perf(ci): 扩展CI触发分支模式+增加手动触发入口

PR #343: 修复Worker镜像瘦身后shared模块无法导入的PYTHONPATH问题

将ENV PYTHONPATH=/app改为ENV PYTHONPATH=/app:/app/packages，解决ModuleNotFoundError: No module named "shared"问题。

.gitea/workflows/ci-cd.yml

@@ -5,19 +5,32 @@ on:
     - main
     - develop
     - feature/**
+    - feat/**
     - bugfix/**
+    - fix/**
     - hotfix/**
     - release/**
+    - refactor/**
+    - perf/**
+    - docs/**
+    - chore/**
+    - ci/**
     tags:
     - v*
   pull_request:
     branches:
     - main
     - develop
+  workflow_dispatch:
+    inputs:
+      reason:
+        description: "触发原因"
+        required: false
+        default: "手动触发 - CI漏触发补跑"
 permissions:
   contents: read
 concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
+  group: ${{ github.workflow }}-${{ github.event_name }}-${{ github.ref }}
   cancel-in-progress: true
 jobs:
   validate:
@@ -68,7 +81,7 @@ jobs:
 
         python3 -m isort --version-number
 
-        python3 -m flake8 --version
+        python3 -m ruff --version
 
         bandit --version
 
@@ -79,23 +92,14 @@ jobs:
       shell: sh
       run: "set -eu\necho \"=== Installing detect-secrets ===\"\npython3 -m pip install -q detect-secrets\ndetect-secrets --version\necho \"\"\necho \"=== Running secret scan ===\"\ndetect-secrets scan \\\n  --all-files \\\n  --exclude-files '(^|/)(tests|test|e2e|__tests__|spec|docs|node_modules|site-packages|migrations|alembic|.gitea|.git|.pytest_cache|.next|dist|build)/' \\\n  --exclude-files '\\.(md|rst|txt|lock|example|sample|min\\.js|min\\.css|spec\\.ts|test\\.ts|test\\.py)$' \\\n  --exclude-files '(package-lock|yarn\\.lock|poetry\\.lock|Pipfile\\.lock)$' \\\n  --disable-plugin Base64HighEntropyString \\\n  --disable-plugin HexHighEntropyString \\\n  --disable-plugin BasicAuthDetector \\\n  --disable-plugin KeywordDetector \\\n  --disable-plugin IPPublicDetector \\\n  2>&1 | tee /tmp/secrets-scan.json\n\nFOUND=$(python3 -c \"\nimport json\ntry:\n    with open('/tmp/secrets-scan.json') as f:\n        data = json.load(f)\n    results = data.get('results', {})\n    total = sum(len(v) for\
         \ v in results.values())\n    print(total)\nexcept Exception:\n    print('error')\n\")\necho \"\"\necho \"Secrets detected: $FOUND\"\nif [ \"$FOUND\" != \"0\" ] && [ \"$FOUND\" != \"error\" ]; then\n  echo \"\"\n  echo \"=== Secret details ===\"\n  python3 -c \"\nimport json\nwith open('/tmp/secrets-scan.json') as f:\n    data = json.load(f)\nfor fpath, items in data.get('results', {}).items():\n    for item in items:\n        line = item.get('line_number', '?')\n        stype = item.get('type', '?')\n        hashed = item.get('hashed_secret', '')[:16]\n        print(f'  {fpath}:{line} [{stype}] {hashed}...')\n\"\n  echo \"\"\n  echo \"ERROR: Potential secrets detected in code!\"\n  echo \"If these are false positives, add exclusions in the CI workflow.\"\n  exit 1\nfi\necho \"Secret scan completed - no secrets detected\"\n"
+    - name: Calculate changed Python files (incremental scan)
+      shell: sh
+      env:
+        GITHUB_TOKEN: ${{ github.token }}
+      run: "set -eu\nSCAN_MODE=\"full\"\nCHANGED_PY_FILES=\"\"\n\nif [ \"${GITHUB_EVENT_NAME:-}\" = \"pull_request\" ] && [ -n \"${GITHUB_BASE_REF:-}\" ]; then\n  echo \"PR mode (base: ${GITHUB_BASE_REF}) - preparing incremental scan\"\n  git init -q\n  git config user.email \"ci@localhost\"\n  git config user.name \"CI\"\n  git add .\n  git commit -q -m \"current\"\n  REPO_URL=\"https://x-access-token:${GITHUB_TOKEN}@${GITHUB_SERVER_URL#https://}/${GITHUB_REPOSITORY}.git\"\n  git remote add origin \"$REPO_URL\"\n  if git fetch origin \"${GITHUB_BASE_REF}\" --depth=50 -q 2>/dev/null; then\n    CHANGED_PY_FILES=$(git diff --name-only --diff-filter=ACMRT \"origin/${GITHUB_BASE_REF}\" HEAD -- '*.py' 2>/dev/null | tr '\\n' ' ')\n    if [ -n \"$CHANGED_PY_FILES\" ]; then\n      SCAN_MODE=\"incremental\"\n      FILE_COUNT=$(echo \"$CHANGED_PY_FILES\" | wc -w)\n      echo \"Changed Python files: ${FILE_COUNT}\"\n      echo \"$CHANGED_PY_FILES\" | tr ' ' '\\n' | grep -v '^$'\n    else\n      SCAN_MODE=\"skip_py\"\n      echo \"No Python files changed in this PR\"\n    fi\n  else\n    echo \"WARN: failed to fetch base branch, falling back to full scan\"\n  fi\nelse\n  echo \"Full scan mode (not a PR event)\"\nfi\n\necho \"SCAN_MODE=$SCAN_MODE\" >> $GITHUB_ENV\necho \"CHANGED_PY_FILES=$CHANGED_PY_FILES\" >> $GITHUB_ENV\n"
     - name: Run code quality checks
       shell: sh
-      run: 'set -eu
-
-        python3 -m compileall -q alembic apps packages tests scripts
-
-        python3 -m black --check --fast alembic apps packages tests scripts
-
-        python3 -m isort --check-only alembic apps packages tests scripts
-
-        python3 -m flake8 apps packages tests --count --statistics
-
-        '
-    - name: Ruff lint (advisory mode - 摸底阶段)
-      if: always()
-      shell: sh
-      run: "set +e\necho \"=== Installing ruff ===\"\npython3 -m pip install -q ruff\nruff --version\necho \"\"\necho \"=== Running ruff lint (advisory mode) ===\"\necho \"告警模式，不阻断CI。用于摸底问题数量，后续分批修复后正式替换flake8。\"\necho \"\"\nruff check apps packages tests scripts             --statistics             --output-format concise             2>&1 | tail -30\nEXIT_CODE=$?\necho \"\"\nif [ \"$EXIT_CODE\" != \"0\" ]; then\n  echo \"ruff 发现 lint 问题（告警模式，不阻断）\"\n  echo \"问题分类统计见上方，后续将分批修复\"\nelse\n  echo \"ruff 检查全部通过 ✅\"\nfi\nexit 0\n"
+      run: "set -eu\n\nif [ \"$SCAN_MODE\" = \"incremental\" ]; then\n  echo \"=== Incremental scan mode ===\"\n\n  python3 -m compileall -q $CHANGED_PY_FILES\n\n  python3 -m black --check --fast $CHANGED_PY_FILES\n\n  python3 -m isort --check-only $CHANGED_PY_FILES\n\n  RUFF_FILES=$(echo \"$CHANGED_PY_FILES\" | tr ' ' '\\n' | grep -v '^scripts/' | tr '\\n' ' ')\n  if [ -n \"$RUFF_FILES\" ]; then\n    python3 -m ruff check $RUFF_FILES --statistics\n  else\n    echo \"No ruff-checkable files changed, skipping\"\n  fi\n\nelif [ \"$SCAN_MODE\" = \"skip_py\" ]; then\n  echo \"No Python files changed - skipping Python lint checks\"\n\nelse\n  echo \"=== Full scan mode ===\"\n\n  python3 -m compileall -q alembic apps packages tests scripts\n\n  python3 -m black --check --fast alembic apps packages tests scripts\n\n  python3 -m isort --check-only alembic apps packages tests scripts\n\n  python3 -m ruff check apps packages tests --statistics\nfi\n"
     - name: Type check (mypy, advisory mode)
       if: always()
       shell: sh
@@ -142,7 +146,7 @@ jobs:
       shell: sh
       env:
         GITHUB_TOKEN: ${{ github.token }}
-      run: "set -eu\nif command -v git >/dev/null 2>&1; then\n  git init -q\n  git config user.email \"ci@localhost\"\n  git config user.name \"CI\"\n  git add .\n  git commit -q -m \"current\"\n  REPO_URL=\"https://x-access-token:${GITHUB_TOKEN}@${GITHUB_SERVER_URL#https://}/${GITHUB_REPOSITORY}.git\"\n  git remote add origin \"$REPO_URL\"\n  git fetch origin main --depth=1 -q 2>/dev/null || echo \"WARN: cannot fetch main, will check all migrations\"\nelse\n  echo \"WARN: git not available, will check all migrations\"\nfi\n"
