────────────────────────────────────────────────────────
@@ -65,7 +58,6 @@ class PlanGeneratorService:
        project_id: str = "",
        created_by_user_id: str = "",
        name: str = "",
        random_preview: bool = False,
    ) -> dict[str, Any]:
        """基于模板+素材生成剪辑计划
@@ -76,7 +68,6 @@ class PlanGeneratorService:
            project_id: 所属项目 ID
            created_by_user_id: 创建者用户 ID
            name: 计划名称（为空则自动取模板名）
            random_preview: 是否启用随机预览模式（随机选素材+随机截取片段）
        Returns:
            dict: {"plan": EditPlan, "clips": List[EditPlanClip]}
@@ -118,17 +109,7 @@ class PlanGeneratorService:
        # 4. 按 editing_mode 分配素材
        if asset_ids:
-            # 如果是随机预览模式，获取素材时长信息
+            self._distribute_assets(clips, asset_ids, editing_mode)
            asset_durations = None
            if random_preview and self._asset_repo:
                asset_durations = self._fetch_asset_durations(asset_ids)
            self._distribute_assets(
                clips,
                asset_ids,
                editing_mode,
                random_selection=random_preview,
                asset_durations=asset_durations,
            )
        # 5. 持久化所有 clips 并计算总时长
        created_clips: List[EditPlanClip] = []
@@ -182,18 +163,83 @@ class PlanGeneratorService:
        plan_id: str,
        clip_configs: List[TemplateClipConfig],
    ) -> List[EditPlanClip]:
-        """从 TemplateClipConfig 列表创建 EditPlanClip 列表（未持久化）.
+        """从 TemplateClipConfig 列表创建 EditPlanClip 列表（未持久化）"""
        clips: List[EditPlanClip] = []
        # 按 order 排序
        sorted_configs = sorted(clip_configs, key=lambda c: c.order)
-        委托给 plan_generator_utils.create_clips_from_configs 纯函数。
+        for cfg in sorted_configs:
-        """
+            # 计算时长：取 min_duration 和 max_duration 的中间值
-        return create_clips_from_configs(plan_id, clip_configs)
+            if cfg.min_duration > 0 and cfg.max_duration > 0:
                duration = (cfg.min_duration + cfg.max_duration) / 2
            elif cfg.min_duration > 0:
                duration = cfg.min_duration
            elif cfg.max_duration > 0:
                duration = cfg.max_duration
            else:
                duration = _DEFAULT_CLIP_DURATION
            # clip_type 可能是枚举或字符串
            clip_type = cfg.clip_type.value if hasattr(cfg.clip_type, "value") else cfg.clip_type
            # transition_effect 可能是枚举或字符串
            transition = (
                cfg.transition_effect.value if hasattr(cfg.transition_effect, "value") else cfg.transition_effect
            )
            # 从 clip config 中解析 playback_speed（兼容 speed_ratio 字段名）
            clip_cfg = cfg.config or {}
            playback_speed = clip_cfg.get("playback_speed", clip_cfg.get("speed_ratio", 1.0)) or 1.0
            clip = EditPlanClip.create(
                plan_id=plan_id,
                clip_type=clip_type,
                order=cfg.order,
                template_clip_config_id=cfg.id,
                text_content=getattr(cfg, "text_template", "") or "",
                duration=duration,
                transition_effect=transition or "cut",
                playback_speed=playback_speed,
                config=clip_cfg,
            )
            clips.append(clip)
        return clips
    def _map_clip_types_for_mode(self, clips: List[EditPlanClip], editing_mode: str) -> None:
-        """将 MAIN 类型片段按 editing_mode 映射为对应角色类型.
+        """将模板 clip_config 生成的 MAIN 类型片段，按 editing_mode 映射为对应角色类型。
-        委托给 plan_generator_utils.map_clip_types_for_mode 纯函数。
+        模板的 clip_config 使用 ClipType 枚举（main/intro/outro 等），
        但 PIP / VOICE_PIP 模式的素材分配和渲染分层依赖特定的 clip_type 命名
        （overlay / background / corner_voice / b_roll）。
        映射规则（仅修改 MAIN 类型片段，非 MAIN 片段保持原类型）：
        - PIP: 第1个 MAIN → main（背景），其余 MAIN → overlay（画中画）
        - VOICE_PIP: 第1个 → background，第2个 → corner_voice，第3+个 → b_roll
        - ONE_TAKE / VOICE_OVER: 保持 main 不变
        """
-        map_clip_types_for_mode(clips, editing_mode)
+        from packages.domain.template_clip_config import ClipType
        main_clips = [c for c in clips if c.clip_type == ClipType.MAIN.value]
        if not main_clips:
            return
        if editing_mode == EditingMode.PIP.value:
            # 第1个 main 保持（背景层），其余改为 overlay（画中画层）
            for i, clip in enumerate(main_clips):
                if i > 0:
                    clip.clip_type = "overlay"
        elif editing_mode == EditingMode.VOICE_PIP.value:
            for i, clip in enumerate(main_clips):
                if i == 0:
                    clip.clip_type = "background"
                elif i == 1:
                    clip.clip_type = "corner_voice"
                else:
                    clip.clip_type = "b_roll"
        # ONE_TAKE / VOICE_OVER: 保持 main 不变，无需处理
    def _generate_default_clips(
        self,
@@ -201,45 +247,191 @@ class PlanGeneratorService:
        editing_mode: str,
        asset_count: int,
    ) -> List[EditPlanClip]:
-        """无 clip_configs 时，根据 editing_mode 生成默认 clip 结构.
+        """无 clip_configs 时，根据 editing_mode 生成默认 clip 结构
-        委托给 plan_generator_utils.generate_default_clips 纯函数。
+        - ONE_TAKE: N 个 main clips（N = asset_count，至少1个）
        - PIP: 1 个 main + (N-1) 个 overlay（N = asset_count）
        - VOICE_OVER: N 个 main clips + 标记需要配音
        - VOICE_PIP: 1 个 background + 1 个 corner_voice + (N-2) 个 b_roll
        """
-        return generate_default_clips(plan_id, editing_mode, asset_count)
+        n = max(asset_count, 1)
        clips: List[EditPlanClip] = []
        order = 0
        if editing_mode == EditingMode.PIP.value:
            # 1 个 main（全屏背景）
            clips.append(
                EditPlanClip.create(
                    plan_id=plan_id,
                    clip_type=ClipType.MAIN.value,
                    order=order,
                    duration=_DEFAULT_CLIP_DURATION,
                )
            )
            order += 1
            # 剩余为 overlay
            for _ in range(1, n):
                clips.append(
                    EditPlanClip.create(
                        plan_id=plan_id,
                        clip_type="overlay",
                        order=order,
                        duration=_DEFAULT_CLIP_DURATION,
                    )
                )
                order += 1
        elif editing_mode == EditingMode.VOICE_OVER.value:
            # N 个 main clips（B-roll）
            for _ in range(n):
                clips.append(
                    EditPlanClip.create(
                        plan_id=plan_id,
                        clip_type=ClipType.MAIN.value,
                        order=order,
                        duration=_DEFAULT_CLIP_DURATION,
                        config={"role": "b_roll"},
                    )
                )
                order += 1
        elif editing_mode == EditingMode.VOICE_PIP.value:
            # 1 个 background
            clips.append(
                EditPlanClip.create(
                    plan_id=plan_id,
                    clip_type="background",
                    order=order,
                    duration=_DEFAULT_CLIP_DURATION,
                )
            )
            order += 1
            # 1 个 corner_voice
            clips.append(
                EditPlanClip.create(
                    plan_id=plan_id,
                    clip_type="corner_voice",
                    order=order,
                    duration=_DEFAULT_CLIP_DURATION,
                )
            )
            order += 1
            # 剩余为 b_roll
            for _ in range(2, n):
                clips.append(
                    EditPlanClip.create(
                        plan_id=plan_id,
                        clip_type="b_roll",
                        order=order,
                        duration=_DEFAULT_CLIP_DURATION,
                    )
                )
                order += 1
        else:
            # ONE_TAKE: N 个 main clips
            for _ in range(n):
                clips.append(
                    EditPlanClip.create(
                        plan_id=plan_id,
                        clip_type=ClipType.MAIN.value,
                        order=order,
                        duration=_DEFAULT_CLIP_DURATION,
                    )
                )
                order += 1
        return clips
    def _distribute_assets(
        self,
        clips: List[EditPlanClip],
        asset_ids: List[str],
        editing_mode: str,
        *,
        random_selection: bool = False,
        asset_durations: dict[str, float] | None = None,
    ) -> None:
-        """按 editing_mode 将素材分配到 clips（就地修改，未持久化）.
+        """按 editing_mode 将素材分配到 clips（就地修改，未持久化）
-        委托给 plan_generator_utils.distribute_assets 纯函数。
+        分配策略:
        - ONE_TAKE: 素材按顺序依次分配给 main 类型 clips
        - PIP: 第1个素材→main(全屏背景)，其余→交替分配给 overlay clips
        - VOICE_OVER: 素材→main clips (B-roll)
        - VOICE_PIP: 第1个→background, 第2个→corner_voice, 其余→b_roll
        """
-        distribute_assets(
+        if not asset_ids or not clips:
-            clips,
+            return
            asset_ids,
            editing_mode,
            random_selection=random_selection,
            asset_durations=asset_durations,
        )
-    def _fetch_asset_durations(self, asset_ids: List[str]) -> dict[str, float]:
+        if editing_mode == EditingMode.ONE_TAKE.value:
-        """从数据库获取素材时长信息.
+            self._distribute_one_take(clips, asset_ids)
        elif editing_mode == EditingMode.PIP.value:
            self._distribute_pip(clips, asset_ids)
        elif editing_mode == EditingMode.VOICE_OVER.value:
            self._distribute_voice_over(clips, asset_ids)
        elif editing_mode == EditingMode.VOICE_PIP.value:
            self._distribute_voice_pip(clips, asset_ids)
        else:
            # 未知模式，退化为 one_take
            self._distribute_one_take(clips, asset_ids)
-        Args:
+    def _distribute_one_take(
-            asset_ids: 素材 ID 列表
+        self,
        clips: List[EditPlanClip],
        asset_ids: List[str],
    ) -> None:
        """ONE_TAKE: 素材按顺序依次分配给 main 类型 clips"""
        main_clips = [c for c in clips if c.clip_type == ClipType.MAIN.value]
        for i, clip in enumerate(main_clips):
            if i < len(asset_ids):
                clip.assign_asset(asset_ids[i])
-        Returns:
+    def _distribute_pip(
-            dict: 素材 ID -> 时长（秒）映射
+        self,
-        """
+        clips: List[EditPlanClip],
-        durations: dict[str, float] = {}
+        asset_ids: List[str],
-        for asset_id in asset_ids:
+    ) -> None:
-            asset = self._asset_repo.get(asset_id)
+        """PIP: 第1个素材→main(全屏背景)，其余→overlay clips"""
-            if asset and hasattr(asset, "duration"):
+        # 第1个素材 → main clip
-                durations[asset_id] = float(asset.duration or 0.0)
+        main_clips = [c for c in clips if c.clip_type == ClipType.MAIN.value]
-        return durations
+        if main_clips and asset_ids:
            main_clips[0].assign_asset(asset_ids[0])
        # 其余素材 → overlay clips
        overlay_clips = [c for c in clips if c.clip_type == "overlay"]
        remaining = asset_ids[1:]
        for i, clip in enumerate(overlay_clips):
            if i < len(remaining):
                clip.assign_asset(remaining[i])
    def _distribute_voice_over(
        self,
        clips: List[EditPlanClip],
        asset_ids: List[str],
    ) -> None:
        """VOICE_OVER: 素材→main clips (B-roll)"""
        main_clips = [c for c in clips if c.clip_type == ClipType.MAIN.value]
        for i, clip in enumerate(main_clips):
            if i < len(asset_ids):
                clip.assign_asset(asset_ids[i])
    def _distribute_voice_pip(
        self,
        clips: List[EditPlanClip],
        asset_ids: List[str],
    ) -> None:
        """VOICE_PIP: 第1个→background, 第2个→corner_voice, 其余→b_roll"""
        bg_clips = [c for c in clips if c.clip_type == "background"]
        corner_clips = [c for c in clips if c.clip_type == "corner_voice"]
        broll_clips = [c for c in clips if c.clip_type == "b_roll"]
        # 第1个素材 → background
        if bg_clips and len(asset_ids) > 0:
            bg_clips[0].assign_asset(asset_ids[0])
        # 第2个素材 → corner_voice
        if corner_clips and len(asset_ids) > 1:
            corner_clips[0].assign_asset(asset_ids[1])
        # 其余素材 → b_roll
        remaining = asset_ids[2:]
        for i, clip in enumerate(broll_clips):
            if i < len(remaining):
                clip.assign_asset(remaining[i])
@@ -0,0 +1,326 @@
 """SmartAssetSelector — 智能素材选择服务.
 根据多维度评分从素材库中自动选择最优视频素材，
 用于一键生成等需要自动选取素材的场景。
 评分维度（加权求和，总分 0-1）：
  - 质量分（quality_score）：权重 0.5 — 来自人工或AI的质量评分
  - 分辨率适配：权重 0.2 — 分辨率越接近 1080p 得分越高
  - 时长合理性：权重 0.2 — 3-30 秒区间最佳，过短/过长扣分
  - 码率质量：权重 0.1 — 用文件大小/时长估算，码率适中得分高
 特性：
  - 最低质量分门槛：自动过滤低质量素材
  - 时长多样性：保证选出的素材时长分布均匀（短/中/长各占一定比例）
  - 兼容全部模式：素材库模式和项目模式都可用
 """
 from __future__ import annotations
 import logging
 from dataclasses import dataclass
 logger = logging.getLogger(__name__)
 # ── 评分权重 ──────────────────────────────────────────────────────────────────
 _WEIGHT_QUALITY = 0.5
 _WEIGHT_RESOLUTION = 0.2
 _WEIGHT_DURATION = 0.2
 _WEIGHT_BITRATE = 0.1
 # ── 评分参数 ──────────────────────────────────────────────────────────────────
 _TARGET_WIDTH = 1920  # 目标分辨率宽度基准
 _TARGET_HEIGHT = 1080  # 目标分辨率高度基准
 _MIN_QUALITY_SCORE = 30.0  # 最低质量分门槛（低于此值的素材直接排除）
 _OPTIMAL_DURATION_MIN = 3.0  # 最佳时长区间（秒）
 _OPTIMAL_DURATION_MAX = 30.0
 # ── 多样性分桶 ───────────────────────────────────────────────────────────────
 _SHORT_BUCKET_MAX = 5.0  # 短素材：< 5s
 _MEDIUM_BUCKET_MAX = 15.0  # 中素材：5-15s
 # 长素材：> 15s
@dataclass
 class SmartSelectResult:
    """智能选择结果."""
    selected_ids: list[str]
    total_candidates: int
    filtered_out: int  # 被质量门槛过滤的数量
    avg_score: float
    details: list[AssetScoreDetail]
@dataclass
 class AssetScoreDetail:
    """单个素材的评分详情."""
    asset_id: str
    total_score: float
    quality_score: float
    resolution_score: float
    duration_score: float
    bitrate_score: float
    duration: float | None
 class SmartAssetSelector:
    """智能素材选择器.
    从一组素材中按综合评分选择最优的 N 个，
    同时保证时长分布的多样性。
    """
    def __init__(
        self,
        min_quality_score: float = _MIN_QUALITY_SCORE,
        target_width: int = _TARGET_WIDTH,
        target_height: int = _TARGET_HEIGHT,
    ):
        self.min_quality_score = min_quality_score
        self.target_width = target_width
        self.target_height = target_height
    # ── 公开方法 ──────────────────────────────────────────────────────────────
    def select(
        self,
        assets: list,
        count: int = 0,
        *,
        ensure_diversity: bool = True,
    ) -> SmartSelectResult:
        """从素材列表中智能选择最优素材.
        Args:
            assets: Asset 实体列表（需要有 id/quality_score/width/height/duration/file_size 属性）
            count: 选取数量，0 表示全部符合条件的
            ensure_diversity: 是否保证时长多样性（默认开启）
        Returns:
            SmartSelectResult 选择结果
        """
        # 1. 过滤：只保留 ready 状态的视频素材 + 最低质量分门槛
        candidates = []
        filtered_out = 0
        for asset in assets:
            status = getattr(asset, "status", None)
            status_val = status.value if hasattr(status, "value") else str(status)
            if status_val != "ready":
                continue
            mime_type = getattr(asset, "mime_type", "") or ""
            if not mime_type.startswith("video"):
                continue
            quality = getattr(asset, "quality_score", None)
            if quality is not None and quality < self.min_quality_score:
                filtered_out += 1
                continue
            candidates.append(asset)
        if not candidates:
            return SmartSelectResult(
                selected_ids=[],
                total_candidates=0,
                filtered_out=filtered_out,
                avg_score=0.0,
                details=[],
            )
        # 2. 对每个候选素材评分
        scored: list[AssetScoreDetail] = []
        for asset in candidates:
            detail = self._score_asset(asset)
            scored.append(detail)
        # 3. 按总分降序排列
        scored.sort(key=lambda d: d.total_score, reverse=True)
        # 4. 多样性选择（如果需要且数量有限制）
        if ensure_diversity and count > 0 and len(scored) > count:
            selected = self._diverse_selection(scored, count)
        else:
            # 无数量限制或不要求多样性，直接按排名取
            selected = scored if count <= 0 else scored[:count]
        avg_score = sum(d.total_score for d in selected) / len(selected) if selected else 0.0
        result = SmartSelectResult(
            selected_ids=[d.asset_id for d in selected],
            total_candidates=len(candidates),
            filtered_out=filtered_out,
            avg_score=avg_score,
            details=selected,
        )
        logger.info(
            "智能素材选择完成: 候选=%d, 过滤=%d, 选中=%d, 平均分=%.3f",
            result.total_candidates,
            result.filtered_out,
            len(result.selected_ids),
            result.avg_score,
        )
        return result
    # ── 内部方法 ──────────────────────────────────────────────────────────────
    def _score_asset(self, asset) -> AssetScoreDetail:
        """对单个素材进行多维度评分."""