+      run: "set -eu\nif command -v git >/dev/null 2>&1; then\n  if [ ! -d .git ]; then\n    git init -q\n    git config user.email \"ci@localhost\"\n    git config user.name \"CI\"\n    git add .\n    git commit -q -m \"current\"\n    REPO_URL=\"https://x-access-token:${GITHUB_TOKEN}@${GITHUB_SERVER_URL#https://}/${GITHUB_REPOSITORY}.git\"\n    git remote add origin \"$REPO_URL\"\n  fi\n  git fetch origin main --depth=1 -q 2>/dev/null || echo \"WARN: cannot fetch main, will check all migrations\"\nelse\n  echo \"WARN: git not available, will check all migrations\"\nfi\n"
     - name: Check migration safety
       shell: sh
       run: "set -eu\nif git rev-parse origin/main >/dev/null 2>&1; then\n  python3 scripts/check_migration_safety.py --allow-medium-risk --diff-against origin/main\nelse\n  python3 scripts/check_migration_safety.py --allow-medium-risk\nfi\n"
@@ -371,19 +375,19 @@ jobs:
         '
     - name: Install dependencies
       shell: sh
-      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm \\\n  -v \"$PWD:/workspace\" \\\n  -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\" -w /workspace/apps/web \\\n  docker.m.daocloud.io/library/node:20 \\\n  sh -lc 'rm -rf node_modules/* 2>/dev/null; npm ci --include=dev'\n"
+      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm \\\n  -v \"$PWD:/workspace\" \\\n  -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\" -w /workspace/apps/web \\\n  docker.m.daocloud.io/library/node:20 \\\n  sh -lc 'PACKAGE_LOCK_HASH=$(md5sum package-lock.json 2>/dev/null | cut -d\" \" -f1)\nCACHE_HASH_FILE=\"node_modules/.package-lock-hash\"\nCACHE_VALID=false\nif [ -f \"$CACHE_HASH_FILE\" ] && [ \"$(cat \"$CACHE_HASH_FILE\")\" = \"$PACKAGE_LOCK_HASH\" ] && [ -x \"node_modules/.bin/eslint\" ]; then\n  CACHE_VALID=true\n  echo \"Cache hit: dependencies valid, skipping npm ci\"\nfi\nif [ \"$CACHE_VALID\" = \"false\" ]; then\n  echo \"Cache miss or invalid: running npm ci...\"\n  if ! npm ci --include=dev; then\n    echo \"npm ci failed, cleaning node_modules and retrying...\"\n    rm -rf node_modules\n    mkdir -p node_modules\n    npm ci --include=dev\n  fi\n  echo \"$PACKAGE_LOCK_HASH\" > \"$CACHE_HASH_FILE\"\n  echo \"Dependencies installed, cache updated\"\nfi'\n"
     - name: Run ESLint
       shell: sh
-      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm \\\n  -v \"$PWD:/workspace\" \\\n  -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\" -w /workspace/apps/web \\\n  docker.m.daocloud.io/library/node:20 \\\n  sh -lc 'npx eslint src --ext .ts,.tsx --max-warnings 50'\n"
+      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm \\\n  -v \"$PWD:/workspace\" \\\n  -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\" -w /workspace/apps/web \\\n  docker.m.daocloud.io/library/node:20 \\\n  sh -lc 'npx --no-install eslint src --ext .ts,.tsx --max-warnings 50'\n"
     - name: Run TypeScript type check
       shell: sh
-      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm \\\n  -v \"$PWD:/workspace\" \\\n  -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\" -w /workspace/apps/web \\\n  docker.m.daocloud.io/library/node:20 \\\n  sh -lc 'npx tsc --noEmit'\n"
+      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm \\\n  -v \"$PWD:/workspace\" \\\n  -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\" -w /workspace/apps/web \\\n  docker.m.daocloud.io/library/node:20 \\\n  sh -lc 'npx --no-install tsc --noEmit'\n"
     - name: Run Prettier check
       shell: sh
-      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm \\\n  -v \"$PWD:/workspace\" \\\n  -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\" -w /workspace/apps/web \\\n  docker.m.daocloud.io/library/node:20 \\\n  sh -lc 'npx prettier --check \"src/**/*.{ts,tsx,md}\"'\n"
+      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm \\\n  -v \"$PWD:/workspace\" \\\n  -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\" -w /workspace/apps/web \\\n  docker.m.daocloud.io/library/node:20 \\\n  sh -lc 'npx --no-install prettier --check \"src/**/*.{ts,tsx,md}\"'\n"
     - name: Run Vitest tests
       shell: sh
-      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm \\\n  -v \"$PWD:/workspace\" \\\n  -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\" -w /workspace/apps/web \\\n  docker.m.daocloud.io/library/node:20 \\\n  sh -lc 'ls node_modules/.bin/vitest 2>/dev/null && node_modules/.bin/vitest run src/test || npm exec vitest run src/test'\n"
+      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm \\\n  -v \"$PWD:/workspace\" \\\n  -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\" -w /workspace/apps/web \\\n  docker.m.daocloud.io/library/node:20 \\\n  sh -lc 'npx --no-install vitest run src/test'\n"
     - name: Job duration summary
       if: always()
       shell: sh
@@ -565,7 +569,7 @@ jobs:
       run: "set -eu\n# develop/main 分支写回缓存，其他分支只读\nif [ \"${GITHUB_REF_NAME}\" = \"develop\" ] || [ \"${GITHUB_REF_NAME}\" = \"main\" ]; then\n  echo \"CACHE_MODE=read-write\" >> $GITHUB_ENV\n  echo \"Cache mode: read-write (will push cache)\"\nelse\n  echo \"CACHE_MODE=read-only\" >> $GITHUB_ENV\n  echo \"Cache mode: read-only\"\nfi\n"
     - name: Build frontend assets (npm build)
       shell: sh
-      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm           -v \"$PWD:/workspace\"           -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\"           -w /workspace/apps/web           docker.m.daocloud.io/library/node:20           sh -lc \"npm ci && npm run build\"\n\ntest -f apps/web/dist/index.html\necho \"Frontend build complete: $(ls apps/web/dist/ | head -5)\"\n"
+      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\n  echo \"Created npm cache volume: $NPM_CACHE_VOLUME\"\nfi\n\ndocker run --rm           -v \"$PWD:/workspace\"           -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\"           -w /workspace/apps/web           docker.m.daocloud.io/library/node:20           sh -lc \"npm ci && npx tsc --incremental --tsBuildInfoFile node_modules/.tsbuildinfo && npx vite build\"\n\ntest -f apps/web/dist/index.html\necho \"Frontend build complete: $(ls apps/web/dist/ | head -5)\"\n"
     - name: Setup buildx builder (docker-container driver)
       shell: sh
       run: "set -eu\n# 确保使用 docker-container driver 以支持 cache export 功能\nif ! docker buildx inspect ci-builder > /dev/null 2>&1; then\n  docker buildx create --use --name ci-builder --driver docker-container\n  echo \"Created ci-builder (docker-container driver)\"\nelse\n  docker buildx use ci-builder\n  echo \"Using existing ci-builder\"\nfi\ndocker buildx inspect --bootstrap\n"
@@ -620,7 +624,7 @@ jobs:
         CI_NOTIFY_WEBHOOK: ${{ secrets.CI_NOTIFY_WEBHOOK }}
       run: 'set +e
 
-        NOTIFY_MODE=start JOB_NAME="Deploy Staging (Watchtower auto-deploy)" python3 scripts/ci_notify.py
+        NOTIFY_MODE=start JOB_NAME="Deploy Staging" python3 scripts/ci_notify.py
 
         '
     - name: Docker login to Registry
@@ -634,14 +638,30 @@ jobs:
         echo "Docker login successful"
 