        # 质量分
        quality = getattr(asset, "quality_score", None)
        quality_score = (quality / 100.0) if quality is not None else 0.5
        # 分辨率评分：越接近目标分辨率得分越高
        width = getattr(asset, "width", None)
        height = getattr(asset, "height", None)
        resolution_score = self._score_resolution(width, height)
        # 时长评分：在最佳区间内得分高，过短过长扣分
        duration = getattr(asset, "duration", None)
        duration_score = self._score_duration(duration)
        # 码率评分：用 file_size/duration 估算，适中得分高
        file_size = getattr(asset, "file_size", 0) or 0
        bitrate_score = self._score_bitrate(file_size, duration)
        # 加权总分
        total = (
            _WEIGHT_QUALITY * quality_score
            + _WEIGHT_RESOLUTION * resolution_score
            + _WEIGHT_DURATION * duration_score
            + _WEIGHT_BITRATE * bitrate_score
        )
        return AssetScoreDetail(
            asset_id=asset.id,
            total_score=round(total, 4),
            quality_score=round(quality_score, 4),
            resolution_score=round(resolution_score, 4),
            duration_score=round(duration_score, 4),
            bitrate_score=round(bitrate_score, 4),
            duration=duration,
        )
    def _score_resolution(self, width: int | None, height: int | None) -> float:
        """分辨率评分：越接近目标分辨率得分越高，低于480p扣分严重."""
        if width is None or height is None or width <= 0 or height <= 0:
            return 0.5  # 未知分辨率给中评分
        target_pixels = self.target_width * self.target_height
        actual_pixels = width * height
        # 计算像素数比例
        ratio = actual_pixels / target_pixels
        if ratio >= 1.0:
            # 高于或等于目标分辨率：满分，略高不扣分（4K也给满分）
            return 1.0
        else:
            # 低于目标分辨率：线性衰减，但最低不低于 0.1
            # 例如：720p (921600) / 1080p (2073600) = 0.44 → 得分 0.6
            score = 0.3 + 0.7 * ratio
            return max(0.1, min(1.0, score))
    def _score_duration(self, duration: float | None) -> float:
        """时长评分：3-30秒最佳，过短或过长都扣分."""
        if duration is None or duration <= 0:
            return 0.5  # 未知时长给中评分
        if _OPTIMAL_DURATION_MIN <= duration <= _OPTIMAL_DURATION_MAX:
            # 最佳区间：满分
            return 1.0
        if duration < _OPTIMAL_DURATION_MIN:
            # 太短：线性衰减，1秒以下给 0.3
            ratio = duration / _OPTIMAL_DURATION_MIN
            return 0.3 + 0.7 * ratio
        # 太长：每超过最佳区间上限10秒扣 0.1 分，最低 0.2
        excess = duration - _OPTIMAL_DURATION_MAX
        penalty = min(0.8, excess / 10.0 * 0.1)
        return max(0.2, 1.0 - penalty)
    def _score_bitrate(self, file_size: int, duration: float | None) -> float:
        """码率评分：根据文件大小和时长估算码率，适中得分高."""
        if not file_size or not duration or duration <= 0:
            return 0.5  # 未知给中评分
        # 估算码率（bps）
        bitrate = (file_size * 8) / duration
        # 最佳码率范围：2-8 Mbps
        optimal_low = 2_000_000  # 2 Mbps
        optimal_high = 8_000_000  # 8 Mbps
        if optimal_low <= bitrate <= optimal_high:
            return 1.0
        if bitrate < optimal_low:
            # 码率太低：线性衰减
            ratio = bitrate / optimal_low
            return 0.3 + 0.7 * ratio
        # 码率太高（文件太大）：适度扣分
        excess = bitrate / optimal_high - 1.0
        penalty = min(0.5, excess * 0.2)
        return max(0.5, 1.0 - penalty)
    def _diverse_selection(self, scored: list[AssetScoreDetail], count: int) -> list[AssetScoreDetail]:
        """多样性选择：按时长分桶，保证每个桶都有素材.
        策略：
        1. 按时长分为三桶：短(<5s)、中(5-15s)、长(>15s)
        2. 每个桶配额 = max(1, count / 3)
        3. 先从每桶按配额取最高分的
        4. 剩余名额从全局最高分中取（不重复）
        """
        # 分桶
        short_bucket = [d for d in scored if d.duration is not None and d.duration < _SHORT_BUCKET_MAX]
        medium_bucket = [
            d for d in scored if d.duration is not None and _SHORT_BUCKET_MAX <= d.duration < _MEDIUM_BUCKET_MAX
        ]
        long_bucket = [d for d in scored if d.duration is not None and d.duration >= _MEDIUM_BUCKET_MAX]
        unknown_bucket = [d for d in scored if d.duration is None]
        buckets = [short_bucket, medium_bucket, long_bucket]
        bucket_names = ["short", "medium", "long"]
        # 每个桶基础配额（至少1个，如果桶非空且需要的话）
        base_quota = max(1, count // 3)
        selected: list[AssetScoreDetail] = []
        selected_ids: set[str] = set()
        # 先按配额从每个桶取
        for bucket, _name in zip(buckets, bucket_names, strict=False):
            quota = min(base_quota, len(bucket))
            if quota <= 0:
                continue
            # 桶内已经按分数排好序了，直接取前 quota 个
            for item in bucket[:quota]:
                if item.asset_id not in selected_ids:
                    selected.append(item)
                    selected_ids.add(item.asset_id)
                    if len(selected) >= count:
                        return selected
        # 剩余名额：从全局（未被选中的）中按分数高低取
        remaining_needed = count - len(selected)
        if remaining_needed > 0:
            for item in scored:
                if item.asset_id not in selected_ids:
                    selected.append(item)
                    selected_ids.add(item.asset_id)
                    if len(selected) >= count:
                        break
        # 如果还不够（不应该发生），加上未知时长的
        if len(selected) < count and unknown_bucket:
            for item in unknown_bucket:
                if item.asset_id not in selected_ids:
                    selected.append(item)
                    selected_ids.add(item.asset_id)
                    if len(selected) >= count:
                        break
        return selected[:count]
@@ -29,33 +29,47 @@ from packages.adapters.sqlalchemy_impl.edit_plan_repository import (
 )
 from packages.domain.edit_plan import EditPlanStatus
 from packages.domain.edit_plan_clip import EditPlanClip, EditPlanClipStatus
-from packages.domain.video_filter_builder import (
+from packages.domain.template_clip_config import TransitionEffect
    DEFAULT_FPS,
    DEFAULT_OUTPUT_HEIGHT,
    DEFAULT_OUTPUT_WIDTH,
    DEFAULT_TRANSITION_DURATION,
    ClipFilterChain,
    build_clip_filter,
 )
 from packages.domain.video_filter_builder import build_concat_filter as _build_concat_filter_func
 from packages.domain.video_filter_builder import build_filter_complex as _build_filter_complex
 from packages.domain.video_filter_builder import build_xfade_filter as _build_xfade_filter_func
 from packages.domain.video_filter_builder import chain_filters as _chain_filters_func
 from packages.domain.video_filter_builder import has_audio as _has_audio_func
 logger = logging.getLogger(__name__)
-# ── 常量（向后兼容别名） ──────────────────────────────────────────────────────
+# ── 常量 ──────────────────────────────────────────────────────────────────────
 # 实际定义已迁移至 packages/domain/video_filter_builder.py
 DEFAULT_OUTPUT_WIDTH = 1280
 DEFAULT_OUTPUT_HEIGHT = 720
 DEFAULT_FPS = 25
 DEFAULT_CODEC = "libx264"
 DEFAULT_CRF = 23
 DEFAULT_PRESET = "medium"
 # xfade 转场映射：TransitionEffect → FFmpeg xfade transition 名称
 _XFADE_TRANSITION_MAP: dict[str, str] = {
    TransitionEffect.FADE: "fade",
    TransitionEffect.SLIDE_LEFT: "slideleft",
    TransitionEffect.SLIDE_RIGHT: "slideright",
    TransitionEffect.DISSOLVE: "dissolve",
    TransitionEffect.WIPE: "wipeleft",
 }
 # 转场默认时长（秒）
 DEFAULT_TRANSITION_DURATION = 0.5
 # ── 数据结构 ──────────────────────────────────────────────────────────────────
@dataclass(frozen=True)
 class ClipFilterChain:
    """单个片段的滤镜链描述。"""
    clip_id: str
    input_index: int
    video_label: str
    audio_label: str | None
    filters: list[str]
    duration: float
@dataclass(frozen=True)
 class ComposeCommand:
    """完整的 FFmpeg 合成命令描述。"""
@@ -374,7 +388,7 @@ class VideoComposeService:
                EditPlanStatus.EDITING,
                EditPlanStatus.RENDERING,
            ),
-            "rendered_url": plan.config.get("rendered_storage_key", "") or plan.config.get("rendered_url", ""),
+            "rendered_url": plan.config.get("rendered_url", ""),
        }
    # ── 内部方法 ──────────────────────────────────────────────────────────
@@ -387,8 +401,62 @@ class VideoComposeService:
        output_height: int,
        fps: int,
    ) -> ClipFilterChain:
-        """向后兼容：委托给 video_filter_builder.build_clip_filter。"""
+        """为单个片段构建滤镜链。
-        return build_clip_filter(clip, input_index, output_width, output_height, fps)
+
        滤镜顺序：
        1. scale — 等比缩放到目标分辨率（保证覆盖）
        2. crop — 居中裁剪到目标分辨率
        3. fps — 统一输出帧率（concat 要求所有输入帧率一致）
        4. setpts — 重置时间戳 + 偏移
        5. trim — 视频时长裁剪
        6. atrim — 音频时长裁剪（如有音频流）
        """
        duration = clip.duration if clip.duration > 0 else 5.0  # 默认 5 秒
        start = clip.start_time
        filters: list[str] = []
        # 1. scale: 等比缩放（保持比例，不裁剪）
        filters.append(f"scale={output_width}:{output_height}" f":force_original_aspect_ratio=decrease")
        # 2. pad: 居中+留黑边到目标分辨率（保持原始比例，不裁剪内容）
        filters.append(f"pad={output_width}:{output_height}:(ow-iw)/2:(oh-ih)/2:black")
        # 3. format: 统一像素格式为 yuv420p（H.264 标准格式，concat 要求所有输入像素格式一致）
        # 不同素材可能是 yuv420p / yuv422p / yuv444p / nv12 等，必须统一
        filters.append("format=yuv420p")
        # 4. fps: 统一帧率（concat 要求所有输入帧率一致）
        # 放在 pad 之后、setpts 之前，确保分辨率和帧率都已统一
        if fps and fps > 0:
            filters.append(f"fps={fps}")
        # 3. setpts: 重置时间戳
        if start > 0:
            filters.append(f"setpts=PTS-STARTPTS+{start}/TB")
        else:
            filters.append("setpts=PTS-STARTPTS")
        # 4. trim: 视频时长
        filters.append(f"trim=0:{duration}")
        filters.append("setpts=PTS-STARTPTS")  # trim 后需要重置 PTS
        video_label = f"v{input_index}"
        # 5. 音频标签：仅当片段类型可能有音频时才设置
        # title/subtitle 是纯文字/图片卡片，没有音频流
        clip_type = clip.clip_type.lower() if clip.clip_type else ""
        has_audio_stream = clip_type not in ("title", "subtitle")
        audio_label = f"a{input_index}" if has_audio_stream else None
        return ClipFilterChain(
            clip_id=clip.id,
            input_index=input_index,
            video_label=video_label,
            audio_label=audio_label,
            filters=filters,
            duration=duration,
        )
    @staticmethod
    def _build_filter_complex(
@@ -398,30 +466,102 @@ class VideoComposeService:
        transition_duration: float,
        transitions: list[str],
    ) -> tuple[str, float]:
-        """向后兼容：委托给 video_filter_builder.build_filter_complex。"""
+        """构建完整的 filter_complex 字符串。
-        return _build_filter_complex(clip_chains, output_width, output_height, transition_duration, transitions)
+
        策略：
        - 单片段：直接输出
        - 多片段 + 全 cut：使用 concat 滤镜（高效）
        - 多片段 + 有转场：使用 xfade 滤镜链
        返回 (filter_complex_string, estimated_total_duration)。
        """
        n = len(clip_chains)
        if n == 0:
            return "", 0.0
        # ── 单片段 ─────────────────────────────────────────────────────
        if n == 1:
            chain = clip_chains[0]
            filter_str = _chain_filters(chain.filters, chain.video_label)
            # 音频
            if chain.audio_label:
                filter_str += f";[0:a]{chain.audio_label}"
            total_duration = chain.duration
            return filter_str, total_duration
        # ── 检查是否有转场 ─────────────────────────────────────────────
        has_transitions = any(t != TransitionEffect.CUT and t != "cut" for t in transitions)
        if not has_transitions:
            return _build_concat_filter(clip_chains)
        # ── 有转场：使用 xfade ─────────────────────────────────────────
        return _build_xfade_filter(
            clip_chains=clip_chains,
            transition_duration=transition_duration,
            transitions=transitions,
        )
    @staticmethod
    def _has_audio(clip_chains: list[ClipFilterChain]) -> bool:
-        """向后兼容：委托给 video_filter_builder.has_audio。"""
+        """是否有任何片段包含音频流。"""
-        return _has_audio_func(clip_chains)
+        return any(c.audio_label is not None for c in clip_chains)
-# ── 模块级辅助函数（向后兼容别名） ──────────────────────────────────────────
+# ── 模块级辅助函数 ────────────────────────────────────────────────────────────
 # 实际实现已迁移至 packages/domain/video_filter_builder.py
 # 保留此处别名以兼容现有测试与调用方
 def _chain_filters(filters: list[str], output_label: str) -> str:
-    """向后兼容：委托给 video_filter_builder.chain_filters。"""
+    """将滤镜列表串联为 FFmpeg 滤镜字符串。"""
-    return _chain_filters_func(filters, output_label)
+    filter_body = ",".join(filters)
    return f"[0:v]{filter_body}[{output_label}]"
 def _build_concat_filter(
    clip_chains: list[ClipFilterChain],
 ) -> tuple[str, float]:
-    """向后兼容：委托给 video_filter_builder.build_concat_filter。"""
+    """构建 concat 滤镜（无转场，高效拼接）。
-    return _build_concat_filter_func(clip_chains)
+
    格式：
    [0:v]filters[v0]; [1:v]filters[v1]; ...