         '
-    - name: Tag and push :staging images (Watchtower auto-update)
+    - name: Install SSH client
+      if: success()
       shell: sh
-      run: "set -eu\nREGISTRY=\"git.xiaoxiajianji.com/xiaoxia/xiaoxia-saas\"\n\nfor svc in api worker web; do\n  echo \"Pulling ${REGISTRY}/xiaoxia-saas-${svc}:${GITHUB_SHA} ...\"\n  docker pull \"${REGISTRY}/xiaoxia-saas-${svc}:${GITHUB_SHA}\"\n  docker tag \"${REGISTRY}/xiaoxia-saas-${svc}:${GITHUB_SHA}\" \"${REGISTRY}/xiaoxia-saas-${svc}:staging\"\n  docker push \"${REGISTRY}/xiaoxia-saas-${svc}:staging\"\n  echo \"$svc :staging tagged and pushed\"\ndone\necho \"All :staging images pushed. Watchtower will auto-deploy within 60s.\"\n"
+      run: "set -eu\napt-get update -qq \u0026\u0026 apt-get install -y -qq openssh-client \u003e/dev/null 2\u003e\u00261\necho \"openssh-client installed\"\n"
+
+    - name: Deploy staging over SSH (Registry pull)
+      if: success()
+      shell: sh
+      env:
+        STAGING_SSH_HOST: ${{ secrets.STAGING_SSH_HOST }}
+        STAGING_SSH_USER: ${{ secrets.STAGING_SSH_USER }}
+        STAGING_SSH_PORT: ${{ secrets.STAGING_SSH_PORT }}
+        STAGING_SSH_KEY: ${{ secrets.STAGING_SSH_KEY }}
+        REGISTRY_TOKEN: ${{ secrets.REGISTRY_TOKEN }}
+      run: "set -eux\nstaging_host=\"${STAGING_SSH_HOST:-47.98.113.167}\"\nstaging_user=\"${STAGING_SSH_USER:-root}\"\nstaging_port=\"${STAGING_SSH_PORT:-22222}\"\necho \"Host: $staging_host\"\necho \"Port: $staging_port\"\n\nmkdir -p ~/.ssh\n\n# 查找可用的SSH密钥\nkey_path=\"\"\nif [ -f /root/.ssh/xiaoxia_runtime_builder ]; then\n  key_path=\"/root/.ssh/xiaoxia_runtime_builder\"\n  echo \"Using key: $key_path (builder key)\"\nelif [ -f \"$HOME/.ssh/xiaoxia_runtime_builder\" ]; then\n  key_path=\"$HOME/.ssh/xiaoxia_runtime_builder\"\n  echo \"Using key: $key_path (home key)\"\nelif [ -n \"${STAGING_SSH_KEY:-}\" ]; then\n  key_path=\"$HOME/.ssh/id_ed25519\"\n  printf \'%s\\n\' \"$STAGING_SSH_KEY\" \u003e \"$key_path\"\n  chmod 600 \"$key_path\"\n  echo \"Using key from STAGING_SSH_KEY secret\"\nelse\n  echo \"ERROR: No SSH key available\"\n  ls -la ~/.ssh/ 2\u003e/dev/null || true\n  ls -la /root/.ssh/ 2\u003e/dev/null || true\n  exit 1\nfi\n\nssh-keyscan -p \"$staging_port\" -H \"$staging_host\" \u003e\u003e ~/.ssh/known_hosts 2\u003e/dev/null\necho \"SSH keyscan done\"\n\n# 测试SSH连接\nssh -p \"$staging_port\" -i \"$key_path\" -o StrictHostKeyChecking=no \"${staging_user}@${staging_host}\" \"echo SSH_CONNECTION_OK \u0026\u0026 hostname\"\necho \"SSH connection verified\"\n\n# 读取部署脚本并通过SSH执行\ncat scripts/ci_staging_deploy.sh | ssh -p \"$staging_port\" -i \"$key_path\" -o StrictHostKeyChecking=no \"${staging_user}@${staging_host}\" \"IMAGE_TAG=\'${GITHUB_SHA}\' REGISTRY_TOKEN=\'${REGISTRY_TOKEN}\' sh\"\n"
     - name: Staging health check + auto rollback
       if: success()
       shell: sh
       env:
         CI_NOTIFY_WEBHOOK: ${{ secrets.CI_NOTIFY_WEBHOOK }}
+        STAGING_SSH_HOST: ${{ secrets.STAGING_SSH_HOST }}
+        STAGING_SSH_USER: ${{ secrets.STAGING_SSH_USER }}
+        STAGING_SSH_PORT: ${{ secrets.STAGING_SSH_PORT }}
+        STAGING_SSH_KEY: ${{ secrets.STAGING_SSH_KEY }}
         REGISTRY_TOKEN: ${{ secrets.REGISTRY_TOKEN }}
       run: 'set -eu
 
@@ -669,7 +689,7 @@ jobs:
         CI_NOTIFY_WEBHOOK: ${{ secrets.CI_NOTIFY_WEBHOOK }}
       run: 'set +e
 
-        NOTIFY_MODE=success JOB_NAME="Deploy Staging (Watchtower auto-deploy)" python3 scripts/ci_notify.py
+        NOTIFY_MODE=success JOB_NAME="Deploy Staging" python3 scripts/ci_notify.py
 
         '
     - name: Notify on failure
@@ -680,7 +700,7 @@ jobs:
         CI_NOTIFY_WEBHOOK: ${{ secrets.CI_NOTIFY_WEBHOOK }}
       run: 'set +e
 
-        NOTIFY_MODE=failure JOB_NAME="Deploy Staging (Watchtower auto-deploy)" python3 scripts/ci_notify.py
+        NOTIFY_MODE=failure JOB_NAME="Deploy Staging" python3 scripts/ci_notify.py
 
         '
   staging-e2e:
@@ -952,7 +972,7 @@ jobs:
         '
     - name: Build frontend assets (npm build)
       shell: sh
-      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\nfi\n\ndocker run --rm           -v \"$PWD:/workspace\"           -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\"           -w /workspace/apps/web           docker.m.daocloud.io/library/node:20           sh -lc \"npm ci && npm run build\"\n\ntest -f apps/web/dist/index.html\necho \"Frontend build complete\"\n"
+      run: "set -eu\nNPM_CACHE_VOLUME=\"xiaoxia-npm-cache\"\nif ! docker volume inspect \"$NPM_CACHE_VOLUME\" >/dev/null 2>&1; then\n  docker volume create \"$NPM_CACHE_VOLUME\" >/dev/null\nfi\n\ndocker run --rm           -v \"$PWD:/workspace\"           -v \"$NPM_CACHE_VOLUME:/workspace/apps/web/node_modules\"           -w /workspace/apps/web           docker.m.daocloud.io/library/node:20           sh -lc \"npm ci && npx tsc --incremental --tsBuildInfoFile node_modules/.tsbuildinfo && npx vite build\"\n\ntest -f apps/web/dist/index.html\necho \"Frontend build complete\"\n"
     - name: Setup buildx builder (docker-container driver)
       shell: sh
       run: "set -eu\n# 确保使用 docker-container driver 以支持 cache export 功能\nif ! docker buildx inspect ci-builder > /dev/null 2>&1; then\n  docker buildx create --use --name ci-builder --driver docker-container\n  echo \"Created ci-builder (docker-container driver)\"\nelse\n  docker buildx use ci-builder\n  echo \"Using existing ci-builder\"\nfi\ndocker buildx inspect --bootstrap\n"

apps/web/package-lock.json

@@ -33,6 +33,7 @@
         "eslint-plugin-react-hooks": "^4.6.2",
         "eslint-plugin-react-refresh": "^0.4.7",
         "jsdom": "^24.1.0",
+        "prettier": "^3.0.0",
         "typescript": "^5.5.3",
         "vite": "^5.3.1",
         "vitest": "^1.6.0"
@@ -4828,6 +4829,22 @@
         "node": ">= 0.8.0"
       }
     },
+    "node_modules/prettier": {
+      "version": "3.9.5",
+      "resolved": "https://registry.npmmirror.com/prettier/-/prettier-3.9.5.tgz",
+      "integrity": "sha512-/FVl766LpUfB5vXgCYOYa0MeV/441Ia99AeICQIQFTY/Nw0roZwULcXpku5i1/m5kt/baz+s4Zogspd839HSMg==",
+      "dev": true,
+      "license": "MIT",
+      "bin": {
+        "prettier": "bin/prettier.cjs"
+      },
+      "engines": {
+        "node": ">=14"
+      },
+      "funding": {
+        "url": "https://github.com/prettier/prettier?sponsor=1"
+      }
+    },
     "node_modules/pretty-format": {
       "version": "27.5.1",
       "resolved": "https://registry.npmjs.org/pretty-format/-/pretty-format-27.5.1.tgz",

apps/web/package.json

@@ -42,6 +42,7 @@
     "eslint-plugin-react-hooks": "^4.6.2",
     "eslint-plugin-react-refresh": "^0.4.7",
     "jsdom": "^24.1.0",
+    "prettier": "^3.0.0",
     "typescript": "^5.5.3",
     "vite": "^5.3.1",
     "vitest": "^1.6.0"

apps/web/vite.config.ts

@@ -37,6 +37,7 @@ export default defineConfig({
     },
   },
   build: {
+    cache: true,
     rollupOptions: {
       output: {
         manualChunks: {

infra/docker/worker.Dockerfile

@@ -1,48 +1,104 @@
 # ============================================================
-# Worker Dockerfile - 专门用于 Celery Worker
-# 优化：依赖分层缓存，基础大包和业务依赖分开
+# Worker Dockerfile - 优化版（多阶段构建 + 镜像瘦身）
+# 优化项：
+#   1. 多阶段构建：builder 阶段安装编译依赖，runtime 阶段只保留运行时
+#   2. ffmpeg 静态编译替换：从 apt 安装(457MB) 改为静态二进制(~80MB)
+#   3. Python 依赖瘦身：strip .so 调试符号 + 清理测试文件 + 清理缓存
 # ============================================================
 