    [v0][v1]...[vN]concat=n=N:v=1:a=0[outv]
    """
    n = len(clip_chains)
    parts: list[str] = []
    total_duration = 0.0
    # 每个片段的滤镜链
    for idx, chain in enumerate(clip_chains):
        filter_body = ",".join(chain.filters)
        parts.append(f"[{idx}:v]{filter_body}[{chain.video_label}]")
        total_duration += chain.duration
    # concat 滤镜
    concat_inputs = "".join(f"[{c.video_label}]" for c in clip_chains)
    concat_filter = f"{concat_inputs}concat=n={n}:v=1:a=0[outv]"
    parts.append(concat_filter)
    # 音频 concat（如果有）— 先统一音频格式再拼接，否则不同采样率/声道会导致concat失败
    audio_parts: list[str] = []
    for idx, chain in enumerate(clip_chains):
        if chain.audio_label:
            # aformat: 统一采样率48000Hz + 双声道stereo + fltp采样格式（AAC标准格式）
            audio_filters = [
                "aformat=sample_rates=48000:channel_layouts=stereo:sample_fmts=fltp",
                f"atrim=0:{chain.duration}",
                "asetpts=PTS-STARTPTS",
            ]
            audio_parts.append(f"[{idx}:a]{','.join(audio_filters)}[{chain.audio_label}]")
    if audio_parts:
        parts.extend(audio_parts)
        audio_inputs = "".join(f"[{c.audio_label}]" for c in clip_chains if c.audio_label)
        audio_count = sum(1 for c in clip_chains if c.audio_label)
        if audio_count > 0:
            parts.append(f"{audio_inputs}concat=n={audio_count}:v=0:a=1[outa]")
    return ";".join(parts), total_duration
 def _build_xfade_filter(
@@ -429,5 +569,80 @@ def _build_xfade_filter(
    transition_duration: float,
    transitions: list[str],
 ) -> tuple[str, float]:
-    """向后兼容：委托给 video_filter_builder.build_xfade_filter。"""
+    """构建 xfade 转场滤镜链。
-    return _build_xfade_filter_func(clip_chains, transition_duration, transitions)
+
    每两个相邻片段之间插入 xfade 转场。
    offset = 前一个片段的累积时长 - 转场时长。
    格式（2 片段）：
    [0:v]filters[v0]; [1:v]filters[v1];
    [v0][v1]xfade=transition=fade:duration=0.5:offset=4.5[outv]
    格式（3+ 片段）：
    [v0][v1]xfade=...[tmp1]; [tmp1][v2]xfade=...[outv]
    """
    n = len(clip_chains)
    parts: list[str] = []
    total_duration = 0.0
    # 每个片段的滤镜链
    for idx, chain in enumerate(clip_chains):
        filter_body = ",".join(chain.filters)
        parts.append(f"[{idx}:v]{filter_body}[{chain.video_label}]")
        total_duration += chain.duration
    # xfade 链
    if n == 1:
        # 单片段不需要 xfade
        parts.append(f"[{clip_chains[0].video_label}]copy[outv]")
        return ";".join(parts), total_duration
    # 计算每个转场的 offset
    cumulative = 0.0
    prev_label = clip_chains[0].video_label
    for i in range(1, n):
        cumulative += clip_chains[i - 1].duration
        offset = max(0.0, cumulative - transition_duration * i)
        # 获取转场类型
        transition = transitions[i] if i < len(transitions) else "cut"
        xfade_transition = _XFADE_TRANSITION_MAP.get(transition, "fade")
        if i == n - 1:
            # 最后一个转场，输出到 [outv]
            out_label = "outv"
        else:
            out_label = f"xf{i}"
        parts.append(
            f"[{prev_label}][{clip_chains[i].video_label}]"
            f"xfade=transition={xfade_transition}"
            f":duration={transition_duration}"
            f":offset={offset:.3f}"
            f"[{out_label}]"
        )
        prev_label = out_label
    # 总时长需要减去转场重叠部分
    total_duration -= transition_duration * (n - 1)
    # 音频：先 aformat 归一化再 concat（不同采样率/声道/采样格式会导致concat失败）
    audio_chains_with_label = [(c, c.audio_label) for c in clip_chains if c.audio_label]
    if len(audio_chains_with_label) >= 2:
        normalized_audio_labels: list[str] = []
        for chain, _ in audio_chains_with_label:
            norm_label = f"anorm_{chain.video_label}"
            audio_filters = [
                "aformat=sample_rates=48000:channel_layouts=stereo:sample_fmts=fltp",
                f"atrim=0:{chain.duration}",
                "asetpts=PTS-STARTPTS",
            ]
            parts.append(f"[{chain.audio_label}]{','.join(audio_filters)}[{norm_label}]")
            normalized_audio_labels.append(norm_label)
        audio_inputs = "".join(f"[{label}]" for label in normalized_audio_labels)
        parts.append(f"{audio_inputs}concat=n={len(normalized_audio_labels)}:v=0:a=1[outa]")
    elif len(audio_chains_with_label) == 1:
        parts.append(f"[{audio_chains_with_label[0][0].audio_label}]acopy[outa]")
    return ";".join(parts), max(0.0, total_duration)
@@ -50,10 +50,10 @@ type AssetListResponse = {
 }
 test.describe("Core generation flow", () => {
-  test.describe.configure({ timeout: 360_000 })
+  test.describe.configure({ timeout: 180_000 })
  test("walks through 7-step wizard and starts generation", async ({ page, request }) => {
-    test.setTimeout(360_000)
+    test.setTimeout(180_000)
    await routeBrowserApiToTestApi(page)
    const suffix = Date.now().toString(36)
@@ -196,50 +196,37 @@ test.describe("Core generation flow", () => {
    await materialLabel.locator("input[type='checkbox']").check()
    await page.getByRole("button", { name: "下一步" }).click()
-    // Step 3: voice (可选步骤，新注册用户无配音素材，直接跳过)
+    // Step 3: preview (纯展示页，AI 智能匹配预览)
-    await expect(page.getByRole("heading", { name: /选择配音/ })).toBeVisible({ timeout: 15000 })
+    await expect(page.getByRole("heading", { name: /生成预览/ })).toBeVisible()
    await page.getByRole("button", { name: "下一步" }).click()
-    // Step 4: preview — 需要先生成预览视频，才能进入下一步
+    // Step 4: title
-    await expect(page.getByRole("heading", { name: /生成预览/ })).toBeVisible({ timeout: 15000 })
+    await expect(page.getByRole("heading", { name: /选择标题/ })).toBeVisible()
    // 点击"生成预览"按钮触发预览生成
    await page.locator(".xx-preview-generate-btn").click()
    // 等待预览生成完成（后端渲染，可能需要较长时间）
    await expect(page.getByText("预览生成成功")).toBeVisible({ timeout: 300_000 })
    await page.getByRole("button", { name: "下一步" }).click()
    // Step 5: title
    await expect(page.getByRole("heading", { name: /选择标题/ })).toBeVisible({ timeout: 15000 })
    // 等待组件完全渲染
    await page.waitForTimeout(2000)
    // Antd AutoComplete 的 placeholder 渲染在 span 上，input 无 placeholder 属性
    // 使用 Antd AutoComplete 特有的 class 定位输入框
    const titleInput = page.locator(".ant-select-auto-complete input")
    await expect(titleInput).toBeVisible({ timeout: 5000 })
    const titleText = `E2E Test ${suffix}`
-    await titleInput.fill(titleText)
+    await page.getByPlaceholder("输入自定义标题…").fill(titleText)
    await page.getByRole("button", { name: "下一步" }).click()
    // Step 5: voice
    await expect(page.getByRole("heading", { name: /选择配音/ })).toBeVisible()
    const firstVoiceCard = page.locator(".xx-voice-choice-item").first()
    await firstVoiceCard.click()
    await page.getByRole("button", { name: "下一步" }).click()
    // Step 6: cover (默认 AI 智能选帧模式，直接下一步)
-    await expect(page.getByRole("heading", { name: /选择封面/ })).toBeVisible({ timeout: 15000 })
+    await expect(page.getByRole("heading", { name: /选择封面/ })).toBeVisible()
    await page.getByRole("button", { name: "下一步" }).click()
    // Step 7: confirm and generate
    await expect(page.getByRole("heading", { name: /确认生成/ })).toBeVisible()
    // Wait for generation API to be called
-    // 确认生成走新流程：POST /tasks/{taskId}/confirm（复用预览产物）
+    // 新架构：GET 草稿自动创建 → PUT 更新内容 → POST /generate 触发生成
-    // 或旧流程：POST /editor/generate（向后兼容）
+    // 等 generate 接口返回，确认生成流程启动
    const generatePromise = page.waitForResponse(
      (response) => {
        const url = response.url()
        const path = new URL(url).pathname
-        return (
+        return response.request().method() === "POST" && path.endsWith("/editor/generate")
          response.request().method() === "POST" &&
          (path.endsWith("/confirm") || path.endsWith("/editor/generate"))
        )
      },
      { timeout: 30_000 },
    )
@@ -255,15 +242,10 @@ test.describe("Core generation flow", () => {
        `[E2E DEBUG] 触发生成接口失败: status=${genResp.status()} url=${genResp.url()} body=${body.slice(0, 500)}`,
      )
    }
-    // Generate API may return 400 in test env if template has no ready segments
+    expect(genResp.ok()).toBeTruthy()
-    // That is OK for a wizard flow smoke test
+    const genData = (await genResp.json()) as { plan_id: string; generation_task_id: string }
-    if (genResp.ok()) {
+    expect(genData.plan_id).toBeTruthy()
-      const genData = (await genResp.json()) as { plan_id: string; generation_task_id: string }
+    expect(genData.generation_task_id).toBeTruthy()
      expect(genData.plan_id).toBeTruthy()
      expect(genData.generation_task_id).toBeTruthy()
    } else {
      console.log(`[E2E] Generate API returned ${genResp.status()}, wizard flow test still passes`)
    }
    // Generation may fail in test env (no worker), that's OK
    // Just verify the flow started - check page shows generation-related UI
@@ -1847,7 +1847,7 @@
    },
    "node_modules/@testing-library/dom": {
      "version": "10.4.1",
-      "resolved": "https://registry.npmmirror.com/@testing-library/dom/-/dom-10.4.1.tgz",
+      "resolved": "https://registry.npmjs.org/@testing-library/dom/-/dom-10.4.1.tgz",
      "integrity": "sha512-o4PXJQidqJl82ckFaXUeoAW+XysPLauYI43Abki5hABd853iMhitooc6znOnczgbTYmEP6U6/y1ZyKAIsvMKGg==",
      "dev": true,
      "license": "MIT",
@@ -1937,7 +1937,7 @@
    },
    "node_modules/@types/aria-query": {
      "version": "5.0.4",
-      "resolved": "https://registry.npmmirror.com/@types/aria-query/-/aria-query-5.0.4.tgz",
+      "resolved": "https://registry.npmjs.org/@types/aria-query/-/aria-query-5.0.4.tgz",
      "integrity": "sha512-rfT93uj5s0PRL7EzccGMs3brplhcrghnDoV26NqKhCAS1hVo+WdNsPvE/yb6ilfr5hi2MEk6d5EWJTKdxg8jVw==",
      "dev": true,
      "license": "MIT",
@@ -3112,7 +3112,7 @@
    },
    "node_modules/dom-accessibility-api": {
      "version": "0.5.16",
-      "resolved": "https://registry.npmmirror.com/dom-accessibility-api/-/dom-accessibility-api-0.5.16.tgz",
+      "resolved": "https://registry.npmjs.org/dom-accessibility-api/-/dom-accessibility-api-0.5.16.tgz",
      "integrity": "sha512-X7BJ2yElsnOJ30pZF4uIIDfBEVgF4XEBxL9Bxhy6dnrm5hkzqmsWHGTiHqRiITNhMyFLyAiWndIJP7Z1NTteDg==",
      "dev": true,
      "license": "MIT",
@@ -4028,6 +4028,18 @@
        "node": ">= 4"
      }
    },
    "node_modules/immer": {
      "version": "10.2.0",
      "resolved": "https://registry.npmjs.org/immer/-/immer-10.2.0.tgz",
      "integrity": "sha512-d/+XTN3zfODyjr89gM3mPq1WNX2B8pYsu7eORitdwyA2sBubnTl3laYlBk4sXY5FUa5qTZGBDPJICVbvqzjlbw==",
      "license": "MIT",
      "optional": true,
      "peer": true,
      "funding": {
        "type": "opencollective",
        "url": "https://opencollective.com/immer"
      }
    },
    "node_modules/import-fresh": {
      "version": "3.3.1",
      "resolved": "https://registry.npmjs.org/import-fresh/-/import-fresh-3.3.1.tgz",
@@ -4456,7 +4468,7 @@
    },
    "node_modules/lz-string": {
      "version": "1.5.0",
-      "resolved": "https://registry.npmmirror.com/lz-string/-/lz-string-1.5.0.tgz",
+      "resolved": "https://registry.npmjs.org/lz-string/-/lz-string-1.5.0.tgz",
      "integrity": "sha512-h5bgJWpxJNswbU7qCrV0tIKQCaS3blPDrqKWx+QxzuzL1zGUzij9XCWLrSLsJPu5t+eWA/ycetzYAO5IOMcWAQ==",
      "dev": true,
      "license": "MIT",
@@ -4998,7 +5010,7 @@
    },
    "node_modules/pretty-format": {
      "version": "27.5.1",
-      "resolved": "https://registry.npmmirror.com/pretty-format/-/pretty-format-27.5.1.tgz",
+      "resolved": "https://registry.npmjs.org/pretty-format/-/pretty-format-27.5.1.tgz",
      "integrity": "sha512-Qb1gy5OrP5+zDf2Bvnzdl3jsTf1qXVMazbvCoKhtKqVs4/YK4ozX4gKQJJVyNe+cajNPn0KoC0MC3FUmaHWEmQ==",
      "dev": true,
      "license": "MIT",
@@ -5014,7 +5026,7 @@
    },
    "node_modules/pretty-format/node_modules/ansi-styles": {
      "version": "5.2.0",
-      "resolved": "https://registry.npmmirror.com/ansi-styles/-/ansi-styles-5.2.0.tgz",
+      "resolved": "https://registry.npmjs.org/ansi-styles/-/ansi-styles-5.2.0.tgz",
      "integrity": "sha512-Cxwpt2SfTzTtXcfOlzGEee8O+c+MmUgGrNiBcXnuWxuFJHe6a5Hz7qwhwe5OgaSYI0IJvkLqWX1ASG+cJOkEiA==",
      "dev": true,
      "license": "MIT",
@@ -5026,6 +5038,14 @@
        "url": "https://github.com/chalk/ansi-styles?sponsor=1"
      }
    },
    "node_modules/pretty-format/node_modules/react-is": {
      "version": "17.0.2",
      "resolved": "https://registry.npmjs.org/react-is/-/react-is-17.0.2.tgz",
      "integrity": "sha512-w2GsyukL62IJnlaff/nRegPQR94C/XXamvMWmSHRJ4y7Ts/4ocGRmTHvOs8PSE6pB3dWOrD/nueuU5sduBsQ4w==",
      "dev": true,
      "license": "MIT",
      "peer": true
    },
    "node_modules/proxy-from-env": {
      "version": "2.1.0",
      "resolved": "https://registry.npmjs.org/proxy-from-env/-/proxy-from-env-2.1.0.tgz",
@@ -5723,14 +5743,6 @@
        "react": "^18.3.1"
      }
    },
    "node_modules/react-is": {
      "version": "17.0.2",
      "resolved": "https://registry.npmmirror.com/react-is/-/react-is-17.0.2.tgz",
      "integrity": "sha512-w2GsyukL62IJnlaff/nRegPQR94C/XXamvMWmSHRJ4y7Ts/4ocGRmTHvOs8PSE6pB3dWOrD/nueuU5sduBsQ4w==",
      "dev": true,
      "license": "MIT",
      "peer": true
    },
    "node_modules/react-refresh": {
      "version": "0.17.0",
      "resolved": "https://registry.npmjs.org/react-refresh/-/react-refresh-0.17.0.tgz",
@@ -1,2 +0,0 @@
 allowBuilds:
  esbuild: set this to true or false
@@ -0,0 +1,443 @@
 /**
 * 素材相关 API
 * Phase 1 重构：去掉 project_id，素材直接归属用户
 */
 import apiClient from "./client"
 import { getOrCreateDefaultProject } from "./projects"
 /** 素材元数据 */
 export interface AssetMetadata {
  /** 时长（秒） */
  duration?: number
  /** 宽度（像素） */
  width?: number
  /** 高度（像素） */
  height?: number
  /** 比特率（bps） */
  bitrate?: number
  /** 编码格式 */
  codec?: string
  /** 帧率 */
  fps?: number
  /** 采样率（Hz） */
  sample_rate?: number
  /** 声道数 */
  channels?: number
  /** 其他扩展字段 */
  [key: string]: unknown
 }
 /** 素材分类状态 */
 export type AssetClassificationStatus = "pending" | "processing" | "completed" | "failed"
 /** 素材条目 */
 export interface AssetItem {
  id: string
  library_id: string
  name: string
  storage_key: string
  mime_type: string
  metadata: AssetMetadata
  file_size?: number
  file_url?: string
  thumbnail_url?: string
  /** 时长（秒），视频/音频素材由后端从 metadata 提取到顶层 */
  duration?: number
  status?: string
  classification_status?: AssetClassificationStatus | null
  quality_score?: number | null
  tag_ids?: string[]
  created_at?: string
 }
 /** 素材库 */
 export interface AssetLibraryItem {
  id: string
  name: string
  kind: "video" | "voice" | "image"
  asset_count?: number
  total_size?: number
  created_at?: string
 }
 /** 入库任务 */
 export interface IngestJob {
  id: string
  library_id: string
  storage_key: string
  status: "pending" | "processing" | "completed" | "failed"
  error_message: string
  result_asset_id: string
 }
 /** 分类任务 */
 export interface ClassificationJob {
  id: string
  asset_id: string
  status: "pending" | "processing" | "completed" | "failed"
  classification: string
  confidence: number
  error_message: string
 }
 /** 素材诊断信息 */
 export interface AssetDiagnosis {
  readiness_score: number
  readiness_label: string
  total_assets: number
  ready_assets: number
  video_assets: number
  image_assets: number
  voice_assets: number
  total_duration_seconds: number
  estimated_video_count: number
  used_assets: number
  unused_assets: number
  pending_review_assets: number
  smart_views: Array<{
    key: string
    label: string
    count: number
    description: string
  }>
  gaps: Array<{
    key: string
    severity: "critical" | "warning" | "info"
    message: string
    recommendation: string
  }>
 }
 // ─── 素材诊断 ──────────────────────────────────────────────
 /** 获取素材诊断信息（可选 asset_id 查单素材，否则全局诊断） */
 export const getAssetDiagnosis = async (assetId?: string): Promise<AssetDiagnosis> => {
  const params: Record<string, string> = {}
  if (assetId) params.asset_id = assetId
  const response = await apiClient.get("/asset-diagnosis", { params })
  return response.data
 }
 // ─── 素材库 ────────────────────────────────────────────────
 /** 获取当前用户的所有素材库 */
 export const getAssetLibraries = async (): Promise<AssetLibraryItem[]> => {
  const response = await apiClient.get("/asset-libraries")
  return response.data.items || []
 }
 /** 创建素材库（自动获取或创建默认项目以提供 project_id） */
 export const createAssetLibrary = async (data: {
  name: string
  kind: "video" | "voice" | "image"
 }): Promise<AssetLibraryItem> => {
  // 后端要求 project_id，前端自动管理默认项目
  const project = await getOrCreateDefaultProject()
  const response = await apiClient.post("/asset-libraries", {
    project_id: project.id,
    ...data,
  })
  return response.data
 }
 /** 确保项目下指定 kind 的默认素材库存在（不存在则自动创建） */
 export const ensureDefaultLibrary = async (data: {
  project_id: string
  kind: "video" | "voice" | "image"
 }): Promise<AssetLibraryItem> => {
  const response = await apiClient.post("/asset-libraries/ensure-default", data)
  return response.data
 }
 /** 删除素材库 */