-# 基础镜像：Python 3.12 + ffmpeg
-FROM git.xiaoxiajianji.com/xiaoxia/base/python:3.12-slim
+# ==================== Builder 阶段 ====================
+FROM git.xiaoxiajianji.com/xiaoxia/base/python:3.12-slim AS builder
 
-# 构建参数：版本号（CI 传入 commit hash）
+# 使用阿里云镜像加速
+RUN sed -i 's|deb.debian.org|mirrors.aliyun.com|g' /etc/apt/sources.list.d/debian.sources 2>/dev/null || \
+    sed -i 's|deb.debian.org|mirrors.aliyun.com|g' /etc/apt/sources.list 2>/dev/null || true
+
+# 安装编译工具（仅 builder 需要）
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    gcc \
+    g++ \
+    python3-dev \
+    binutils \
+    wget \
+    xz-utils \
+    && rm -rf /var/lib/apt/lists/*
+
+# ---- 下载静态编译 ffmpeg ----
+# 使用 johnvansickle.com 的静态编译版本（业界标准）
+RUN cd /tmp \
+    && wget -q https://johnvansickle.com/ffmpeg/releases/ffmpeg-release-amd64-static.tar.xz \
+    && tar xf ffmpeg-release-amd64-static.tar.xz \
+    && cp ffmpeg-*-amd64-static/ffmpeg /usr/local/bin/ffmpeg \
+    && cp ffmpeg-*-amd64-static/ffprobe /usr/local/bin/ffprobe \
+    && chmod +x /usr/local/bin/ffmpeg /usr/local/bin/ffprobe \
+    && rm -rf ffmpeg-*
+
+# ---- 安装 Python 依赖 ----
+WORKDIR /tmp
+
+# 创建 venv
+RUN python -m venv /opt/venv
+ENV PATH="/opt/venv/bin:$PATH"
+
+# 基础依赖
+COPY requirements-base.txt /tmp/requirements-base.txt
+RUN pip install --no-cache-dir -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host mirrors.aliyun.com \
+    -r /tmp/requirements-base.txt \
+    && rm /tmp/requirements-base.txt
+
+# Worker 专属大包
+COPY requirements-worker.txt /tmp/requirements-worker.txt
+RUN pip install --no-cache-dir -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host mirrors.aliyun.com \
+    -r /tmp/requirements-worker.txt \
+    && rm /tmp/requirements-worker.txt
+
+# 业务依赖
+COPY requirements.txt /tmp/requirements.txt
+RUN pip install --no-cache-dir -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host mirrors.aliyun.com \
+    -r /tmp/requirements.txt \
+    && rm /tmp/requirements.txt
+
+# ---- Python 依赖瘦身 ----
+# 1. strip .so 文件的调试符号（节省约 80-100MB）
+RUN find /opt/venv -name "*.so" -type f -exec strip --strip-all {} \; 2>/dev/null || true
+
+# 2. 清理测试文件（节省约 20MB）
+RUN find /opt/venv -type d -name "tests" -exec rm -rf {} + 2>/dev/null; \
+    find /opt/venv -type d -name "test" -exec rm -rf {} + 2>/dev/null; \
+    find /opt/venv -name "test_*.py" -delete 2>/dev/null || true
+
+# 3. 清理 .pyc 缓存和 __pycache__（节省约 10MB，运行时按需生成）
+RUN find /opt/venv -type d -name "__pycache__" -exec rm -rf {} + 2>/dev/null; \
+    find /opt/venv -name "*.pyc" -delete 2>/dev/null || true
+
+# 4. 清理 dist-info 中的文档
+RUN find /opt/venv -name "*.dist-info" -type d -exec sh -c 'rm -f "$1"/DESCRIPTION.rst "$1"/INSTALLER "$1"/LICENSE* "$1"/WHEEL "$1"/entry_points.txt' _ {} \; 2>/dev/null || true
+
+# ==================== Runtime 阶段 ====================
+FROM git.xiaoxiajianji.com/xiaoxia/base/python:3.12-slim AS runtime
+
+# 构建参数：版本号
 ARG APP_VERSION=dev
 
 # 使用阿里云镜像加速
 RUN sed -i 's|deb.debian.org|mirrors.aliyun.com|g' /etc/apt/sources.list.d/debian.sources 2>/dev/null || \
     sed -i 's|deb.debian.org|mirrors.aliyun.com|g' /etc/apt/sources.list 2>/dev/null || true
 
-# 安装系统依赖
+# 安装最小运行时依赖（opencv-python-headless 需要 libglib2.0-0）
 RUN apt-get update && apt-get install -y --no-install-recommends \
-    ffmpeg \
-    libsm6 \
-    libxext6 \
-    libgl1 \
+    libglib2.0-0 \
     && rm -rf /var/lib/apt/lists/*
 
+# 从 builder 复制 ffmpeg 静态二进制
+COPY --from=builder /usr/local/bin/ffmpeg /usr/local/bin/ffmpeg
+COPY --from=builder /usr/local/bin/ffprobe /usr/local/bin/ffprobe
+
+# 从 builder 复制 Python 虚拟环境
+COPY --from=builder /opt/venv /opt/venv
+
 # 设置工作目录
 WORKDIR /app
 
-# ---- 依赖分层：基础依赖（变化少，缓存命中率高）----
-COPY requirements-base.txt /tmp/requirements-base.txt
-
-RUN python -m venv /opt/venv \
-    && /opt/venv/bin/pip install --no-cache-dir -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host mirrors.aliyun.com -r /tmp/requirements-base.txt \
-    && rm /tmp/requirements-base.txt
-
-# ---- 依赖分层：Worker 专属大包（视频处理，变化极少）----
-COPY requirements-worker.txt /tmp/requirements-worker.txt
-
-RUN /opt/venv/bin/pip install --no-cache-dir -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host mirrors.aliyun.com -r /tmp/requirements-worker.txt \
-    && rm /tmp/requirements-worker.txt
-
-# ---- 依赖分层：业务依赖（变化频繁）----
-COPY requirements.txt /tmp/requirements.txt
-
-RUN /opt/venv/bin/pip install --no-cache-dir -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host mirrors.aliyun.com -r /tmp/requirements.txt \
-    && rm /tmp/requirements.txt
-
 # 复制应用代码
 COPY apps/worker/ /app/apps/worker/
 COPY apps/api/app/config.py /app/apps/api/app/config.py
@@ -51,13 +107,13 @@ COPY packages/ /app/packages/
 COPY alembic.ini /app/alembic.ini
 COPY migrations/ /app/migrations/
 
-# 复制 Worker 启动脚本（支持 WORKER_CONCURRENCY 环境变量）
+# 复制 Worker 启动脚本
 COPY infra/docker/entrypoint-worker.sh /usr/local/bin/entrypoint-worker.sh
 RUN chmod +x /usr/local/bin/entrypoint-worker.sh
 
 # 设置 Python 路径
 ENV PATH="/opt/venv/bin:$PATH"
-ENV PYTHONPATH=/app
+ENV PYTHONPATH=/app:/app/packages
 ENV PYTHONUNBUFFERED=1
 ENV APP_VERSION=$APP_VERSION
 

packages/application/tts_job/audio_merger.py

@@ -11,7 +11,7 @@ import shutil
 import tempfile
 from subprocess import CalledProcessError, TimeoutExpired
 
-from shared.ffmpeg_utils import FFMPEG_BIN, run_ffmpeg
+from packages.shared.ffmpeg_utils import FFMPEG_BIN, run_ffmpeg
 
 logger = logging.getLogger(__name__)
 

pyproject.toml

@@ -66,18 +66,18 @@ exclude = [
 ]
 
 [tool.ruff.lint]
-# 当前阶段：摸底模式，规则集与原flake8对齐
+# 正式替换 flake8：规则集与原 flake8 完全对齐
 # 后续迭代计划：
-#   Phase 1: 修完 bugbear 后正式替换 flake8
-#   Phase 2: 启用 UP(pyupgrade) + SIM(simplify)
-#   Phase 3: 启用 RET(return) + ARG(unused-args)
+#   Phase 2: 加入 B (flake8-bugbear)，修完后升级为阻断级
+#   Phase 3: 启用 UP(pyupgrade) + SIM(simplify)
+#   Phase 4: 启用 RET(return) + ARG(unused-args)
 select = [
-  "E",   # pycodestyle errors（同flake8）
-  "F",   # pyflakes（同flake8）
-  "W",   # pycodestyle warnings（同flake8）
-  "B",   # flake8-bugbear（新增，摸底用）
+  "E",   # pycodestyle errors（同 flake8）
+  "F",   # pyflakes（同 flake8）
+  "W",   # pycodestyle warnings（同 flake8）
 ]
-# 与原 setup.cfg flake8 配置对齐，确保不新增阻断
+# 与原 setup.cfg + .flake8 的 flake8 配置完全对齐
+# 注意：W503 在 ruff≥0.14 中已被移除（行为变默认），故不列入
 ignore = [
   "E203",
   "E501",  # line-too-long（black管）
@@ -90,12 +90,14 @@ ignore = [
   "F403",
   "F405",
   "F841",  # unused-variable
-  "B008",  # do-not-perform-callback-from-arg（fastapi依赖注入）
 ]
 
 [tool.ruff.lint.per-file-ignores]
 "__init__.py" = ["F401", "F403", "F405"]
-"tests/*" = ["E402", "F401", "F841"]
-"packages/ports/*" = ["E301"]
+"tests/*" = ["E402", "F401", "F821", "F841"]
+"packages/ports/*" = ["E301"]  # E704 在 ruff≥0.14 已移除
+"apps/api/app/api/routes/auth.py" = ["ALL"]
+"apps/api/app/api/routes/workspaces.py" = ["ALL"]
+"apps/api/app/middleware/auth.py" = ["ALL"]
 "apps/*/migrations/*" = ["ALL"]
 "alembic/*" = ["ALL"]

requirements-dev.txt

@@ -3,7 +3,7 @@
 # 代码质量
 black==26.5.1
 isort==8.0.1
-flake8==7.3.0
+ruff==0.14.0
 bandit==1.9.4
 
 # 测试

scripts/ci_staging_deploy.sh

@@ -0,0 +1,250 @@
+#!/bin/sh
+# ===========================================
+# Staging 部署脚本（SSH 模式）
+# ===========================================
+# 通过 SSH 在 staging 服务器上执行
+#
+# 环境变量:
+#   IMAGE_TAG       - 镜像版本 tag（如 commit SHA 或分支名）
+#   REGISTRY_TOKEN  - Registry 访问令牌
+#   REGISTRY        - Registry 地址（默认 git.xiaoxiajianji.com/xiaoxia/xiaoxia-saas）
+#   REGISTRY_USER   - Registry 用户名（默认 xiaoxia）
+#   ENV_FILE        - 环境变量文件路径
+#   GENERATED_DIR   - 生成文件目录
+#   SKIP_MIGRATION  - 跳过数据库迁移（true/false，默认 false）
+
+set -eu
+
+IMAGE_TAG="${IMAGE_TAG:-}"
+REGISTRY="${REGISTRY:-git.xiaoxiajianji.com/xiaoxia/xiaoxia-saas}"
+REGISTRY_USER="${REGISTRY_USER:-xiaoxia}"
+REGISTRY_TOKEN="${REGISTRY_TOKEN:-}"
+
+ENV_FILE="${ENV_FILE:-/var/lib/xiaoxia-saas-staging/.env}"
+GENERATED_DIR="${GENERATED_DIR:-/var/lib/xiaoxia-saas-staging/generated}"
+LEGACY_ASSETS_DIR="${LEGACY_ASSETS_DIR:-/var/lib/xiaoxia-saas-staging/legacy-assets}"
+
+SKIP_MIGRATION="${SKIP_MIGRATION:-false}"
+
+if [ -z "$IMAGE_TAG" ]; then
+  echo "ERROR: IMAGE_TAG is required"
+  exit 1
+fi
+
+test -f "$ENV_FILE"
+mkdir -p "$GENERATED_DIR"
+mkdir -p "$LEGACY_ASSETS_DIR"
+
+echo "=========================================="
+echo "  Staging 部署 - $IMAGE_TAG"
+echo "=========================================="
+
+# ---- 登录 Registry ----
+if [ -n "$REGISTRY_TOKEN" ]; then
+  echo "Logging in to registry: $REGISTRY"
+  REGISTRY_HOST=$(echo "$REGISTRY" | cut -d/ -f1)
+  printf %s "$REGISTRY_TOKEN" | docker login "$REGISTRY_HOST" -u "$REGISTRY_USER" --password-stdin 2>/dev/null || {
+    echo "WARN: docker login failed, will try to pull anyway"
+  }
+fi
+
+# ---- Pull 新版本镜像 ----
+REGISTRY_API="${REGISTRY}/xiaoxia-saas-api:${IMAGE_TAG}"
+REGISTRY_WORKER="${REGISTRY}/xiaoxia-saas-worker:${IMAGE_TAG}"
+REGISTRY_WEB="${REGISTRY}/xiaoxia-saas-web:${IMAGE_TAG}"
+
+LOCAL_API="xiaoxia-saas-api:${IMAGE_TAG}"
+LOCAL_WORKER="xiaoxia-saas-worker:${IMAGE_TAG}"
+LOCAL_WEB="xiaoxia-saas-web:${IMAGE_TAG}"
+
+echo "Pulling API image..."
+docker pull "$REGISTRY_API"
+echo "Pulling Worker image..."
+docker pull "$REGISTRY_WORKER"
+echo "Pulling Web image..."
+docker pull "$REGISTRY_WEB"
+
+# Re-tag 成本地名
+docker tag "$REGISTRY_API" "$LOCAL_API"
+docker tag "$REGISTRY_WORKER" "$LOCAL_WORKER"
+docker tag "$REGISTRY_WEB" "$LOCAL_WEB"
+echo "All images pulled and tagged."
+
+# ---- 备份 legacy assets ----
+echo "Backing up legacy assets from current web container..."
+if docker inspect xiaoxia-web-staging >/dev/null 2>&1; then
+  _tmpdir="/tmp/legacy-assets-$$"
+  rm -rf "$_tmpdir"
+  mkdir -p "$_tmpdir"
+  docker cp xiaoxia-web-staging:/usr/share/nginx/html/assets/. "$_tmpdir/" 2>/dev/null || true
+  # 只有目录非空才拷贝，避免覆盖有内容的 legacy assets
+  if [ -d "$_tmpdir" ] && [ "$(ls -A "$_tmpdir" 2>/dev/null)" ]; then
+    cp -an "$_tmpdir"/. "$LEGACY_ASSETS_DIR"/ 2>/dev/null || true
+    echo "Legacy assets backed up: $(ls "$_tmpdir" | wc -l) files"
+  fi
+  rm -rf "$_tmpdir"
+else
+  echo "No existing web container, skipping legacy assets backup"
+fi
+
+# 清理 7 天前的 legacy assets
+if [ -d "$LEGACY_ASSETS_DIR" ]; then
+  find "$LEGACY_ASSETS_DIR" -type f -mtime +7 -delete 2>/dev/null || true
+  echo "Legacy assets cleanup done (retain 7 days)"
+fi
+
+# ---- 检查基础设施容器 ----
+echo "Checking infrastructure containers..."
+for c in xiaoxia-postgres-staging xiaoxia-redis-staging; do
+  if ! docker inspect "$c" >/dev/null 2>&1; then
+    echo "ERROR: Required container not found: $c"
+    exit 1
+  fi
+  state=$(docker inspect -f '{{.State.Status}}' "$c")
+  if [ "$state" != "running" ]; then
+    echo "ERROR: Container not running: $c ($state)"
+    exit 1
+  fi
+done
+
+# ---- 创建网络（不存在则创建） ----
+docker network create xiaoxia-net-staging 2>/dev/null || true
+
+# ---- 数据库迁移 ----
+if [ "$SKIP_MIGRATION" != "true" ]; then
+  echo "Running database migrations..."
+  docker run --rm \
+    --env-file "$ENV_FILE" \
+    --network xiaoxia-net-staging \
+    -e APP_ENV=staging \
+    "$LOCAL_API" sh -c "cd /app && alembic upgrade head"
+  echo "Migrations completed."
+else
+  echo "Skipping migrations (SKIP_MIGRATION=true)"
+fi
+
+# ---- 停止旧容器 ----
+echo "Stopping old containers..."
+docker rm -f xiaoxia-api-staging 2>/dev/null || true
+docker rm -f xiaoxia-worker-staging 2>/dev/null || true
+docker rm -f xiaoxia-web-staging 2>/dev/null || true
+
+LOG_OPTS="--log-driver json-file --log-opt max-size=50m --log-opt max-file=3"
+
+# ---- 启动 API ----
+echo "Starting API container..."
+docker run -d \
+  --name xiaoxia-api-staging \
+  --env-file "$ENV_FILE" \
+  --network xiaoxia-net-staging \
+  -p 127.0.0.1:8000:8000 \
+  -e APP_ENV=staging \
+  -e APP_VERSION="$IMAGE_TAG" \
+  -e GENERATED_FILES_DIR=/app/generated \
+  -e GENERATED_FILES_URL_PREFIX=/generated-files \
+  -e PUBLIC_API_BASE_URL=https://staging-api.xiaoxiajianji.com \
+  -v "$GENERATED_DIR:/app/generated" \
+  --restart unless-stopped \
+  --health-cmd "python -c \"import urllib.request; urllib.request.urlopen('http://localhost:8000/health', timeout=5)\"" \
+  --health-interval 30s \
+  --health-timeout 10s \
+  --health-retries 3 \
+  --health-start-period 40s \
+  $LOG_OPTS \
+  "$LOCAL_API"
+
+# ---- 启动 Worker ----
+echo "Starting Worker container..."
+docker run -d \
+  --name xiaoxia-worker-staging \
+  --env-file "$ENV_FILE" \
+  --network xiaoxia-net-staging \
+  -e APP_ENV=staging \
+  -e APP_VERSION="$IMAGE_TAG" \
+  -e WORKER_CONCURRENCY=1 \
+  -e WORKER_MAX_TASKS_PER_CHILD=100 \
+  -e GENERATED_FILES_DIR=/app/generated \
+  -e GENERATED_FILES_URL_PREFIX=/generated-files \
+  -e PUBLIC_API_BASE_URL=https://staging-api.xiaoxiajianji.com \
+  -v "$GENERATED_DIR:/app/generated" \
+  --restart unless-stopped \
+  --health-cmd "sh -c \"grep -q celery /proc/1/cmdline || exit 1\"" \
+  --health-interval 30s \
+  --health-timeout 10s \
+  --health-retries 3 \
+  --health-start-period 30s \
+  $LOG_OPTS \
+  "$LOCAL_WORKER"
+
+# ---- 启动 Web ----
+LEGACY_VOLUME=""
+if [ -d "$LEGACY_ASSETS_DIR" ] && [ "$(ls -A "$LEGACY_ASSETS_DIR" 2>/dev/null)" ]; then
+  LEGACY_VOLUME="-v ${LEGACY_ASSETS_DIR}:/usr/share/nginx/html/assets-legacy/assets:ro"
+  echo "Web container: legacy assets mounted (fallback)"
+else
+  echo "Web container: no legacy assets to mount"
+fi
+
+echo "Starting Web container..."
+docker run -d \
+  --name xiaoxia-web-staging \
+  --network xiaoxia-net-staging \
+  -p 127.0.0.1:3001:80 \
+  --restart unless-stopped \
+  $LEGACY_VOLUME \
+  --health-cmd "wget --spider -q http://127.0.0.1:80" \
+  --health-interval 30s \
+  --health-timeout 5s \
+  --health-retries 3 \
+  $LOG_OPTS \
+  "$LOCAL_WEB"
+
+# ---- 等待 API 健康 ----
+echo "Waiting for API to become healthy..."
+i=0
+while [ "$i" -lt 40 ]; do
+  if curl -sf --max-time 5 http://127.0.0.1:8000/health >/dev/null 2>&1; then
+    echo "API is healthy!"
+    break
+  fi
+  i=$((i + 1))
+  echo "  Waiting... ($i/40)"
+  sleep 3
+done
+
+if [ "$i" -ge 40 ]; then
+  echo "ERROR: API did not become healthy within 120s"
+  docker logs --tail 50 xiaoxia-api-staging
+  exit 1
+fi
+
+# ---- 等待 Web 健康 ----
+echo "Waiting for Web to become healthy..."
+i=0
+while [ "$i" -lt 15 ]; do
+  if curl -sf --max-time 5 http://127.0.0.1:3001/ >/dev/null 2>&1; then
+    echo "Web is healthy!"
+    break
+  fi
+  i=$((i + 1))
+  echo "  Waiting... ($i/15)"
+  sleep 2
+done
+
+if [ "$i" -ge 15 ]; then
+  echo "ERROR: Web did not become healthy within 30s"
+  docker logs --tail 30 xiaoxia-web-staging
+  exit 1
+fi
+
+# ---- 清理旧镜像 ----
+echo "Cleaning up old images..."
+docker image prune -af --filter "until=168h" 2>/dev/null || true
+docker builder prune -af --filter "until=168h" 2>/dev/null || true
+
+echo ""
+echo "=== Staging deployment complete ==="
+echo "API:      http://127.0.0.1:8000"
+echo "Web:      http://127.0.0.1:3001"
+echo "Version:  $IMAGE_TAG"
+docker ps --format "table {{.Names}}\t{{.Status}}\t{{.Image}}" | grep staging