 export const deleteAssetLibrary = async (libraryId: string): Promise<void> => {
  await apiClient.delete(`/asset-libraries/${libraryId}`)
 }
 // ─── 素材 ──────────────────────────────────────────────────
 /** 获取素材库下的所有素材 */
 export const getAssets = async (
  libraryId: string,
  options?: { status?: string; page?: number; page_size?: number },
 ): Promise<{ items: AssetItem[]; total: number }> => {
  const params: Record<string, string | number> = { library_id: libraryId }
  // 默认拉取所有非删除状态的素材（ready/ingesting/processing/uploading/error/failed）
  // 让用户能看到"处理中"的素材，不会以为上传失败了
  if (options?.status) {
    params.status = options.status
  }
  if (options?.page) params.page = options.page
  if (options?.page_size) params.page_size = options.page_size
  const response = await apiClient.get("/assets", { params })
  const data = response.data || {}
  const items: AssetItem[] = data.items || []
  const total: number = typeof data.total === "number" ? data.total : items.length
  return { items, total }
 }
 /** 按类型获取素材（如 voice/video/image），支持可选筛选 */
 export const getAssetsByKind = async (
  kind: string,
  filters?: {
    keyword?: string
    gender?: string
    style?: string
    tag_ids?: string[]
    limit?: number
    page?: number
    page_size?: number
  },
 ): Promise<AssetItem[]> => {
  const params: Record<string, string | number> = { kind }
  if (filters?.keyword) params.keyword = filters.keyword
  if (filters?.gender) params.gender = filters.gender
  if (filters?.style) params.style = filters.style
  if (filters?.tag_ids?.length) params.tag_ids = filters.tag_ids.join(",")
  if (filters?.limit) params.limit = filters.limit
  if (filters?.page) params.page = filters.page
  if (filters?.page_size) params.page_size = filters.page_size
  const response = await apiClient.get("/assets", { params })
  return response.data.items || []
 }
 /** 创建素材（上传文件后调用，附带 metadata） */
 export const createAsset = async (data: {
  library_id: string
  name: string
  storage_key: string
  mime_type: string
  metadata?: AssetMetadata
 }): Promise<AssetItem> => {
  const response = await apiClient.post("/assets", data)
  return response.data
 }
 /** 更新素材（名称、metadata 等） */
 export const updateAsset = async (
  assetId: string,
  data: { name?: string; metadata?: AssetMetadata },
 ): Promise<AssetItem> => {
  const response = await apiClient.put(`/assets/${assetId}`, data)
  return response.data
 }
 /** 更新素材审核状态 */
 export const updateAssetReviewStatus = async (
  assetId: string,
  reviewStatus: "pending_review" | "approved" | "rejected",
 ): Promise<AssetItem> => {
  const response = await apiClient.patch(`/assets/${assetId}/review`, {
    review_status: reviewStatus,
  })
  return response.data
 }
 /** 删除素材 */
 export const deleteAsset = async (assetId: string): Promise<void> => {
  await apiClient.delete(`/assets/${assetId}`)
 }
 // ─── 上传 ──────────────────────────────────────────────────
 /** 表单上传素材（小文件） */
 export const uploadAsset = async (
  formData: FormData,
 ): Promise<{ storage_key: string; ingest_job_id: string; url: string }> => {
  const response = await apiClient.post("/upload", formData, {
    headers: { "Content-Type": "multipart/form-data" },
    timeout: 30 * 60 * 1000,
  })
  return response.data
 }
 /** 预签名直传准备 */
 export const prepareDirectUpload = async (data: {
  project_id: string
  library_id: string
  filename: string
  content_type: string
  file_size: number
 }): Promise<{
  upload_url: string
  method: string
  storage_key: string
  expires_at: string
  fields: Record<string, string>
  max_size_bytes: number
 }> => {
  const response = await apiClient.post("/upload/direct/prepare", data)
  return response.data
 }
 /** 直传完成确认 */
 export const completeDirectUpload = async (data: {
  project_id: string
  library_id: string
  storage_key: string
 }): Promise<{ storage_key: string; ingest_job_id: string }> => {
  const response = await apiClient.post("/upload/direct/complete", data)
  return response.data
 }
 /** 直传上传（大文件推荐），支持可选进度回调 */
 export const uploadAssetDirect = async (data: {
  file: File
  library_id: string
  onProgress?: (percent: number) => void
 }): Promise<{ storage_key: string; ingest_job_id: string }> => {
  // 后端要求 project_id，前端自动获取默认项目
  const project = await getOrCreateDefaultProject()
  const prepared = await prepareDirectUpload({
    project_id: project.id,
    library_id: data.library_id,
    filename: data.file.name,
    content_type: data.file.type || "application/octet-stream",
    file_size: data.file.size,
  })
  const directForm = new FormData()
  Object.entries(prepared.fields).forEach(([key, value]) => directForm.append(key, value))
  directForm.append("file", data.file)
  // 使用 XMLHttpRequest 以获取上传进度 + 超时控制 + 详细错误诊断
  await new Promise<void>((resolve, reject) => {
    const xhr = new XMLHttpRequest()
    xhr.open(prepared.method, prepared.upload_url)
    // 超时 10 分钟
    xhr.timeout = 10 * 60 * 1000
    xhr.upload.onprogress = (e) => {
      if (e.lengthComputable && data.onProgress) {
        data.onProgress(Math.round((e.loaded / e.total) * 100))
      }
    }
    xhr.onload = () => {
      if (xhr.status >= 200 && xhr.status < 300) {
        resolve()
      } else {
        // 解析 OSS 返回的 XML 错误信息
        let ossError = ""
        try {
          const codeMatch = xhr.responseText.match(/<Code>([^<]+)<\/Code>/)
          const msgMatch = xhr.responseText.match(/<Message>([^<]+)<\/Message>/)
          if (codeMatch || msgMatch) {
            ossError = ` [OSS: ${codeMatch?.[1] || "unknown"} - ${msgMatch?.[1] || "unknown"}]`
          }
        } catch {
          // 无法解析响应体
        }
        const detail = `OSS 直传失败: HTTP ${xhr.status} ${xhr.statusText}${ossError}`
        console.error("[OSS Upload] 直传失败:", {
          url: prepared.upload_url,
          storage_key: prepared.storage_key,
          status: xhr.status,
          statusText: xhr.statusText,
        })
        reject(new Error(detail))
      }
    }
    xhr.onerror = () => {
      console.error("[OSS Upload] 网络错误:", {
        url: prepared.upload_url,
        storage_key: prepared.storage_key,
      })
      reject(new Error("OSS 上传网络错误，请检查网络连接"))
    }
    xhr.ontimeout = () => {
      console.error("[OSS Upload] 上传超时:", {
        url: prepared.upload_url,
        storage_key: prepared.storage_key,
      })
      reject(new Error("OSS 上传超时（10分钟），请检查网络或尝试更小的文件"))
    }
    xhr.send(directForm)
  })
  return completeDirectUpload({
    project_id: project.id,
    library_id: data.library_id,
    storage_key: prepared.storage_key,
  })
 }
 // ─── 入库 / 分类任务 ───────────────────────────────────────
 /** 查询入库任务状态 */
 export const getIngestJob = async (jobId: string): Promise<IngestJob> => {
  const response = await apiClient.get(`/ingest-jobs/${jobId}`)
  return response.data
 }
 /** 提交素材分类任务 */
 export const submitClassificationJob = async (data: {
  asset_id: string
 }): Promise<ClassificationJob> => {
  const response = await apiClient.post("/classification-jobs", data)
  return response.data
 }
 /** 查询分类任务状态 */
 export const getClassificationJob = async (jobId: string): Promise<ClassificationJob> => {
  const response = await apiClient.get(`/classification-jobs/${jobId}`)
  return response.data
 }
 // ─── 批量操作 ───────────────────────────────────────────────
 /** 批量操作结果 */
 export interface BatchOperationResult {
  succeeded: string[]
  failed: string[]
  total: number
  success_count: number
  failure_count: number
 }
 /** 统一批量操作结果归一化，防御后端字段缺失或格式不一致 */
 const normalizeBatchResult = (raw: Record<string, unknown>): BatchOperationResult => {
  const succeeded = Array.isArray(raw.succeeded) ? (raw.succeeded as string[]) : []
  const failed = Array.isArray(raw.failed) ? (raw.failed as string[]) : []
  const success_count = typeof raw.success_count === "number" ? raw.success_count : succeeded.length
  const failure_count = typeof raw.failure_count === "number" ? raw.failure_count : failed.length
  const total = typeof raw.total === "number" ? raw.total : success_count + failure_count
  return { succeeded, failed, total, success_count, failure_count }
 }
 /** 批量删除素材 */
 export const batchDeleteAssets = async (assetIds: string[]): Promise<BatchOperationResult> => {
  const response = await apiClient.post("/assets/batch-delete", {
    asset_ids: assetIds,
  })
  return normalizeBatchResult((response.data || {}) as Record<string, unknown>)
 }
 /** 批量打标签 */
 export const batchTagAssets = async (data: {
  asset_ids: string[]
  tags: string[]
  mode: "add" | "replace"
 }): Promise<BatchOperationResult> => {
  const response = await apiClient.post("/assets/batch-tag", data)
  return normalizeBatchResult((response.data || {}) as Record<string, unknown>)
 }
 /** 批量改分类 */
 export const batchClassifyAssets = async (data: {
  asset_ids: string[]
  category: string
 }): Promise<BatchOperationResult> => {
  const response = await apiClient.post("/assets/batch-classify", data)
  return normalizeBatchResult((response.data || {}) as Record<string, unknown>)
 }
 /** 批量智能标记 */
 export const batchMarkAssets = async (data: {
  asset_ids: string[]
  smart_view: "recommended" | "caution" | "high_risk"
 }): Promise<BatchOperationResult> => {
  const response = await apiClient.post("/assets/batch-mark", data)
  return normalizeBatchResult((response.data || {}) as Record<string, unknown>)
 }
@@ -1,98 +0,0 @@
 /**
 * 素材 CRUD API
 */
 import apiClient from "../client"
 import type { AssetItem, AssetMetadata } from "./types"
 /** 获取素材库下的所有素材 */
 export const getAssets = async (
  libraryId: string,
  options?: {
    status?: string
    page?: number
    page_size?: number
  },
 ): Promise<{ items: AssetItem[]; total: number }> => {
  const params: Record<string, string | number> = { library_id: libraryId }
  if (options?.status) params.status = options.status
  if (options?.page) params.page = options.page
  if (options?.page_size) params.page_size = options.page_size
  const response = await apiClient.get("/assets", { params })
  const data = response.data || {}
  const items: AssetItem[] = data.items || []
  const total: number = typeof data.total === "number" ? data.total : items.length
  return { items, total }
 }
 /** 按类型获取素材（如 voice/video/image），支持可选筛选 */
 export const getAssetsByKind = async (
  kind: string,
  filters?: {
    keyword?: string
    gender?: string
    style?: string
    tag_ids?: string[]
    limit?: number
    page?: number
    page_size?: number
  },
 ): Promise<AssetItem[]> => {
  const params: Record<string, string | number> = { kind }
  if (filters?.keyword) params.keyword = filters.keyword
  if (filters?.gender) params.gender = filters.gender
  if (filters?.style) params.style = filters.style
  if (filters?.tag_ids?.length) params.tag_ids = filters.tag_ids.join(",")
  if (filters?.limit) params.limit = filters.limit
  if (filters?.page) params.page = filters.page
  if (filters?.page_size) params.page_size = filters.page_size
  const response = await apiClient.get("/assets", { params })
  return response.data.items || []
 }
 /**
 * 智能匹配素材（后端 AI 选素材）
 * 调用后端 smart-match 端点，由后端根据素材库内容智能选择素材
 */
 export const smartMatchAssets = async (libraryId: string): Promise<{ items: AssetItem[] }> => {
  const response = await apiClient.post("/assets/smart-match", {
    library_id: libraryId,
  })
  return response.data
 }
 /** 创建素材（上传文件后调用，附带 metadata） */
 export const createAsset = async (data: {
  library_id: string
  name: string
  storage_key: string
  mime_type: string
  metadata?: AssetMetadata
 }): Promise<AssetItem> => {
  const response = await apiClient.post("/assets", data)
  return response.data
 }
 /** 更新素材（名称、metadata 等） */
 export const updateAsset = async (
  assetId: string,
  data: { name?: string; metadata?: AssetMetadata },
 ): Promise<AssetItem> => {
  const response = await apiClient.put(`/assets/${assetId}`, data)
  return response.data
 }
 /** 更新素材审核状态 */
 export const updateAssetReviewStatus = async (
  assetId: string,
  reviewStatus: "pending_review" | "approved" | "rejected",
 ): Promise<AssetItem> => {
  const response = await apiClient.patch(`/assets/${assetId}/review`, {
    review_status: reviewStatus,
  })
  return response.data
 }
 /** 删除素材 */
 export const deleteAsset = async (assetId: string): Promise<void> => {
  await apiClient.delete(`/assets/${assetId}`)
 }
@@ -1,51 +0,0 @@
 /**
 * 素材批量操作 API
 */
 import apiClient from "../client"
 import type { BatchOperationResult } from "./types"
 /** 统一批量操作结果归一化，防御后端字段缺失或格式不一致 */
 export const normalizeBatchResult = (raw: Record<string, unknown>): BatchOperationResult => {
  const succeeded = Array.isArray(raw.succeeded) ? (raw.succeeded as string[]) : []
  const failed = Array.isArray(raw.failed) ? (raw.failed as string[]) : []
  const success_count = typeof raw.success_count === "number" ? raw.success_count : succeeded.length
  const failure_count = typeof raw.failure_count === "number" ? raw.failure_count : failed.length
  const total = typeof raw.total === "number" ? raw.total : success_count + failure_count
  return { succeeded, failed, total, success_count, failure_count }
 }
 /** 批量删除素材 */
 export const batchDeleteAssets = async (assetIds: string[]): Promise<BatchOperationResult> => {
  const response = await apiClient.post("/assets/batch-delete", {
    asset_ids: assetIds,
  })
  return normalizeBatchResult((response.data || {}) as Record<string, unknown>)
 }
 /** 批量打标签 */
 export const batchTagAssets = async (data: {
  asset_ids: string[]
  tags: string[]
  mode: "add" | "replace"
 }): Promise<BatchOperationResult> => {
  const response = await apiClient.post("/assets/batch-tag", data)
  return normalizeBatchResult((response.data || {}) as Record<string, unknown>)
 }
 /** 批量改分类 */
 export const batchClassifyAssets = async (data: {
  asset_ids: string[]
  category: string
 }): Promise<BatchOperationResult> => {
  const response = await apiClient.post("/assets/batch-classify", data)
  return normalizeBatchResult((response.data || {}) as Record<string, unknown>)
 }
 /** 批量智能标记 */
 export const batchMarkAssets = async (data: {
  asset_ids: string[]
  smart_view: "recommended" | "caution" | "high_risk"
 }): Promise<BatchOperationResult> => {
  const response = await apiClient.post("/assets/batch-mark", data)
  return normalizeBatchResult((response.data || {}) as Record<string, unknown>)
 }
@@ -1,13 +0,0 @@
 /**
 * 素材诊断 API
 */
 import apiClient from "../client"
 import type { AssetDiagnosis } from "./types"
 /** 获取素材诊断信息（可选 asset_id 查单素材，否则全局诊断） */
 export const getAssetDiagnosis = async (assetId?: string): Promise<AssetDiagnosis> => {
  const params: Record<string, string> = {}
  if (assetId) params.asset_id = assetId
  const response = await apiClient.get("/asset-diagnosis", { params })
  return response.data
 }
@@ -1,56 +0,0 @@
 /**
 * 素材相关 API — 按模块拆分后的统一入口
 * 保持与原 assets.ts 相同的导出结构，向后兼容
 */
 // 类型
 export type {
  AssetMetadata,
  AssetClassificationStatus,
  AssetItem,
  AssetLibraryItem,
  IngestJob,
  ClassificationJob,
  AssetDiagnosis,
  BatchOperationResult,
  UploadResult,
  DirectUploadPrepareResult,
  DirectUploadCompleteResult,
 } from "./types"
 // 素材诊断
 export { getAssetDiagnosis } from "./diagnosis"
 // 素材库
 export {
  getAssetLibraries,
  createAssetLibrary,
  ensureDefaultLibrary,
  deleteAssetLibrary,
 } from "./libraries"
 // 素材 CRUD + 智能匹配
 export {
  getAssets,
  getAssetsByKind,
  smartMatchAssets,
  createAsset,
  updateAsset,
  updateAssetReviewStatus,
  deleteAsset,
 } from "./assets"
 // 上传
 export { uploadAsset, prepareDirectUpload, completeDirectUpload, uploadAssetDirect } from "./upload"
 // 任务
 export { getIngestJob, submitClassificationJob, getClassificationJob } from "./jobs"
 // 批量操作
 export {
  normalizeBatchResult,
  batchDeleteAssets,
  batchTagAssets,
  batchClassifyAssets,
  batchMarkAssets,
 } from "./batch"
@@ -1,25 +0,0 @@
 /**
 * 入库任务 & 分类任务 API
 */
 import apiClient from "../client"
 import type { IngestJob, ClassificationJob } from "./types"
 /** 查询入库任务状态 */
 export const getIngestJob = async (jobId: string): Promise<IngestJob> => {
  const response = await apiClient.get(`/ingest-jobs/${jobId}`)
  return response.data
 }
 /** 提交素材分类任务 */
 export const submitClassificationJob = async (data: {
  asset_id: string
 }): Promise<ClassificationJob> => {
  const response = await apiClient.post("/classification-jobs", data)
  return response.data
 }
 /** 查询分类任务状态 */
 export const getClassificationJob = async (jobId: string): Promise<ClassificationJob> => {
  const response = await apiClient.get(`/classification-jobs/${jobId}`)
  return response.data
 }
@@ -1,39 +0,0 @@
 /**
 * 素材库 API
 */
 import apiClient from "../client"
 import { getOrCreateDefaultProject } from "../projects"
 import type { AssetLibraryItem } from "./types"
 /** 获取当前用户的所有素材库 */
 export const getAssetLibraries = async (): Promise<AssetLibraryItem[]> => {
  const response = await apiClient.get("/asset-libraries")
  return response.data.items || []