scripts/ci_staging_healthcheck.sh

@@ -1,10 +1,10 @@
 #!/bin/bash
 # ===========================================
-# CI Staging 健康检查 + 自动回滚脚本（Watchtower 模式）
+# CI Staging 健康检查 + 自动回滚脚本（SSH 部署模式）
 # ===========================================
 #
 # 在 CI Runner 上执行，通过公网 URL 检查 Staging 部署健康状态。
-# 不健康则自动回滚（把 :staging tag 指回旧版本镜像，Watchtower 检测到后自动回滚）。
+# 不健康则通过 SSH 自动回滚到上一个版本的镜像。
 #
 # 用法:
 #   ./ci_staging_healthcheck.sh
@@ -13,14 +13,15 @@
 #   STAGING_API_URL       - Staging API 地址 (默认 https://staging-api.xiaoxiajianji.com)
 #   STAGING_WEB_URL       - Staging Web 地址 (默认 https://staging.xiaoxiajianji.com)
 #   HEALTH_CHECK_TIMEOUT  - 健康检查总超时秒数 (默认 120)
-#   WATCHTOWER_WAIT       - 等待 Watchtower 检测更新的秒数 (默认 90)
 #   SKIP_ROLLBACK         - 失败时不自动回滚 (true/false, 默认 false)
 #   SKIP_NOTIFY           - 跳过通知 (true/false, 默认 false)
 #   CI_NOTIFY_WEBHOOK     - 通知 Webhook URL
 #
-#   REGISTRY_TOKEN        - Gitea Registry Token（回滚时需要）
-#   REGISTRY_HOST         - Registry 地址 (默认 git.xiaoxiajianji.com)
-#   REGISTRY_REPO         - 仓库路径 (默认 xiaoxia/xiaoxia-saas)
+#   STAGING_SSH_HOST      - Staging 服务器 SSH 地址 (默认 47.98.113.167)
+#   STAGING_SSH_USER      - SSH 用户名 (默认 root)
+#   STAGING_SSH_PORT      - SSH 端口 (默认 22222)
+#   STAGING_SSH_KEY       - SSH 私钥内容
+#   REGISTRY_TOKEN        - Registry Token（回滚时拉取旧镜像需要）
 #
 #   GITHUB_SHA            - 当前 commit SHA
 #   GITHUB_REF_NAME       - 分支名
@@ -36,12 +37,15 @@ SCRIPT_DIR="$(CDPATH= cd -- "$(dirname -- "$0")" && pwd)"
 STAGING_API_URL="${STAGING_API_URL:-https://staging-api.xiaoxiajianji.com}"
 STAGING_WEB_URL="${STAGING_WEB_URL:-https://staging.xiaoxiajianji.com}"
 HEALTH_CHECK_TIMEOUT="${HEALTH_CHECK_TIMEOUT:-120}"
-WATCHTOWER_WAIT="${WATCHTOWER_WAIT:-90}"
 SKIP_ROLLBACK="${SKIP_ROLLBACK:-false}"
 SKIP_NOTIFY="${SKIP_NOTIFY:-false}"
 
-REGISTRY_HOST="${REGISTRY_HOST:-git.xiaoxiajianji.com}"
-REGISTRY_REPO="${REGISTRY_REPO:-xiaoxia/xiaoxia-saas}"
+STAGING_SSH_HOST="${STAGING_SSH_HOST:-47.98.113.167}"
+STAGING_SSH_USER="${STAGING_SSH_USER:-root}"
+STAGING_SSH_PORT="${STAGING_SSH_PORT:-22222}"
+
+REGISTRY="${REGISTRY:-git.xiaoxiajianji.com/xiaoxia/xiaoxia-saas}"
+REGISTRY_USER="${REGISTRY_USER:-xiaoxia}"
 
 # 颜色
 RED='\033[0;31m'
@@ -56,59 +60,85 @@ log_error() { echo -e "${RED}[ERROR]${NC} $1"; }
 log_step()  { echo -e "${BLUE}[STEP]${NC} $1"; }
 
 # ===========================================
-# 1. 记录部署前 :staging 镜像 digest（用于回滚）
+# SSH 工具函数
 # ===========================================
+SSH_KEY_PATH=""
+
+setup_ssh() {
+    # 查找或创建 SSH 密钥
+    if [ -f /root/.ssh/xiaoxia_runtime_builder ]; then
+        SSH_KEY_PATH="/root/.ssh/xiaoxia_runtime_builder"
+    elif [ -f "$HOME/.ssh/xiaoxia_runtime_builder" ]; then
+        SSH_KEY_PATH="$HOME/.ssh/xiaoxia_runtime_builder"
+    elif [ -n "${STAGING_SSH_KEY:-}" ]; then
+        SSH_KEY_PATH="$HOME/.ssh/staging_deploy_key"
+        mkdir -p "$HOME/.ssh"
+        printf '%s\n' "$STAGING_SSH_KEY" > "$SSH_KEY_PATH"
+        chmod 600 "$SSH_KEY_PATH"
+    else
+        log_error "没有可用的 SSH 密钥"
+        return 1
+    fi
+
+    ssh-keyscan -p "$STAGING_SSH_PORT" -H "$STAGING_SSH_HOST" >> ~/.ssh/known_hosts 2>/dev/null || true
+    log_info "SSH 已配置: ${STAGING_SSH_USER}@${STAGING_SSH_HOST}:${STAGING_SSH_PORT}"
+}
+
+run_ssh() {
+    local cmd="$1"
+    ssh -p "$STAGING_SSH_PORT" -i "$SSH_KEY_PATH" -o StrictHostKeyChecking=no \
+        "${STAGING_SSH_USER}@${STAGING_SSH_HOST}" "$cmd"
+}
+
+# ===========================================
+# 1. 记录部署前各服务的镜像版本（用于回滚）
+# ===========================================
+ROLLBACK_API_TAG=""
+ROLLBACK_WORKER_TAG=""
+ROLLBACK_WEB_TAG=""
+
 save_rollback_target() {
-    log_step "记录当前 :staging 镜像 digest（回滚目标）..."
+    log_step "记录当前 staging 各服务镜像版本（回滚目标）..."
 
-    # 登录 Registry
-    if [ -n "${REGISTRY_TOKEN:-}" ]; then
-        printf '%s' "${REGISTRY_TOKEN}" | docker login "${REGISTRY_HOST}" -u xiaoxia --password-stdin 2>/dev/null || true
-    fi
+    # 通过 SSH 获取当前运行的容器镜像
+    local api_image worker_image web_image
+    api_image=$(run_ssh "docker inspect --format '{{.Config.Image}}' xiaoxia-api-staging 2>/dev/null || echo ''")
+    worker_image=$(run_ssh "docker inspect --format '{{.Config.Image}}' xiaoxia-worker-staging 2>/dev/null || echo ''")
+    web_image=$(run_ssh "docker inspect --format '{{.Config.Image}}' xiaoxia-web-staging 2>/dev/null || echo ''")
 
-    ROLLBACK_IMAGES=""
+    # 提取 tag（镜像名是 xiaoxia-saas-api:abc123 或 git.xiaoxiajianji.com/.../xiaoxia-saas-api:staging 格式）
+    ROLLBACK_API_TAG=$(echo "$api_image" | sed 's/.*://' || echo "")
+    ROLLBACK_WORKER_TAG=$(echo "$worker_image" | sed 's/.*://' || echo "")
+    ROLLBACK_WEB_TAG=$(echo "$web_image" | sed 's/.*://' || echo "")
 
-    for svc in api worker web; do
-        image="${REGISTRY_HOST}/${REGISTRY_REPO}/xiaoxia-saas-${svc}:staging"
-        # pull 当前 :staging 镜像获取 digest
-        if docker pull "$image" 2>/dev/null; then
-            digest=$(docker inspect -f '{{index .RepoDigests 0}}' "$image" 2>/dev/null || echo "")
-            if [ -n "$digest" ]; then
-                ROLLBACK_IMAGES="${ROLLBACK_IMAGES}${svc}:${digest} "
-                log_info "  $svc: $digest"
-            else
-                log_warn "  $svc: 无法获取 digest，该服务将不参与回滚"
-            fi
+    log_info "  API:     ${ROLLBACK_API_TAG:-未知}"
+    log_info "  Worker:  ${ROLLBACK_WORKER_TAG:-未知}"
+    log_info "  Web:     ${ROLLBACK_WEB_TAG:-未知}"
+
+    # 验证三个服务版本是否一致
+    if [ -n "$ROLLBACK_API_TAG" ] && [ -n "$ROLLBACK_WORKER_TAG" ] && [ -n "$ROLLBACK_WEB_TAG" ]; then
+        if [ "$ROLLBACK_API_TAG" = "$ROLLBACK_WORKER_TAG" ] && [ "$ROLLBACK_API_TAG" = "$ROLLBACK_WEB_TAG" ]; then
+            log_info "  ✅ 三个服务版本一致: $ROLLBACK_API_TAG"
+            export ROLLBACK_TAG="$ROLLBACK_API_TAG"
         else
-            log_warn "  $svc: 拉取 :staging 镜像失败，该服务将不参与回滚"
+            log_warn "  ⚠️  三个服务版本不一致，回滚时将分别使用各自版本"
+            export ROLLBACK_API_TAG ROLLBACK_WORKER_TAG ROLLBACK_WEB_TAG
+            export ROLLBACK_TAG_MIXED="true"
         fi
-    done
-
-    if [ -z "$ROLLBACK_IMAGES" ]; then
-        log_warn "没有获取到任何可回滚镜像，回滚功能将不可用"
+    else
+        log_warn "  ⚠️  未能获取全部服务版本，回滚功能可能受限"
     fi
-
-    export ROLLBACK_IMAGES
 }
 
 # ===========================================
-# 2. 等待 Watchtower 更新
-# ===========================================
-wait_for_watchtower() {
-    log_step "等待 Watchtower 检测新镜像并更新（${WATCHTOWER_WAIT}s）..."
-    sleep "$WATCHTOWER_WAIT"
-    log_info "等待结束，开始健康检查"
-}
-
-# ===========================================
-# 3. 健康检查
+# 2. 健康检查（公网视角）
 # ===========================================
 health_check() {
     local timeout="$HEALTH_CHECK_TIMEOUT"
     local start_time
     start_time=$(date +%s)
 
-    log_step "健康检查（超时 ${timeout}s）..."
+    log_step "公网健康检查（超时 ${timeout}s）..."
     log_info "  API: ${STAGING_API_URL}/health"
     log_info "  Web: ${STAGING_WEB_URL}/"
 
@@ -171,69 +201,183 @@ health_check() {
 }
 
 # ===========================================
-# 4. 执行回滚（把 :staging tag 改回旧版本）
+# 3. 执行回滚（SSH 重新部署旧版本）
 # ===========================================
 do_rollback() {
-    log_step "执行回滚：将 :staging tag 指回旧版本镜像..."
+    log_step "执行回滚：通过 SSH 重新部署旧版本镜像..."
 