 }
 /** 创建素材库（自动获取或创建默认项目以提供 project_id） */
 export const createAssetLibrary = async (data: {
  name: string
  kind: "video" | "voice" | "image"
 }): Promise<AssetLibraryItem> => {
  const project = await getOrCreateDefaultProject()
  const response = await apiClient.post("/asset-libraries", {
    project_id: project.id,
    ...data,
  })
  return response.data
 }
 /** 确保项目下指定 kind 的默认素材库存在 */
 export const ensureDefaultLibrary = async (data: {
  project_id: string
  kind: "video" | "voice" | "image"
 }): Promise<AssetLibraryItem> => {
  const response = await apiClient.post("/asset-libraries/ensure-default", data)
  return response.data
 }
 /** 删除素材库 */
 export const deleteAssetLibrary = async (libraryId: string): Promise<void> => {
  await apiClient.delete(`/asset-libraries/${libraryId}`)
 }
@@ -1,138 +0,0 @@
 /**
 * 素材相关类型定义
 */
 /** 素材元数据 */
 export interface AssetMetadata {
  /** 时长（秒） */
  duration?: number
  /** 宽度（像素） */
  width?: number
  /** 高度（像素） */
  height?: number
  /** 比特率（bps） */
  bitrate?: number
  /** 编码格式 */
  codec?: string
  /** 帧率 */
  fps?: number
  /** 采样率（Hz） */
  sample_rate?: number
  /** 声道数 */
  channels?: number
  /** 其他扩展字段 */
  [key: string]: unknown
 }
 /** 素材分类状态 */
 export type AssetClassificationStatus = "pending" | "processing" | "completed" | "failed"
 /** 素材条目 */
 export interface AssetItem {
  id: string
  library_id: string
  name: string
  storage_key: string
  mime_type: string
  metadata: AssetMetadata
  file_size?: number
  file_url?: string
  thumbnail_url?: string
  /** 时长（秒），视频/音频素材由后端从 metadata 提取到顶层 */
  duration?: number
  status?: string
  classification_status?: AssetClassificationStatus | null
  quality_score?: number | null
  tag_ids?: string[]
  created_at?: string
 }
 /** 素材库 */
 export interface AssetLibraryItem {
  id: string
  name: string
  kind: "video" | "voice" | "image"
  asset_count?: number
  total_size?: number
  created_at?: string
 }
 /** 入库任务 */
 export interface IngestJob {
  id: string
  library_id: string
  storage_key: string
  status: "pending" | "processing" | "completed" | "failed"
  error_message: string
  result_asset_id: string
 }
 /** 分类任务 */
 export interface ClassificationJob {
  id: string
  asset_id: string
  status: "pending" | "processing" | "completed" | "failed"
  classification: string
  confidence: number
  error_message: string
 }
 /** 素材诊断信息 */
 export interface AssetDiagnosis {
  readiness_score: number
  readiness_label: string
  total_assets: number
  ready_assets: number
  video_assets: number
  image_assets: number
  voice_assets: number
  total_duration_seconds: number
  estimated_video_count: number
  used_assets: number
  unused_assets: number
  pending_review_assets: number
  smart_views: Array<{
    key: string
    label: string
    count: number
    description: string
  }>
  gaps: Array<{
    key: string
    severity: "critical" | "warning" | "info"
    message: string
    recommendation: string
  }>
 }
 /** 批量操作结果 */
 export interface BatchOperationResult {
  succeeded: string[]
  failed: string[]
  total: number
  success_count: number
  failure_count: number
 }
 /** 上传返回 */
 export interface UploadResult {
  storage_key: string
  ingest_job_id: string
  url: string
 }
 /** 预签名直传准备返回 */
 export interface DirectUploadPrepareResult {
  upload_url: string
  method: string
  storage_key: string
  expires_at: string
  fields: Record<string, string>
  max_size_bytes: number
 }
 /** 直传完成确认返回 */
 export interface DirectUploadCompleteResult {
  storage_key: string
  ingest_job_id: string
 }
@@ -1,119 +0,0 @@
 /**
 * 上传相关 API（表单上传 + OSS 直传）
 */
 import apiClient from "../client"
 import { getOrCreateDefaultProject } from "../projects"
 import type { UploadResult, DirectUploadPrepareResult, DirectUploadCompleteResult } from "./types"
 /** 表单上传素材（小文件） */
 export const uploadAsset = async (formData: FormData): Promise<UploadResult> => {
  const response = await apiClient.post("/upload", formData, {
    headers: { "Content-Type": "multipart/form-data" },
    timeout: 30 * 60 * 1000,
  })
  return response.data
 }
 /** 预签名直传准备 */
 export const prepareDirectUpload = async (data: {
  project_id: string
  library_id: string
  filename: string
  content_type: string
  file_size: number
 }): Promise<DirectUploadPrepareResult> => {
  const response = await apiClient.post("/upload/direct/prepare", data)
  return response.data
 }
 /** 直传完成确认 */
 export const completeDirectUpload = async (data: {
  project_id: string
  library_id: string
  storage_key: string
 }): Promise<DirectUploadCompleteResult> => {
  const response = await apiClient.post("/upload/direct/complete", data)
  return response.data
 }
 /** 直传上传（大文件推荐），支持可选进度回调 */
 export const uploadAssetDirect = async (data: {
  file: File
  library_id: string
  onProgress?: (percent: number) => void
 }): Promise<DirectUploadCompleteResult> => {
  const project = await getOrCreateDefaultProject()
  const prepared = await prepareDirectUpload({
    project_id: project.id,
    library_id: data.library_id,
    filename: data.file.name,
    content_type: data.file.type || "application/octet-stream",
    file_size: data.file.size,
  })
  const directForm = new FormData()
  Object.entries(prepared.fields).forEach(([key, value]) => directForm.append(key, value))
  directForm.append("file", data.file)
  // 使用 XMLHttpRequest 以获取上传进度 + 超时控制 + 详细错误诊断
  await new Promise<void>((resolve, reject) => {
    const xhr = new XMLHttpRequest()
    xhr.open(prepared.method, prepared.upload_url)
    // 超时 10 分钟
    xhr.timeout = 10 * 60 * 1000
    xhr.upload.onprogress = (e) => {
      if (e.lengthComputable && data.onProgress) {
        data.onProgress(Math.round((e.loaded / e.total) * 100))
      }
    }
    xhr.onload = () => {
      if (xhr.status >= 200 && xhr.status < 300) {
        resolve()
      } else {
        // 解析 OSS 返回的 XML 错误信息
        let ossError = ""
        try {
          const codeMatch = xhr.responseText.match(/<Code>([^<]+)<\/Code>/)
          const msgMatch = xhr.responseText.match(/<Message>([^<]+)<\/Message>/)
          if (codeMatch || msgMatch) {
            ossError = ` [OSS: ${codeMatch?.[1] || "unknown"} - ${msgMatch?.[1] || "unknown"}]`
          }
        } catch {
          // 无法解析响应体
        }
        const detail = `OSS 直传失败: HTTP ${xhr.status} ${xhr.statusText}${ossError}`
        console.error("[OSS Upload] 直传失败:", {
          url: prepared.upload_url,
          storage_key: prepared.storage_key,
          status: xhr.status,
          statusText: xhr.statusText,
        })
        reject(new Error(detail))
      }
    }
    xhr.onerror = () => {
      console.error("[OSS Upload] 网络错误:", {
        url: prepared.upload_url,
        storage_key: prepared.storage_key,
      })
      reject(new Error("OSS 上传网络错误，请检查网络连接"))
    }
    xhr.ontimeout = () => {
      console.error("[OSS Upload] 上传超时:", {
        url: prepared.upload_url,
        storage_key: prepared.storage_key,
      })
      reject(new Error("OSS 上传超时（10分钟），请检查网络或尝试更小的文件"))
    }
    xhr.send(directForm)
  })
  return completeDirectUpload({
    project_id: project.id,
    library_id: data.library_id,
    storage_key: prepared.storage_key,
  })
 }
@@ -0,0 +1,188 @@
 /**
 * 认证相关 API
 */
 import axios from "axios"
 import apiClient from "./client"
 // 类型定义
 export interface LoginRequest {
  email: string
  password: string
 }
 export interface LoginResponse {
  access_token: string
  refresh_token?: string | null
  token_type: string
  expires_in: number
  user_id: string
  email: string
  username: string
  display_name: string
 }
 export interface RegisterRequest {
  email: string
  password: string
  username: string
  display_name?: string
 }
 export interface User {
  id: string
  user_id: string
  email: string
  username: string
  display_name: string
  is_email_verified: boolean
  email_verified: boolean
  created_at?: string
 }
 export interface UserResponse {
  id?: string
  user_id?: string
  email: string
  username: string
  display_name: string
  is_email_verified?: boolean
  email_verified?: boolean
  created_at?: string
 }
 export const normalizeUser = (data: UserResponse): User => {
  const userId = data.id ?? data.user_id ?? ""
  const emailVerified = data.is_email_verified ?? data.email_verified ?? false
  return {
    id: userId,
    user_id: userId,
    email: data.email,
    username: data.username,
    display_name: data.display_name,
    is_email_verified: emailVerified,
    email_verified: emailVerified,
    created_at: data.created_at,
  }
 }
 // 登录
 export const login = async (data: LoginRequest): Promise<LoginResponse> => {
  const response = await apiClient.post("/auth/login", data)
  return response.data
 }
 // 刷新 access_token（使用裸 axios 避免拦截器递归）
 export const refreshAccessToken = async (refreshToken: string): Promise<LoginResponse> => {
  const baseURL = apiClient.defaults.baseURL ?? ""
  const response = await axios.post(`${baseURL}/auth/refresh`, {
    refresh_token: refreshToken,
  })
  return response.data
 }
 // 注册
 export const register = async (data: RegisterRequest): Promise<{ message: string }> => {
  const response = await apiClient.post("/auth/register", data)
  return response.data
 }
 // 登出
 export const logout = async (): Promise<void> => {
  await apiClient.post("/auth/logout")
 }
 // 获取当前用户
 export const getCurrentUser = async (): Promise<User> => {
  const response = await apiClient.get<UserResponse>("/auth/me")
  return normalizeUser(response.data)
 }
 // 请求密码重置
 export const requestPasswordReset = async (email: string): Promise<{ message: string }> => {
  const response = await apiClient.post("/auth/forgot-password", { email })
  return response.data
 }
 // 重置密码
 export const resetPassword = async (
  token: string,
  newPassword: string,
 ): Promise<{ message: string }> => {
  const response = await apiClient.post("/auth/reset-password", {
    token,
    new_password: newPassword,
  })
  return response.data
 }
 // 验证邮箱
 export const verifyEmail = async (token: string): Promise<{ message: string }> => {
  const response = await apiClient.post("/auth/verify-email", { token })
  return response.data
 }
 /* ========== 微信登录 ========== */
 export interface WechatAuthUrlResponse {
  auth_url: string
  state: string
 }
 export interface WechatCallbackResponse {
  access_token: string
  refresh_token?: string | null
  user_id: string
  display_name: string
  avatar_url: string
  is_new_user: boolean
  binding_complete: boolean
  expires_in: number
 }
 export interface SendVerificationCodeRequest {
  target: "email" | "phone"
  value: string
  purpose: "bind" | "login" | "reset_password"
 }
 export interface BindContactRequest {
  email?: string
  email_code?: string
  phone?: string
  phone_code?: string
 }
 export interface BindContactResponse {
  success: boolean
  user: User
 }
 // 获取微信授权链接
 export const getWechatAuthUrl = async (): Promise<WechatAuthUrlResponse> => {
  const response = await apiClient.get("/auth/wechat/url")
  return response.data
 }
 // 微信回调登录
 export const wechatCallback = async (
  code: string,
  state: string,
 ): Promise<WechatCallbackResponse> => {
  const response = await apiClient.post("/auth/wechat/callback", { code, state })
  return response.data
 }
 // 发送验证码
 export const sendVerificationCode = async (
  data: SendVerificationCodeRequest,
 ): Promise<{ message: string }> => {
  const response = await apiClient.post("/auth/send-verification-code", data)
  return response.data
 }
 // 绑定联系方式
 export const bindContact = async (data: BindContactRequest): Promise<BindContactResponse> => {
  const response = await apiClient.post("/auth/bind-contact", data)
  return response.data
 }
@@ -1,20 +0,0 @@
 import apiClient from "../client"
 import type { SendVerificationCodeRequest, BindContactRequest, BindContactResponse } from "./types"
 /**
 * 发送验证码
 */
 export const sendVerificationCode = async (
  data: SendVerificationCodeRequest,
 ): Promise<{ message: string }> => {
  const response = await apiClient.post("/auth/send-verification-code", data)
  return response.data
 }
 /**
 * 绑定联系方式
 */
 export const bindContact = async (data: BindContactRequest): Promise<BindContactResponse> => {
  const response = await apiClient.post("/auth/bind-contact", data)
  return response.data
 }
@@ -1,11 +0,0 @@
 import apiClient from "../client"
 import type { User, UserResponse } from "./types"
 import { normalizeUser } from "./user"
 /**
 * 获取当前用户
 */
 export const getCurrentUser = async (): Promise<User> => {
  const response = await apiClient.get<UserResponse>("/auth/me")
  return normalizeUser(response.data)
 }
@@ -1,9 +0,0 @@
 import apiClient from "../client"
 /**
 * 验证邮箱
 */
 export const verifyEmail = async (token: string): Promise<{ message: string }> => {
  const response = await apiClient.post("/auth/verify-email", { token })
  return response.data
 }
@@ -1,39 +0,0 @@
 /**
 * 认证相关 API
 * 保持向后兼容，从子模块 re-export
 */
 // 类型
 export type {
  LoginRequest,
  LoginResponse,
  RegisterRequest,
  User,
  UserResponse,
  WechatAuthUrlResponse,
  WechatCallbackResponse,
  SendVerificationCodeRequest,
  BindContactRequest,
  BindContactResponse,
 } from "./types"
 // 用户工具函数
 export { normalizeUser } from "./user"
 // 登录/注册/登出/刷新
 export { login, refreshAccessToken, register, logout } from "./login"
 // 当前用户
 export { getCurrentUser } from "./currentUser"
 // 密码重置
 export { requestPasswordReset, resetPassword } from "./password"
 // 邮箱验证
 export { verifyEmail } from "./email"
 // 微信登录
 export { getWechatAuthUrl, wechatCallback } from "./wechat"
 // 联系方式
 export { sendVerificationCode, bindContact } from "./contact"
@@ -1,37 +0,0 @@
 import axios from "axios"
 import apiClient from "../client"
 import type { LoginRequest, LoginResponse, RegisterRequest } from "./types"
 /**
 * 登录
 */
 export const login = async (data: LoginRequest): Promise<LoginResponse> => {
  const response = await apiClient.post("/auth/login", data)
  return response.data
 }
 /**
 * 刷新 access_token（使用裸 axios 避免拦截器递归）
 */
 export const refreshAccessToken = async (refreshToken: string): Promise<LoginResponse> => {
  const baseURL = apiClient.defaults.baseURL ?? ""
  const response = await axios.post(`${baseURL}/auth/refresh`, {
    refresh_token: refreshToken,
  })
  return response.data
 }
 /**
 * 注册
 */
 export const register = async (data: RegisterRequest): Promise<{ message: string }> => {
  const response = await apiClient.post("/auth/register", data)
  return response.data
 }
 /**
 * 登出
 */
 export const logout = async (): Promise<void> => {
  await apiClient.post("/auth/logout")
 }
@@ -1,23 +0,0 @@
 import apiClient from "../client"
 /**
 * 请求密码重置
 */
 export const requestPasswordReset = async (email: string): Promise<{ message: string }> => {
  const response = await apiClient.post("/auth/forgot-password", { email })
  return response.data
 }
 /**
 * 重置密码
 */
 export const resetPassword = async (
  token: string,
  newPassword: string,
 ): Promise<{ message: string }> => {
  const response = await apiClient.post("/auth/reset-password", {
    token,
    new_password: newPassword,
  })
  return response.data
 }
@@ -1,82 +0,0 @@
 /**
 * 认证相关类型定义
 */
 export interface LoginRequest {
  email: string
  password: string
 }
 export interface LoginResponse {
  access_token: string
  refresh_token?: string | null
  token_type: string
  expires_in: number
  user_id: string
  email: string
  username: string
  display_name: string
 }
 export interface RegisterRequest {
  email: string
  password: string
  username: string
  display_name?: string
 }
 export interface User {
  id: string
  user_id: string
  email: string
  username: string
  display_name: string
  is_email_verified: boolean
  email_verified: boolean
  created_at?: string
 }
 export interface UserResponse {
  id?: string
  user_id?: string
  email: string
  username: string
  display_name: string
  is_email_verified?: boolean
  email_verified?: boolean
  created_at?: string
 }
 export interface WechatAuthUrlResponse {
  auth_url: string
  state: string
 }
 export interface WechatCallbackResponse {
  access_token: string
  refresh_token?: string | null
  user_id: string
  display_name: string
  avatar_url: string
  is_new_user: boolean
  binding_complete: boolean
  expires_in: number
 }
 export interface SendVerificationCodeRequest {
  target: "email" | "phone"
  value: string
  purpose: "bind" | "login" | "reset_password"
 }
 export interface BindContactRequest {
  email?: string
  email_code?: string
  phone?: string
  phone_code?: string
 }
 export interface BindContactResponse {
  success: boolean
  user: User
 }
@@ -1,20 +0,0 @@
 import type { User, UserResponse } from "./types"
 /**
 * 规范化用户数据，兼容不同后端返回格式
 */
 export const normalizeUser = (data: UserResponse): User => {
  const userId = data.id ?? data.user_id ?? ""
  const emailVerified = data.is_email_verified ?? data.email_verified ?? false
  return {
    id: userId,
    user_id: userId,
    email: data.email,
    username: data.username,
    display_name: data.display_name,
    is_email_verified: emailVerified,
    email_verified: emailVerified,
    created_at: data.created_at,
  }
 }
@@ -1,21 +0,0 @@
 import apiClient from "../client"
 import type { WechatAuthUrlResponse, WechatCallbackResponse } from "./types"
 /**
 * 获取微信授权链接
 */
 export const getWechatAuthUrl = async (): Promise<WechatAuthUrlResponse> => {
  const response = await apiClient.get("/auth/wechat/url")
  return response.data
 }
 /**
 * 微信回调登录
 */
 export const wechatCallback = async (
  code: string,
  state: string,
 ): Promise<WechatCallbackResponse> => {
  const response = await apiClient.post("/auth/wechat/callback", { code, state })
  return response.data
 }
@@ -0,0 +1,68 @@
 /**
 * BGM 预设音乐 API
 * 对接后端 BGM 混音能力：预设列表查询（按风格分类 + 关键词搜索）
 */
 import apiClient from "./client"
 /* ──────────── 类型 ──────────── */
 /** BGM 风格分类 */
 export type BgmCategory = "轻快" | "治愈" | "科技" | "电商"
 /** BGM 预设项 */
 export interface BgmPreset {
  id: string
  name: string
  category: BgmCategory
  /** 音频文件 URL */
  url: string
  /** 时长（秒） */
  duration: number
  /** 关键词标签 */
  tags: string[]
  /** 封面图 URL */
  cover_url?: string
 }
 /** BGM 预设列表查询参数 */
 export interface BgmPresetsQuery {
  category?: BgmCategory | string
  keyword?: string
 }
 /** BGM 混音配置（嵌入模板） */
 export interface BgmMixConfig {
  /** 是否启用 BGM */
  enabled: boolean
  /** 选中的 BGM ID */
  music_id: string
  /** BGM 音量 0-100 */
  volume: number
  /** 淡入时长（秒） 0-3 */
  fade_in: number
  /** 淡出时长（秒） 0-3 */
  fade_out: number
  /** 人声闪避（sidechain） */
  voice_dodge: boolean
 }
 /** 默认 BGM 混音配置 */
 export const DEFAULT_BGM_MIX_CONFIG: BgmMixConfig = {
  enabled: false,
  music_id: "",
  volume: 50,
  fade_in: 0.5,
  fade_out: 0.5,
  voice_dodge: true,
 }
 /* ──────────── API ──────────── */
 /** 获取 BGM 预设列表 */
 export const getBgmPresets = async (params?: BgmPresetsQuery): Promise<BgmPreset[]> => {
  const searchParams: Record<string, string> = {}
  if (params?.category) searchParams.category = params.category
  if (params?.keyword) searchParams.keyword = params.keyword
  const res = await apiClient.get("/bgm/presets", { params: searchParams })
  return res.data?.data ?? res.data ?? []
 }
@@ -1,14 +0,0 @@
 /**
 * BGM 预设音乐 API 函数
 */
 import apiClient from "../client"
 import type { BgmPreset, BgmPresetsQuery } from "./types"
 /** 获取 BGM 预设列表 */
 export const getBgmPresets = async (params?: BgmPresetsQuery): Promise<BgmPreset[]> => {
  const searchParams: Record<string, string> = {}
  if (params?.category) searchParams.category = params.category
  if (params?.keyword) searchParams.keyword = params.keyword
  const res = await apiClient.get("/bgm/presets", { params: searchParams })
  return res.data?.data ?? res.data ?? []
 }
@@ -1,14 +0,0 @@
 /**
 * BGM 相关常量
 */
 import type { BgmMixConfig } from "./types"
 /** 默认 BGM 混音配置 */
 export const DEFAULT_BGM_MIX_CONFIG: BgmMixConfig = {
  enabled: false,
  music_id: "",
  volume: 50,
  fade_in: 0.5,
  fade_out: 0.5,
  voice_dodge: true,
 }
@@ -1,13 +0,0 @@
 /**
 * BGM API — 目录化入口
 * 保持与原 bgm.ts 相同导出，向后兼容
 */
 // 类型
 export type { BgmCategory, BgmPreset, BgmPresetsQuery, BgmMixConfig } from "./types"
 // 常量
 export { DEFAULT_BGM_MIX_CONFIG } from "./constants"
 // API 函数
 export { getBgmPresets } from "./bgm"
@@ -1,43 +0,0 @@
 /**
 * BGM 相关类型定义
 */
 /** BGM 风格分类 */
 export type BgmCategory = "轻快" | "治愈" | "科技" | "电商"
 /** BGM 预设项 */
 export interface BgmPreset {
  id: string
  name: string
  category: BgmCategory
  /** 音频文件 URL */
  url: string
  /** 时长（秒） */
  duration: number
  /** 关键词标签 */
  tags: string[]
  /** 封面图 URL */
  cover_url?: string
 }
 /** BGM 预设列表查询参数 */
 export interface BgmPresetsQuery {
  category?: BgmCategory | string
  keyword?: string
 }
 /** BGM 混音配置（嵌入模板） */
 export interface BgmMixConfig {
  /** 是否启用 BGM */
  enabled: boolean
  /** 选中的 BGM ID */
  music_id: string
  /** BGM 音量 0-100 */
  volume: number
  /** 淡入时长（秒） 0-3 */
  fade_in: number
  /** 淡出时长（秒） 0-3 */
  fade_out: number
  /** 人声闪避（sidechain） */
  voice_dodge: boolean
 }
@@ -57,21 +57,7 @@ apiClient.interceptors.response.use(
    }
    // 401 → 尝试刷新 Token
-    // 排除 auth 端点：登录/注册/找回密码的 401 是正常业务响应（如密码错误），
+    if (error.response?.status === 401 && originalRequest && !originalRequest._retry) {
    // 不应触发 token 刷新或登出跳转，走后面的错误提示逻辑即可
    const requestUrl = originalRequest?.url || ""
    const isAuthEndpoint =
      requestUrl.includes("/auth/login") ||
      requestUrl.includes("/auth/register") ||
      requestUrl.includes("/auth/forgot-password") ||
      requestUrl.includes("/auth/reset-password")
    if (
      error.response?.status === 401 &&
      originalRequest &&
      !originalRequest._retry &&
      !isAuthEndpoint
    ) {
      const refreshToken = useAuthStore.getState().refreshToken
      // 无 refresh_token → 直接登出
@@ -1,53 +0,0 @@
 /**
 * 封面模板 CRUD API
 * 后端路由: /api/v1/cover-templates
 */
 import apiClient from "./client"
 import type { CoverTemplate } from "@/pages/generate/types/cover"
 export interface CoverTemplateListResponse {
  items: CoverTemplate[]
  total: number
 }
 export interface CoverTemplateCreateRequest {
  name: string
  config?: {
    background_enabled?: boolean
    background_color?: string
    portrait_enabled?: boolean
    title_text?: string
    subtitle_text?: string
    mask_enabled?: boolean
  }
 }
 export type CoverTemplateUpdateRequest = Partial<CoverTemplateCreateRequest>
 /** 获取封面模板列表 */
 export async function fetchCoverTemplates(): Promise<CoverTemplateListResponse> {
  const response = await apiClient.get<CoverTemplateListResponse>("/cover-templates")
  return response.data
 }
 /** 创建封面模板 */
 export async function createCoverTemplate(
  data: CoverTemplateCreateRequest,
 ): Promise<CoverTemplate> {
  const response = await apiClient.post<CoverTemplate>("/cover-templates", data)
  return response.data
 }
 /** 更新封面模板 */
 export async function updateCoverTemplate(
  id: string,
  data: CoverTemplateUpdateRequest,
 ): Promise<CoverTemplate> {
  const response = await apiClient.put<CoverTemplate>(`/cover-templates/${id}`, data)
  return response.data
 }
 /** 删除封面模板（系统模板不可删） */
 export async function deleteCoverTemplate(id: string): Promise<void> {
  await apiClient.delete(`/cover-templates/${id}`)
 }
@@ -1,6 +1,8 @@
 /**
- * 查重相关类型定义
+ * 查重 API 模块
 * 提供视频查重相关接口
 */
 import apiClient from "./client"
 /** 查重记录状态 */
 export type DuplicationStatus = "pending" | "processing" | "completed" | "failed"
@@ -60,3 +62,38 @@ export interface DuplicationUploadResponse {
  /** 消息 */
  message: string
 }
 // ============ API 函数 ============
 /** 上传视频进行查重 */
 export const uploadForDuplication = async (file: File): Promise<DuplicationUploadResponse> => {
  const formData = new FormData()
  formData.append("file", file)
  const response = await apiClient.post("/duplication/upload", formData, {
    headers: { "Content-Type": "multipart/form-data" },
  })
  return response.data
 }
 /** 获取查重记录列表 */
 export const getDuplicationRecords = async (): Promise<DuplicationRecord[]> => {
  const response = await apiClient.get("/duplication/records")
  return response.data
 }
 /** 获取查重详情 */
 export const getDuplicationDetail = async (recordId: string): Promise<DuplicationDetail> => {
  const response = await apiClient.get(`/duplication/records/${recordId}`)
  return response.data
 }
 /** 删除查重记录 */
 export const deleteDuplicationRecord = async (recordId: string): Promise<void> => {
  await apiClient.delete(`/duplication/records/${recordId}`)
 }
 /** 重新查重 */
 export const retryDuplication = async (recordId: string): Promise<DuplicationUploadResponse> => {
  const response = await apiClient.post(`/duplication/records/${recordId}/retry`)
  return response.data
 }
@@ -1,38 +0,0 @@
 /**
 * 查重 API 函数
 */
 import apiClient from "../client"
 import type { DuplicationDetail, DuplicationRecord, DuplicationUploadResponse } from "./types"
 /** 上传视频进行查重 */
 export const uploadForDuplication = async (file: File): Promise<DuplicationUploadResponse> => {
  const formData = new FormData()
  formData.append("file", file)
  const response = await apiClient.post("/duplication/upload", formData, {
    headers: { "Content-Type": "multipart/form-data" },
  })
  return response.data
 }
 /** 获取查重记录列表 */
 export const getDuplicationRecords = async (): Promise<DuplicationRecord[]> => {
  const response = await apiClient.get("/duplication/records")
  return response.data
 }
 /** 获取查重详情 */
 export const getDuplicationDetail = async (recordId: string): Promise<DuplicationDetail> => {
  const response = await apiClient.get(`/duplication/records/${recordId}`)
  return response.data
 }
 /** 删除查重记录 */
 export const deleteDuplicationRecord = async (recordId: string): Promise<void> => {
  await apiClient.delete(`/duplication/records/${recordId}`)
 }
 /** 重新查重 */
 export const retryDuplication = async (recordId: string): Promise<DuplicationUploadResponse> => {
  const response = await apiClient.post(`/duplication/records/${recordId}/retry`)
  return response.data
 }
@@ -1,22 +0,0 @@
 /**
 * 查重 API — 目录化入口
 * 保持与原 duplication.ts 相同导出，向后兼容
 */
 // 类型
 export type {
  DuplicationStatus,
  DuplicationRecord,
  DuplicateSegment,
  DuplicationDetail,
  DuplicationUploadResponse,
 } from "./types"
 // API 函数
 export {
  uploadForDuplication,
  getDuplicationRecords,
  getDuplicationDetail,
  deleteDuplicationRecord,
  retryDuplication,
 } from "./duplication"
@@ -0,0 +1,243 @@
 /**
 * 模板编辑器 API
 * 对接后端 /api/v1/templates 路由
 */
 import apiClient from "./client"
 import type {
  WatermarkConfig,
  IntroOutroConfig,
  PipConfig,
  FilterConfig,
  ChromaKeyConfig,
  StickerConfig,
  CoverConfig,
 } from "@/pages/editing-planner/types"
 /* ──────────── 类型定义 ──────────── */
 /** 模板模式（后端枚举值） */
 export type TemplateMode = "pip" | "voice_over" | "one_take" | "voice_pip"
 /** 模式显示名称映射 */
 export const MODE_LABELS: Record<TemplateMode, string> = {
  pip: "混剪",
  voice_over: "人物口播",
  one_take: "一镜到底",
  voice_pip: "口播+混剪",
 }
 /** 模式颜色映射 */
 export const MODE_COLORS: Record<TemplateMode, string> = {
  pip: "blue",
  voice_over: "green",
  one_take: "orange",
  voice_pip: "purple",
 }
 /** 标题配置 */
 export interface TitleConfig {
  ai_auto_select: boolean
  content: string
  font_preset: string
  font_color: string
  font_size: number
  position: string
 }
 /** 字幕配置 */
 export interface SubtitleConfig {
  enabled: boolean
  position: string
  font: string
  color: string
  size: number
  animation: string
 }
 /** BGM 配置 */
 export interface BgmConfig {
  enabled: boolean
  music_id: string
 }
 /** 模板片段 */
 export interface TemplateSegment {
  id?: string
  segment_order: number
  duration_min: number
  duration_max: number
  material_type: string | null
 }
 /** 剪辑模板 */
 export interface EditingTemplate {
  id: string
  name: string
  mode: TemplateMode
  category: string
  tags: string[]
  title_config: TitleConfig
  subtitle_config: SubtitleConfig
  bgm_config: BgmConfig
  estimated_duration: number
  segments: TemplateSegment[]
  /** 水印配置（后端就绪后启用） */
  watermark_config?: WatermarkConfig
  /** 片头片尾配置（后端就绪后启用） */
  intro_outro_config?: IntroOutroConfig
  /** 混剪配置 */
  pip_config?: PipConfig
  /** 滤镜调色配置 */
  filter_config?: FilterConfig
  /** 绿幕抠像配置 */
  green_screen_config?: ChromaKeyConfig
  /** 贴纸配置 */
  sticker_config?: StickerConfig
  /** 封面配置 */
  cover_config?: CoverConfig
  is_active?: boolean
  created_at: string
  updated_at: string
 }
 /** 模板分类 */
 export interface TemplateCategory {
  id: string
  name: string
  created_at?: string
 }
 /** 创建/更新模板请求体 */
 export interface SaveTemplatePayload {
  name: string
  mode: TemplateMode
  category: string
  tags: string[]
  title_config: TitleConfig
  subtitle_config: SubtitleConfig
  bgm_config: BgmConfig
  estimated_duration: number
  segments: Omit<TemplateSegment, "id">[]
  /** 水印配置（后端就绪后启用） */
  watermark_config?: WatermarkConfig
  /** 片头片尾配置（后端就绪后启用） */
  intro_outro_config?: IntroOutroConfig
  /** 混剪配置 */
  pip_config?: PipConfig
  /** 滤镜调色配置 */
  filter_config?: FilterConfig
  /** 绿幕抠像配置 */
  green_screen_config?: ChromaKeyConfig
  /** 贴纸配置 */
  sticker_config?: StickerConfig
  /** 封面配置 */
  cover_config?: CoverConfig
 }
 /** 使用模板生成请求体 */
 export interface GenerateFromTemplatePayload {
  voiceover_duration: number
 }
 /** 验证警告详情 */
 export interface ValidationWarningDetails {
  /** 相关字段名 */
  field?: string
  /** 期望值 */
  expected?: string | number
  /** 实际值 */
  actual?: string | number
  /** 建议值 */
  suggested?: string | number
 }
 /** 验证/生成响应 */
 export interface ValidateWarning {
  code: string
  message: string
  details?: ValidationWarningDetails
 }
 /** 使用模板生成响应 */
 export interface GenerateFromTemplateResponse {
  template: EditingTemplate
  warnings: ValidateWarning[]
 }
 /** 列表响应（带分页） */
 export interface ListTemplatesResponse {
  items: EditingTemplate[]
  total: number
 }
 /** 分类列表响应 */
 export interface ListCategoriesResponse {
  items: TemplateCategory[]
 }
 // ============ API 函数 ============
 /** 获取模板列表 */
 export const getEditingTemplates = async (params?: {
  category?: string
  tag?: string
  skip?: number
  limit?: number
 }): Promise<EditingTemplate[]> => {
  const response = await apiClient.get<ListTemplatesResponse>("/templates", {
    params: {
      skip: params?.skip ?? 0,
      limit: params?.limit ?? 50,
    },
  })
  let list = response.data.items
  if (params?.category) list = list.filter((t) => t.category === params.category)
  if (params?.tag) list = list.filter((t) => t.tags.includes(params.tag!))
  return list
 }
 /** 获取模板详情 */
 export const getEditingTemplate = async (id: string): Promise<EditingTemplate> => {
  const response = await apiClient.get<EditingTemplate>(`/templates/${id}`)
  return response.data
 }
 /** 创建模板 */
 export const createEditingTemplate = async (
  data: SaveTemplatePayload,
 ): Promise<EditingTemplate> => {
  const response = await apiClient.post<EditingTemplate>("/templates", data)
  return response.data
 }
 /** 更新模板 */
 export const updateEditingTemplate = async (
  id: string,
  data: SaveTemplatePayload,
 ): Promise<EditingTemplate> => {
  const response = await apiClient.patch<EditingTemplate>(`/templates/${id}`, data)
  return response.data
 }
 /** 删除模板 */
 export const deleteEditingTemplate = async (id: string): Promise<void> => {
  await apiClient.delete(`/templates/${id}`)
 }
 /** 获取模板分类列表 */
 export const getTemplateCategories = async (): Promise<TemplateCategory[]> => {
  const response = await apiClient.get<ListCategoriesResponse>("/templates/categories/list")
  return response.data.items
 }
 /** 使用模板生成视频（调用 validate 端点） */
 export const generateFromTemplate = async (
  templateId: string,
  data: GenerateFromTemplatePayload,
 ): Promise<GenerateFromTemplateResponse> => {
  const response = await apiClient.post<GenerateFromTemplateResponse>(
    `/templates/${templateId}/validate`,
    data,
  )
  return response.data
 }
@@ -1,20 +0,0 @@
 /**
 * 模板编辑器常量
 */
 import type { TemplateMode } from "./types"
 /** 模式显示名称映射 */
 export const MODE_LABELS: Record<TemplateMode, string> = {
  pip: "混剪",
  voice_over: "人物口播",
  one_take: "一镜到底",
  voice_pip: "口播+混剪",
 }
 /** 模式颜色映射 */
 export const MODE_COLORS: Record<TemplateMode, string> = {
  pip: "blue",
  voice_over: "green",
  one_take: "orange",
  voice_pip: "purple",
 }
@@ -1,36 +0,0 @@
 /**
 * 模板编辑器 API — 目录化入口
 * 保持与原 editing-planner.ts 相同导出，向后兼容
 */
 // 类型
 export type {
  TemplateMode,
  TitleConfig,
  SubtitleConfig,
  BgmConfig,
  TemplateSegment,
  EditingTemplate,
  TemplateCategory,
  SaveTemplatePayload,
  GenerateFromTemplatePayload,
  ValidationWarningDetails,
  ValidateWarning,
  GenerateFromTemplateResponse,
  ListTemplatesResponse,
  ListCategoriesResponse,
 } from "./types"
 // 常量
 export { MODE_LABELS, MODE_COLORS } from "./constants"
 // API 函数
 export {
  getEditingTemplates,
  getEditingTemplate,
  createEditingTemplate,
  updateEditingTemplate,
  deleteEditingTemplate,
  getTemplateCategories,
  generateFromTemplate,
 } from "./templates"
@@ -1,78 +0,0 @@
 /**
 * 模板编辑器 API 函数
 */
 import apiClient from "../client"
 import type {
  EditingTemplate,
  TemplateCategory,
  SaveTemplatePayload,
  GenerateFromTemplatePayload,
  GenerateFromTemplateResponse,
  ListTemplatesResponse,
  ListCategoriesResponse,
 } from "./types"
 /** 获取模板列表 */
 export const getEditingTemplates = async (params?: {
  category?: string
  tag?: string
  skip?: number
  limit?: number
 }): Promise<EditingTemplate[]> => {
  const response = await apiClient.get<ListTemplatesResponse>("/templates", {
    params: {
      skip: params?.skip ?? 0,
      limit: params?.limit ?? 50,
    },
  })
  let list = response.data.items
  if (params?.category) list = list.filter((t) => t.category === params.category)
  if (params?.tag) list = list.filter((t) => t.tags.includes(params.tag!))
  return list
 }
 /** 获取模板详情 */
 export const getEditingTemplate = async (id: string): Promise<EditingTemplate> => {
  const response = await apiClient.get<EditingTemplate>(`/templates/${id}`)
  return response.data
 }
 /** 创建模板 */
 export const createEditingTemplate = async (
  data: SaveTemplatePayload,
 ): Promise<EditingTemplate> => {
  const response = await apiClient.post<EditingTemplate>("/templates", data)
  return response.data
 }
 /** 更新模板 */
 export const updateEditingTemplate = async (
  id: string,
  data: SaveTemplatePayload,
 ): Promise<EditingTemplate> => {
  const response = await apiClient.patch<EditingTemplate>(`/templates/${id}`, data)
  return response.data
 }
 /** 删除模板 */
 export const deleteEditingTemplate = async (id: string): Promise<void> => {
  await apiClient.delete(`/templates/${id}`)
 }
 /** 获取模板分类列表 */
 export const getTemplateCategories = async (): Promise<TemplateCategory[]> => {
  const response = await apiClient.get<ListCategoriesResponse>("/templates/categories/list")
  return response.data.items
 }
 /** 使用模板生成视频（调用 validate 端点） */
 export const generateFromTemplate = async (
  templateId: string,
  data: GenerateFromTemplatePayload,
 ): Promise<GenerateFromTemplateResponse> => {
  const response = await apiClient.post<GenerateFromTemplateResponse>(
    `/templates/${templateId}/validate`,
    data,
  )
  return response.data
 }
@@ -1,138 +0,0 @@
 /**
 * 模板编辑器类型定义
 */
 import type {
  WatermarkConfig,
  IntroOutroConfig,
  PipConfig,
  FilterConfig,
  ChromaKeyConfig,
  StickerConfig,
 } from "@/pages/editing-planner/types"
 import type { CoverConfig } from "@/pages/generate/types/cover"
 /** 模板模式（后端枚举值） */
 export type TemplateMode = "pip" | "voice_over" | "one_take" | "voice_pip"
 /** 标题配置 */
 export interface TitleConfig {
  ai_auto_select: boolean
  content: string
  font_preset: string
  font_color: string
  font_size: number
  position: string
 }
 /** 字幕配置 */
 export interface SubtitleConfig {
  enabled: boolean
  position: string
  font: string
  color: string
  size: number
  animation: string
 }
 /** BGM 配置 */
 export interface BgmConfig {
  enabled: boolean
  music_id: string
 }
 /** 模板片段 */
 export interface TemplateSegment {
  id?: string
  segment_order: number
  duration_min: number
  duration_max: number
  material_type: string | null
 }
 /** 剪辑模板 */
 export interface EditingTemplate {
  id: string
  name: string
  mode: TemplateMode
  category: string
  tags: string[]
  title_config: TitleConfig
  subtitle_config: SubtitleConfig
  bgm_config: BgmConfig
  estimated_duration: number
  segments: TemplateSegment[]
  watermark_config?: WatermarkConfig
  intro_outro_config?: IntroOutroConfig
  pip_config?: PipConfig
  filter_config?: FilterConfig
  green_screen_config?: ChromaKeyConfig
  sticker_config?: StickerConfig
  cover_config?: CoverConfig
  is_active?: boolean
  created_at: string
  updated_at: string
 }
 /** 模板分类 */
 export interface TemplateCategory {
  id: string
  name: string
  created_at?: string
 }
 /** 创建/更新模板请求体 */
 export interface SaveTemplatePayload {
  name: string
  mode: TemplateMode
  category: string
  tags: string[]
  title_config: TitleConfig
  subtitle_config: SubtitleConfig
  bgm_config: BgmConfig
  estimated_duration: number
  segments: Omit<TemplateSegment, "id">[]
  watermark_config?: WatermarkConfig
  intro_outro_config?: IntroOutroConfig
  pip_config?: PipConfig
  filter_config?: FilterConfig
  green_screen_config?: ChromaKeyConfig
  sticker_config?: StickerConfig
  cover_config?: CoverConfig
 }
 /** 使用模板生成请求体 */
 export interface GenerateFromTemplatePayload {
  voiceover_duration: number
 }
 /** 验证警告详情 */
 export interface ValidationWarningDetails {
  field?: string
  expected?: string | number
  actual?: string | number
  suggested?: string | number
 }
 /** 验证警告 */
 export interface ValidateWarning {
  code: string
  message: string
  details?: ValidationWarningDetails
 }
 /** 使用模板生成响应 */
 export interface GenerateFromTemplateResponse {
  template: EditingTemplate
  warnings: ValidateWarning[]
 }
 /** 列表响应（带分页） */
 export interface ListTemplatesResponse {
  items: EditingTemplate[]
  total: number
 }
 /** 分类列表响应 */
 export interface ListCategoriesResponse {
  items: TemplateCategory[]
 }
@@ -1,14 +0,0 @@
 import apiClient from "../client"
 import type { ConfirmGenerationRequest, ConfirmGenerationResponse } from "./types"
 /** 确认生成 — 基于预览任务创建正式生成任务 */
 export const confirmGeneration = async (
  taskId: string,
  params: ConfirmGenerationRequest,
 ): Promise<ConfirmGenerationResponse> => {
  const response = await apiClient.post<ConfirmGenerationResponse>(
    `/generation/tasks/${taskId}/confirm`,
    params,
  )
  return response.data
 }
@@ -1,35 +0,0 @@
 import apiClient from "../client"
 export interface GenerateCoverRequest {
  asset_ids: string[]
  cover_type?: "ai_frame" | "manual" | "upload" | "ai_regenerate"
  frame_time?: number
 }
 export interface GenerateCoverResponse {
  plan_id: string
  cover: {
    scheme?: string
    asset_id?: string
    frame_time?: number
    image_url?: string
    thumbnail_url?: string
    [key: string]: unknown
  }
 }
 /** AI 生成封面 — 从预览视频中抽帧 */
 export async function generateCover(
  templateId: string,
  data: GenerateCoverRequest,
 ): Promise<GenerateCoverResponse> {
  const response = await apiClient.post<GenerateCoverResponse>(
    "/generation/generate-cover",
    { ...data, template_id: templateId },
    {
      timeout: 300000,
      params: { template_id: templateId },
    },
  )
  return response.data
 }
@@ -1,15 +0,0 @@
 export type {
  PreviewStatus,
  CreatePreviewRequest,
  CreatePreviewResponse,
  PreviewTaskResponse,
  ConfirmGenerationRequest,
  ConfirmGenerationResponse,
  ConfirmGenerationTaskItem,
 } from "./types"
 export { createPreview, getPreviewStatus } from "./preview"
 export { confirmGeneration } from "./confirm"
 export { generateCover } from "./cover"
 export type { GenerateCoverRequest, GenerateCoverResponse } from "./cover"
@@ -1,16 +0,0 @@
 import apiClient from "../client"
 import type { CreatePreviewRequest, CreatePreviewResponse, PreviewTaskResponse } from "./types"
 /** 创建预览生成任务（单版本） */
 export const createPreview = async (
  params: CreatePreviewRequest,
 ): Promise<CreatePreviewResponse> => {
  const response = await apiClient.post<CreatePreviewResponse>("/generation/preview", params)
  return response.data
 }
 /** 查询预览任务状态及结果 */
 export const getPreviewStatus = async (taskId: string): Promise<PreviewTaskResponse> => {
  const response = await apiClient.get<PreviewTaskResponse>(`/generation/preview/${taskId}`)
  return response.data
 }
@@ -1,94 +0,0 @@
 /** 预览任务状态 */
 export type PreviewStatus = "pending" | "generating" | "completed" | "failed" | "cancelled"
 /** 创建预览任务请求 */
 export interface CreatePreviewRequest {
  template_id: string
  asset_ids: string[]
  source_edit_plan_id?: string
  title_ids?: string[]
  voice_ids?: string[]
  /** 配音素材库ID（用户上传的音频或AI配音），对应配音选择页面选择的配音素材 */
  voice_library_id?: string
  video_title?: string
  duration?: number
  video_ratio?: string
  bgm_config?: {
    enabled: boolean
    preset_id?: string
    volume?: number
  }
 }
 /** 创建预览任务响应 */
 export interface CreatePreviewResponse {
  task_id: string
  status: PreviewStatus
  is_preview: boolean
  resolution: string
  created_at: string
 }
 /** 预览任务详情响应 */
 export interface PreviewTaskResponse {
  task_id: string
  status: PreviewStatus
  progress: number
  is_preview: boolean
  resolution: string
  video_url?: string
  duration?: number
  file_size?: number
  clip_count?: number
  transition_count?: number
  material_usage?: number
  error_message?: string
  created_at: string
  started_at?: string
  finished_at?: string
  generate_duration?: number
 }
 /** 确认生成请求体 — 基于预览任务创建正式生成任务 */
 export interface ConfirmGenerationRequest {
  /** 输出视频宽度，默认 1080 */
  output_width?: number
  /** 输出视频高度，默认 1920 */
  output_height?: number
  /** 自定义封面图片 URL */
  cover_url?: string
  /** 自定义视频标题 */
  custom_title?: string
 }
 /** 确认生成响应 */
 export interface ConfirmGenerationResponse {
  items: ConfirmGenerationTaskItem[]
  total: number
 }
 /** 确认生成返回的任务项 */
 export interface ConfirmGenerationTaskItem {
  id: string
  project_id: string
  asset_library_id: string
  strategy_id: string
  voice_library_id: string
  template_id: string
  asset_ids: string[]
  title_ids: string[]
  voice_ids: string[]
  source_edit_plan_id: string
  asset_select_mode: string
  batch_id: string
  is_preview: boolean
  source_task_id: string
  output_width: number
  output_height: number
  cover_url: string
  custom_title: string
  status: string
  progress: number
  result_count: number
  error_message: string
 }
@@ -0,0 +1,174 @@
 /**
 * 成品 / 视频相关 API
 * 后端实际接口：/videos
 */
 import apiClient from "./client"
 /** 复核状态 */
 export type ReviewStatus = "pending_review" | "approved" | "rejected"
 /** 成品条目 */
 export interface ProductItem {
  id: string
  title: string
  video_url?: string
  thumbnail_url?: string
  duration_seconds?: number
  file_size?: number
  resolution?: string
  status: "processing" | "completed" | "failed"
  /** 复核状态 */
  review_status?: ReviewStatus
  /** 所属项目 ID */
  project_id?: string
  /** 所属项目名称 */
  project_name?: string
  /** 查重率（百分比） */
  duplicate_rate?: number
  created_at?: string
  updated_at?: string
 }
 /** 列表查询参数 */
 export interface ProductListParams {
  page?: number
  page_size?: number
  project_id?: string
  review_status?: ReviewStatus | "all"
 }
 /** 分页响应 */
 export interface ProductListResponse {
  items: ProductItem[]
  total: number
  page: number
  page_size: number
 }
 /** 批量下载任务状态 */
 export interface BatchDownloadStatus {
  job_id: string
  status: "processing" | "completed" | "failed"
  /** 完成后返回的下载 URL */
  download_url?: string
  /** 进度百分比 */
  progress?: number
 }
 /** 后端 /videos 接口返回的原始视频条目 */
 interface VideoItem {
  id: string
  project_id: string
  generation_task_id: string
  name: string
  file_url: string
  file_size: number
  duration: number
  thumbnail_url: string | null
  width: number
  height: number
  fps: number
  status: string
  review_status: ReviewStatus
  generation_params: Record<string, unknown>
  download_url: string
  generated_at: string
 }
 /**
 * 将后端 VideoItem 映射为 ProductItem 格式
 */
 function mapVideoToProductItem(video: VideoItem): ProductItem {
  return {
    id: video.id,
    title: video.name || "未命名视频",
    // 优先用 download_url（带签名）播放，file_url 无签名无法访问
    video_url: video.download_url || video.file_url,
    thumbnail_url: video.thumbnail_url || undefined,
    duration_seconds: video.duration,
    file_size: video.file_size,
    resolution: video.width && video.height ? `${video.width}x${video.height}` : undefined,
    status:
      video.status === "completed"
        ? "completed"
        : video.status === "failed"
          ? "failed"
          : "processing",
    review_status: video.review_status,
    project_id: video.project_id,
    // 后端 /videos 接口暂无 project_name 字段
    project_name: undefined,
    // 后端字段名为 generated_at，映射为 created_at 供前端统一使用
    created_at: video.generated_at,
    updated_at: video.generated_at,
    // 后端 /videos 接口暂无 duplicate_rate 字段
    duplicate_rate: undefined,
  }
 }
 /** 获取成品列表（支持分页和筛选） */
 export const getProducts = async (params?: ProductListParams): Promise<ProductItem[]> => {
  const response = await apiClient.get("/videos", { params })
  const data = response.data
  const videos: VideoItem[] = Array.isArray(data?.items)
    ? data.items
    : Array.isArray(data)
      ? data
      : []
  return videos.map(mapVideoToProductItem)
 }
 /** 获取单个成品详情 */
 export const getProduct = async (productId: string): Promise<ProductItem> => {
  const response = await apiClient.get(`/videos/${productId}`)
  return mapVideoToProductItem(response.data as VideoItem)
 }
 /**
 * 删除成品
 * 注意：后端暂未实现 /videos DELETE 接口，调用会返回 405
 * 待后端实现后自动生效
 */
 export const deleteProduct = async (productId: string): Promise<void> => {
  await apiClient.delete(`/videos/${productId}`)
 }
 /**
 * 获取成品下载链接
 * 直接使用列表返回的 download_url（带OSS签名）
 */
 export const getProductDownloadUrl = async (
  productId: string,
 ): Promise<{ url: string; expires_at: string }> => {
  // 优先从列表缓存取；如果没有则调详情接口
  const product = await getProduct(productId)
  if (!product.video_url) throw new Error("下载链接不可用")
  return { url: product.video_url, expires_at: "" }
 }
 /** 更新复核状态 — TODO: 后端暂无对应端点，暂存本地状态 */
 export const updateReviewStatus = async (
  productId: string,
  status: ReviewStatus,
 ): Promise<ProductItem> => {
  // 后端暂无 /videos/{id}/review 端点
  // 暂时返回当前状态，后续可扩展
  const product = await getProduct(productId)
  return { ...product, review_status: status }
 }
 /** 发起批量下载 — TODO: 后端暂无对应端点 */
 export const batchDownload = async (videoIds: string[]): Promise<{ job_id: string }> => {
  // 后端暂无 /videos/batch-download 端点
  // 暂时返回模拟 job_id，后续可扩展
  console.warn("[batchDownload] 后端暂无批量下载端点", videoIds)
  return { job_id: `mock-${Date.now()}` }
 }
 /** 查询批量下载状态 — TODO: 后端暂无对应端点 */
 export const getBatchDownloadStatus = async (jobId: string): Promise<BatchDownloadStatus> => {
  // 后端暂无 /videos/batch-download/{jobId} 端点
  // 暂时返回模拟状态，后续可扩展
  console.warn("[getBatchDownloadStatus] 后端暂无批量下载状态端点", jobId)
  return { job_id: jobId, status: "processing", progress: 0 }
 }
@@ -1,28 +0,0 @@
 /**
 * 成品 / 视频相关 API — 目录化入口
 * 保持与原 products.ts 相同导出，向后兼容
 */
 // 类型
 export type {
  ReviewStatus,
  ProductItem,
  ProductListParams,
  ProductListResponse,
  BatchDownloadStatus,
  VideoItem,
 } from "./types"
 // 工具函数
 export { mapVideoToProductItem } from "./utils"
 // API 函数
 export {
  getProducts,
  getProduct,
  deleteProduct,
  getProductDownloadUrl,
  updateReviewStatus,
  batchDownload,
  getBatchDownloadStatus,
 } from "./products"
@@ -1,80 +0,0 @@
 /**
 * 成品 / 视频相关 API 函数
 * 后端实际接口：/videos
 */
 import apiClient from "../client"
 import type {
  BatchDownloadStatus,
  ProductItem,
  ProductListParams,
  VideoItem,
  ReviewStatus,
 } from "./types"
 import { mapVideoToProductItem } from "./utils"
 /** 获取成品列表（支持分页和筛选） */
 export const getProducts = async (params?: ProductListParams): Promise<ProductItem[]> => {
  const response = await apiClient.get("/videos", { params })
  const data = response.data
  const videos: VideoItem[] = Array.isArray(data?.items)
    ? data.items
    : Array.isArray(data)
      ? data
      : []
  return videos.map(mapVideoToProductItem)
 }
 /** 获取单个成品详情 */
 export const getProduct = async (productId: string): Promise<ProductItem> => {
  const response = await apiClient.get(`/videos/${productId}`)
  return mapVideoToProductItem(response.data as VideoItem)
 }
 /**
 * 删除成品
 * 注意：后端暂未实现 /videos DELETE 接口，调用会返回 405
 * 待后端实现后自动生效
 */
 export const deleteProduct = async (productId: string): Promise<void> => {
  await apiClient.delete(`/videos/${productId}`)
 }
 /**
 * 获取成品下载链接
 * 直接使用列表返回的 download_url（带OSS签名）
 */
 export const getProductDownloadUrl = async (
  productId: string,
 ): Promise<{ url: string; expires_at: string }> => {
  // 优先从列表缓存取；如果没有则调详情接口
  const product = await getProduct(productId)
  if (!product.video_url) throw new Error("下载链接不可用")
  return { url: product.video_url, expires_at: "" }
 }
 /** 更新复核状态 — TODO: 后端暂无对应端点，暂存本地状态 */
 export const updateReviewStatus = async (
  productId: string,
  status: ReviewStatus,
 ): Promise<ProductItem> => {
  // 后端暂无 /videos/{id}/review 端点
  // 暂时返回当前状态，后续可扩展
  const product = await getProduct(productId)
  return { ...product, review_status: status }
 }
 /** 发起批量下载 — TODO: 后端暂无对应端点 */
 export const batchDownload = async (videoIds: string[]): Promise<{ job_id: string }> => {
  // 后端暂无 /videos/batch-download 端点
  // 暂时返回模拟 job_id，后续可扩展
  console.warn("[batchDownload] 后端暂无批量下载端点", videoIds)
  return { job_id: `mock-${Date.now()}` }
 }
 /** 查询批量下载状态 — TODO: 后端暂无对应端点 */
 export const getBatchDownloadStatus = async (jobId: string): Promise<BatchDownloadStatus> => {
  // 后端暂无 /videos/batch-download/{jobId} 端点
  // 暂时返回模拟状态，后续可扩展
  console.warn("[getBatchDownloadStatus] 后端暂无批量下载状态端点", jobId)
  return { job_id: jobId, status: "processing", progress: 0 }
 }
@@ -1,74 +0,0 @@
 /**
 * 成品 / 视频相关类型定义
 */
 /** 复核状态 */
 export type ReviewStatus = "pending_review" | "approved" | "rejected"
 /** 成品条目 */
 export interface ProductItem {
  id: string
  title: string
  video_url?: string
  thumbnail_url?: string
  duration_seconds?: number
  file_size?: number
  resolution?: string
  status: "processing" | "completed" | "failed"
  /** 复核状态 */
  review_status?: ReviewStatus
  /** 所属项目 ID */
  project_id?: string
  /** 所属项目名称 */
  project_name?: string
  /** 查重率（百分比） */
  duplicate_rate?: number
  created_at?: string
  updated_at?: string
 }
 /** 列表查询参数 */
 export interface ProductListParams {
  page?: number
  page_size?: number
  project_id?: string
  review_status?: ReviewStatus | "all"
 }
 /** 分页响应 */
 export interface ProductListResponse {
  items: ProductItem[]
  total: number
  page: number
  page_size: number
 }
 /** 批量下载任务状态 */
 export interface BatchDownloadStatus {
  job_id: string
  status: "processing" | "completed" | "failed"
  /** 完成后返回的下载 URL */
  download_url?: string
  /** 进度百分比 */
  progress?: number
 }
 /** 后端 /videos 接口返回的原始视频条目 */
 export interface VideoItem {
  id: string
  project_id: string
  generation_task_id: string
  name: string
  file_url: string
  file_size: number
  duration: number
  thumbnail_url: string | null
  width: number
  height: number
  fps: number
  status: string
  review_status: ReviewStatus
  generation_params: Record<string, unknown>
  download_url: string
  generated_at: string
 }
@@ -1,35 +0,0 @@
 /**
 * 成品数据转换工具函数
 */
 import type { ProductItem, VideoItem } from "./types"
 /**
 * 将后端 VideoItem 映射为 ProductItem 格式
 */
 export function mapVideoToProductItem(video: VideoItem): ProductItem {
  return {
    id: video.id,
    title: video.name || "未命名视频",
    // 优先用 download_url（带签名）播放，file_url 无签名无法访问
    video_url: video.download_url || video.file_url,
    thumbnail_url: video.thumbnail_url || undefined,
    duration_seconds: video.duration,
    file_size: video.file_size,
    resolution: video.width && video.height ? `${video.width}x${video.height}` : undefined,
    status:
      video.status === "completed"
        ? "completed"
        : video.status === "failed"
          ? "failed"
          : "processing",
    review_status: video.review_status,
    project_id: video.project_id,
    // 后端 /videos 接口暂无 project_name 字段
    project_name: undefined,
    // 后端字段名为 generated_at，映射为 created_at 供前端统一使用
    created_at: video.generated_at,
    updated_at: video.generated_at,
    // 后端 /videos 接口暂无 duplicate_rate 字段
    duplicate_rate: undefined,
  }
 }
@@ -1,10 +1,32 @@
 /**
- * 项目相关 API 函数
+ * 项目相关 API
 * 素材库需要 project_id，前端自动管理默认项目
 */
-import apiClient from "../client"
+import apiClient from "./client"
-import type { BackendListProjectsResponse, BackendProjectResponse, ProjectItem } from "./types"
+
-import { toProjectItem } from "./utils"
+export interface ProjectItem {
  id: string
  name: string
  description: string
 }
 /** 后端 ProjectResponse 只返回 id, name, description */
 interface BackendProjectResponse {
  id: string
  name: string
  description: string
 }
 /** 后端 ListProjectsResponse 返回 { items: [...] } */
 interface BackendListProjectsResponse {
  items: BackendProjectResponse[]
 }
 const toProjectItem = (item: BackendProjectResponse): ProjectItem => ({
  id: item.id,
  name: item.name,
  description: item.description,
 })
 /** 获取当前用户的项目列表 */
 export const getProjects = async (): Promise<ProjectItem[]> => {
@@ -1,13 +0,0 @@
 /**
 * 项目 API — 目录化入口
 * 保持与原 projects.ts 相同导出，向后兼容
 */
 // 类型
 export type { ProjectItem, BackendProjectResponse, BackendListProjectsResponse } from "./types"
 // 工具函数
 export { toProjectItem } from "./utils"
 // API 函数
 export { getProjects, createProject, getOrCreateDefaultProject } from "./projects"
@@ -1,21 +0,0 @@
 /**
 * 项目相关类型定义
 */
 export interface ProjectItem {
  id: string
  name: string
  description: string
 }
 /** 后端 ProjectResponse 只返回 id, name, description */
 export interface BackendProjectResponse {
  id: string
  name: string
  description: string
 }
 /** 后端 ListProjectsResponse 返回 { items: [...] } */
 export interface BackendListProjectsResponse {
  items: BackendProjectResponse[]
 }
@@ -1,10 +0,0 @@
 /**
 * 项目相关工具函数
 */
 import type { BackendProjectResponse, ProjectItem } from "./types"
 export const toProjectItem = (item: BackendProjectResponse): ProjectItem => ({
  id: item.id,
  name: item.name,
  description: item.description,
 })
@@ -1,6 +1,8 @@
 /**
- * 订阅相关类型定义
+ * 订阅 API 模块
 * 对接后端订阅管理接口
 */
 import apiClient from "./client"
 /** 套餐类型 */
 export type PlanType = "free" | "standard" | "pro" | "enterprise"
@@ -63,3 +65,42 @@ export interface ChangePlanResponse {
  message: string
  new_subscription?: SubscriptionInfo
 }
 // ============ API 函数 ============
 /** 获取当前订阅信息 */
 export const getCurrentSubscription = async (): Promise<SubscriptionInfo> => {
  const response = await apiClient.get("/subscription/current")
  return response.data
 }
 /** 获取账单记录列表 */
 export const getBillingRecords = async (): Promise<BillingRecord[]> => {
  const response = await apiClient.get("/subscription/billing-records")
  return response.data
 }
 /** 升级/降级套餐 */
 export const changePlan = async (request: ChangePlanRequest): Promise<ChangePlanResponse> => {
  const response = await apiClient.post("/subscription/change-plan", request)
  return response.data
 }
 /** 取消订阅 */
 export const cancelSubscription = async (): Promise<{
  success: boolean
  message: string
 }> => {
  const response = await apiClient.post("/subscription/cancel")
  return response.data
 }
 /** 切换自动续费 */
 export const toggleAutoRenew = async (
  enabled: boolean,
 ): Promise<{ success: boolean; message: string }> => {
  const response = await apiClient.post("/subscription/toggle-auto-renew", {
    enabled,
  })
  return response.data
 }
@@ -1,26 +0,0 @@
 /**
 * 订阅 API — 目录化入口
 * 保持与原 subscription.ts 相同导出，向后兼容
 */
 // 类型
 export type {
  PlanType,
  SubscriptionStatus,
  BillingStatus,
  BillingCycle,
  Plan,
  SubscriptionInfo,
  BillingRecord,
  ChangePlanRequest,
  ChangePlanResponse,
 } from "./types"
 // API 函数
 export {
  getCurrentSubscription,
  getBillingRecords,
  changePlan,
  cancelSubscription,
  toggleAutoRenew,
 } from "./subscription"
@@ -1,47 +0,0 @@
 /**
 * 订阅相关 API 函数
 */
 import apiClient from "../client"
 import type {
  BillingRecord,
  ChangePlanRequest,
  ChangePlanResponse,
  SubscriptionInfo,
 } from "./types"
 /** 获取当前订阅信息 */
 export const getCurrentSubscription = async (): Promise<SubscriptionInfo> => {
  const response = await apiClient.get("/subscription/current")
  return response.data
 }
 /** 获取账单记录列表 */
 export const getBillingRecords = async (): Promise<BillingRecord[]> => {
  const response = await apiClient.get("/subscription/billing-records")
  return response.data
 }
 /** 升级/降级套餐 */
 export const changePlan = async (request: ChangePlanRequest): Promise<ChangePlanResponse> => {
  const response = await apiClient.post("/subscription/change-plan", request)
  return response.data
 }
 /** 取消订阅 */
 export const cancelSubscription = async (): Promise<{
  success: boolean
  message: string
 }> => {
  const response = await apiClient.post("/subscription/cancel")
  return response.data
 }
 /** 切换自动续费 */
 export const toggleAutoRenew = async (
  enabled: boolean,
 ): Promise<{ success: boolean; message: string }> => {
  const response = await apiClient.post("/subscription/toggle-auto-renew", {
    enabled,
  })
  return response.data
 }
@@ -1,9 +1,15 @@
 /**
- * 标签 CRUD API 函数
+ * 标签 CRUD API
 * P3 标签体系：对接后端标签表
 */
-import apiClient from "../client"
+import apiClient from "./client"
-import type { TagItem } from "./types"
+
 export interface TagItem {
  id: string
  name: string
  created_at?: string
  usage_count?: number
 }
 /** 获取当前用户所有标签 */
 export const getTags = async (): Promise<TagItem[]> => {
@@ -1,10 +0,0 @@
 /**
 * 标签 API — 目录化入口
 * 保持与原 tags.ts 相同导出，向后兼容
 */
 // 类型
 export type { TagItem } from "./types"
 // API 函数
 export { getTags, createTag, deleteTag, tagAsset, untagAsset } from "./tags"
@@ -1,10 +0,0 @@
 /**
 * 标签相关类型定义
 */
 export interface TagItem {
  id: string
  name: string
  created_at?: string
  usage_count?: number
 }
@@ -1,6 +1,14 @@
 /**
- * 任务相关类型定义
+ * 任务相关 API
 * 对接后端任务中心 API：
 * - POST /api/v1/generation/tasks — 创建生成任务
 * - GET  /api/v1/tasks              — 用户级任务列表（支持分页/筛选）
 * - GET  /api/v1/tasks/{task_id}    — 任务详情（含 error_info）
 * - POST /api/v1/tasks/{task_id}/retry — 重试失败任务
 */
 import apiClient from "./client"
 /* ──────────── 类型定义 ──────────── */
 /** 任务状态 */
 export type TaskStatus = "pending" | "waiting" | "running" | "completed" | "failed" | "cancelled"
@@ -77,3 +85,39 @@ export interface CreateGenerationTaskResponse {
  result_count: number
  error_message: string
 }
 /* ──────────── API 函数 ──────────── */
 /** 创建生成任务（智能剪辑） */
 export const createGenerationTask = async (
  params: CreateGenerationTaskRequest,
 ): Promise<CreateGenerationTaskResponse> => {
  const { data } = await apiClient.post<CreateGenerationTaskResponse>("/generation/tasks", params)
  return data
 }
 /** 获取任务列表（支持分页和筛选） */
 export const getTasks = async (params?: TaskListParams): Promise<TaskListResponse> => {
  const { data } = await apiClient.get<TaskListResponse>("/tasks", {
    params,
  })
  return data
 }
 /** 获取当前用户的所有任务（兼容旧接口，跨 project） */
 export const getUserTasks = async (): Promise<TaskItem[]> => {
  const { data } = await apiClient.get("/tasks")
  return data.items || data || []
 }
 /** 获取单个任务详情（含 error_info） */
 export const getTask = async (taskId: string): Promise<TaskItem> => {
  const { data } = await apiClient.get(`/tasks/${taskId}`)
  return data
 }
 /** 重试失败的任务 */
 export const retryTask = async (taskId: string): Promise<TaskItem> => {
  const { data } = await apiClient.post(`/tasks/${taskId}/retry`)
  return data
 }
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
CI Bot	b233f8e539	merge: 解决第56波与develop的add/add冲突（sticker+subtitle） AI Code Review / AI Code Review (pull_request) Failing after 1s Details CI/CD Pipeline / Check if frontend-only change (pull_request) Successful in 36s Details CI/CD Pipeline / Frontend Lint (pull_request) Successful in 38s Details CI/CD Pipeline / Validate - Type Check (mypy) (pull_request) Successful in 1m18s Details CI/CD Pipeline / Build Staging API Image (pull_request) Has been skipped Details CI/CD Pipeline / Build Staging Web Image (pull_request) Has been skipped Details CI/CD Pipeline / Build Staging Worker Image (pull_request) Has been skipped Details CI/CD Pipeline / Validate - Migration (alembic) (pull_request) Successful in 1m49s Details CI/CD Pipeline / PR Build Worker Image (pull_request) Successful in 41s Details Preview Deploy / Deploy Preview Environment (pull_request) Failing after 50s Details CI/CD Pipeline / PR Build Web Image (pull_request) Successful in 1m57s Details CI/CD Pipeline / Frontend Unit Tests (pull_request) Has been skipped Details CI/CD Pipeline / Build Production API Image (pull_request) Has been skipped Details CI/CD Pipeline / Build Production Web Image (pull_request) Has been skipped Details CI/CD Pipeline / Build Production Worker Image (pull_request) Has been skipped Details CI/CD Pipeline / Validate - Code Quality (pull_request) Successful in 4m54s Details CI/CD Pipeline / Deploy Staging (Watchtower auto-deploy) (pull_request) Has been skipped Details CI/CD Pipeline / Deploy Production (pull_request) Has been skipped Details CI/CD Pipeline / Staging E2E Tests (pull_request) Has been skipped Details CI/CD Pipeline / Staging API Integration Tests (pull_request) Has been skipped Details CI/CD Pipeline / ACR Image Cleanup (pull_request) Has been skipped Details CI/CD Pipeline / Production Browser E2E (pull_request) Has been skipped Details CI/CD Pipeline / PR Build API Image (pull_request) Successful in 4m15s Details CI/CD Pipeline / Unit Tests (pull_request) Successful in 2m33s Details PR Automation / Auto Approve on CI Green (pull_request) Successful in 3m54s Details CI/CD Pipeline / Integration Tests (pull_request) Successful in 1m55s Details Preview Cleanup / Cleanup Preview Environment (pull_request) Successful in 23s Details ACR Cleanup / ACR Image Cleanup (pull_request_target) Failing after 22s Details PR Automation / Auto Merge on CI Green + Approved (pull_request) Successful in 51m44s Details	2026-07-25 12:29:34 +08:00
CI Bot	78257fae96	test(P3-1): 第56波 worker层更多引擎纯逻辑单测（+108） CI/CD Pipeline / Check if frontend-only change (pull_request) Successful in 13s Details CI/CD Pipeline / Validate - Type Check (mypy) (pull_request) Successful in 1m4s Details CI/CD Pipeline / Validate - Migration (alembic) (pull_request) Successful in 1m2s Details CI/CD Pipeline / Frontend Lint (pull_request) Successful in 47s Details CI/CD Pipeline / Build Staging API Image (pull_request) Has been skipped Details CI/CD Pipeline / Build Staging Web Image (pull_request) Has been skipped Details CI/CD Pipeline / Build Staging Worker Image (pull_request) Has been skipped Details CI/CD Pipeline / PR Build Worker Image (pull_request) Successful in 38s Details CI/CD Pipeline / Validate - Code Quality (pull_request) Successful in 3m13s Details CI/CD Pipeline / PR Build Web Image (pull_request) Successful in 1m41s Details Preview Deploy / Deploy Preview Environment (pull_request) Failing after 15s Details CI/CD Pipeline / Frontend Unit Tests (pull_request) Has been skipped Details CI/CD Pipeline / Build Production API Image (pull_request) Has been skipped Details CI/CD Pipeline / Build Production Web Image (pull_request) Has been skipped Details CI/CD Pipeline / Build Production Worker Image (pull_request) Has been skipped Details CI/CD Pipeline / Deploy Staging (Watchtower auto-deploy) (pull_request) Has been skipped Details AI Code Review / AI Code Review (pull_request) Successful in 3m0s Details CI/CD Pipeline / Deploy Production (pull_request) Has been skipped Details CI/CD Pipeline / Staging E2E Tests (pull_request) Has been skipped Details CI/CD Pipeline / Staging API Integration Tests (pull_request) Has been skipped Details CI/CD Pipeline / ACR Image Cleanup (pull_request) Has been skipped Details CI/CD Pipeline / Production Browser E2E (pull_request) Has been skipped Details CI/CD Pipeline / PR Build API Image (pull_request) Successful in 4m10s Details CI/CD Pipeline / Integration Tests (pull_request) Successful in 1m59s Details PR Automation / Auto Approve on CI Green (pull_request) Successful in 3m17s Details PR Automation / Auto Merge on CI Green + Approved (pull_request) Successful in 49m36s Details CI/CD Pipeline / Unit Tests (pull_request) Failing after 1h14m0s Details - test_subtitle_render_engine.py: 53个（颜色转换/时间格式化/文字换行/ASS转义/样式） - test_sticker_engine.py: 27个（配置类/位置解析/常量/便捷函数） - test_tts_engine.py: 12个（数据类/入口判断/初始化） - test_render_audio_utils.py: 16个（clip_effective_duration/RenderContext）覆盖worker层4个模块的纯逻辑部分	2026-07-25 09:37:08 +08:00
		`@@ -1,2 +0,0 @@`
			`allowBuilds:`
			`esbuild: set this to true or false`