-    if [ -z "${ROLLBACK_IMAGES:-}" ]; then
-        log_error "没有可回滚的镜像记录，无法自动回滚"
+    local rollback_tag="${ROLLBACK_TAG:-}"
+    if [ -z "$rollback_tag" ] && [ "${ROLLBACK_TAG_MIXED:-}" != "true" ]; then
+        log_error "没有可回滚的版本记录，无法自动回滚"
         return 1
     fi
 
-    local rollback_ok=true
+    # 如果版本不一致，用 API 的版本作为回滚目标
+    if [ -z "$rollback_tag" ]; then
+        rollback_tag="$ROLLBACK_API_TAG"
+    fi
 
-    for entry in $ROLLBACK_IMAGES; do
-        svc="${entry%%:*}"
-        digest="${entry#*:}"
-        # digest 是 host/repo/image@sha256:xxx 格式，entry 里只取了第一部分，需要重新组合
-        # 实际上 ROLLBACK_IMAGES 存的是 svc:repo/image@sha256:xxx
-        # 让我们重新从完整digest中提取
-        image_digest="${entry#*:}:${entry#*:*:}"
-        # 不对，格式是 svc:REGISTRY_HOST/REGISTRY_REPO/xiaoxia-saas-svc@sha256:xxx
-        # 让我重新组织一下
-        :
-    done
+    if [ -z "$rollback_tag" ]; then
+        log_error "无法确定回滚版本"
+        return 1
+    fi
 
-    # 上面的解析有问题，重新来
-    # ROLLBACK_IMAGES 格式: "api:digest1 worker:digest2 web:digest3"
-    # 但digest本身带冒号(@sha256:xxx)，所以分隔有问题
-    # 改用数组方式
-    log_info "开始回滚各服务..."
+    log_info "回滚目标版本: $rollback_tag"
 
-    local IFS=' '
-    for pair in $ROLLBACK_IMAGES; do
-        svc="${pair%%:*}"
-        # 剩余部分是完整的 digest 字符串（含 @sha256:xxx）
-        digest="${pair#*:}"
+    # 构建回滚脚本（直接部署旧版本镜像，不跑 migration）
+    local rollback_script=$(cat << 'ROLLBACK_EOF'
+#!/bin/sh
+set -eu
 
-        image_name="${REGISTRY_HOST}/${REGISTRY_REPO}/xiaoxia-saas-${svc}"
+IMAGE_TAG="$1"
+REGISTRY_TOKEN="$2"
+REGISTRY="${REGISTRY:-git.xiaoxiajianji.com/xiaoxia/xiaoxia-saas}"
+REGISTRY_USER="${REGISTRY_USER:-xiaoxia}"
 
-        log_info "  回滚 $svc 到: $digest"
+ENV_FILE="${ENV_FILE:-/var/lib/xiaoxia-saas-staging/.env}"
+GENERATED_DIR="${GENERATED_DIR:-/var/lib/xiaoxia-saas-staging/generated}"
+LEGACY_ASSETS_DIR="${LEGACY_ASSETS_DIR:-/var/lib/xiaoxia-saas-staging/legacy-assets}"
 
-        # 通过 digest 拉取旧镜像
-        if docker pull "$digest" 2>/dev/null; then
-            # 重新打 :staging tag
-            docker tag "$digest" "${image_name}:staging"
-            docker push "${image_name}:staging"
-            log_info "  ✅ $svc 已回滚"
-        else
-            log_error "  ❌ $svc 回滚失败：无法拉取旧镜像"
-            rollback_ok=false
-        fi
-    done
+echo "=== Rollback to $IMAGE_TAG ==="
 
-    if [ "$rollback_ok" = true ]; then
-        log_info "所有服务回滚命令执行完成，Watchtower 将在 ~60s 内检测到并更新"
+# 登录 Registry
+if [ -n "$REGISTRY_TOKEN" ]; then
+    REGISTRY_HOST=$(echo "$REGISTRY" | cut -d/ -f1)
+    printf %s "$REGISTRY_TOKEN" | docker login "$REGISTRY_HOST" -u "$REGISTRY_USER" --password-stdin 2>/dev/null || true
+fi
+
+# Pull 旧版本镜像
+LOCAL_API="xiaoxia-saas-api:${IMAGE_TAG}"
+LOCAL_WORKER="xiaoxia-saas-worker:${IMAGE_TAG}"
+LOCAL_WEB="xiaoxia-saas-web:${IMAGE_TAG}"
+
+docker pull "${REGISTRY}/xiaoxia-saas-api:${IMAGE_TAG}"
+docker pull "${REGISTRY}/xiaoxia-saas-worker:${IMAGE_TAG}"
+docker pull "${REGISTRY}/xiaoxia-saas-web:${IMAGE_TAG}"
+
+docker tag "${REGISTRY}/xiaoxia-saas-api:${IMAGE_TAG}" "$LOCAL_API"
+docker tag "${REGISTRY}/xiaoxia-saas-worker:${IMAGE_TAG}" "$LOCAL_WORKER"
+docker tag "${REGISTRY}/xiaoxia-saas-web:${IMAGE_TAG}" "$LOCAL_WEB"
+
+echo "Rollback images pulled."
+
+# 停止当前容器（回滚不跑 migration，避免数据问题）
+docker rm -f xiaoxia-api-staging 2>/dev/null || true
+docker rm -f xiaoxia-worker-staging 2>/dev/null || true
+docker rm -f xiaoxia-web-staging 2>/dev/null || true
+
+LOG_OPTS="--log-driver json-file --log-opt max-size=50m --log-opt max-file=3"
+
+# 启动 API（回滚不跑 migration）
+echo "Starting API (rollback)..."
+docker run -d \
+  --name xiaoxia-api-staging \
+  --env-file "$ENV_FILE" \
+  --network xiaoxia-net-staging \
+  -p 127.0.0.1:8000:8000 \
+  -e APP_ENV=staging \
+  -e APP_VERSION="$IMAGE_TAG" \
+  -e GENERATED_FILES_DIR=/app/generated \
+  -e GENERATED_FILES_URL_PREFIX=/generated-files \
+  -e PUBLIC_API_BASE_URL=https://staging-api.xiaoxiajianji.com \
+  -v "$GENERATED_DIR:/app/generated" \
+  --restart unless-stopped \
+  --health-cmd "python -c \"import urllib.request; urllib.request.urlopen('http://localhost:8000/health', timeout=5)\"" \
+  --health-interval 30s \
+  --health-timeout 10s \
+  --health-retries 3 \
+  --health-start-period 40s \
+  $LOG_OPTS \
+  "$LOCAL_API"
+
+# 启动 Worker
+echo "Starting Worker (rollback)..."
+docker run -d \
+  --name xiaoxia-worker-staging \
+  --env-file "$ENV_FILE" \
+  --network xiaoxia-net-staging \
+  -e APP_ENV=staging \
+  -e APP_VERSION="$IMAGE_TAG" \
+  -e WORKER_CONCURRENCY=1 \
+  -e WORKER_MAX_TASKS_PER_CHILD=100 \
+  -e GENERATED_FILES_DIR=/app/generated \
+  -e GENERATED_FILES_URL_PREFIX=/generated-files \
+  -e PUBLIC_API_BASE_URL=https://staging-api.xiaoxiajianji.com \
+  -v "$GENERATED_DIR:/app/generated" \
+  --restart unless-stopped \
+  --health-cmd "sh -c \"grep -q celery /proc/1/cmdline || exit 1\"" \
+  --health-interval 30s \
+  --health-timeout 10s \
+  --health-retries 3 \
+  --health-start-period 30s \
+  $LOG_OPTS \
+  "$LOCAL_WORKER"
+
+# 启动 Web
+LEGACY_VOLUME=""
+if [ -d "$LEGACY_ASSETS_DIR" ] && [ "$(ls -A "$LEGACY_ASSETS_DIR" 2>/dev/null)" ]; then
+  LEGACY_VOLUME="-v ${LEGACY_ASSETS_DIR}:/usr/share/nginx/html/assets-legacy/assets:ro"
+fi
+
+echo "Starting Web (rollback)..."
+docker run -d \
+  --name xiaoxia-web-staging \
+  --network xiaoxia-net-staging \
+  -p 127.0.0.1:3001:80 \
+  --restart unless-stopped \
+  $LEGACY_VOLUME \
+  --health-cmd "wget --spider -q http://127.0.0.1:80" \
+  --health-interval 30s \
+  --health-timeout 5s \
+  --health-retries 3 \
+  $LOG_OPTS \
+  "$LOCAL_WEB"
+
+# 等待 API 健康
+echo "Waiting for API (rollback)..."
+i=0
+while [ "$i" -lt 40 ]; do
+  if curl -sf --max-time 5 http://127.0.0.1:8000/health >/dev/null 2>&1; then
+    echo "API healthy (rollback)."
+    break
+  fi
+  i=$((i + 1))
+  sleep 3
+done
+
+# 等待 Web 健康
+echo "Waiting for Web (rollback)..."
+i=0
+while [ "$i" -lt 15 ]; do
+  if curl -sf --max-time 5 http://127.0.0.1:3001/ >/dev/null 2>&1; then
+    echo "Web healthy (rollback)."
+    break
+  fi
+  i=$((i + 1))
+  sleep 2
+done
+
+echo "=== Rollback complete: $IMAGE_TAG ==="
+docker ps --format "table {{.Names}}\t{{.Status}}\t{{.Image}}" | grep staging
+ROLLBACK_EOF
+)
+
+    # 将脚本 base64 编码后通过 SSH 执行
+    local script_b64
+    script_b64=$(echo "$rollback_script" | base64 -w 0)
+
+    log_info "在 staging 服务器上执行回滚脚本..."
+    if run_ssh "echo '$script_b64' | base64 -d | sh -s -- '$rollback_tag' '${REGISTRY_TOKEN:-}'" 2>&1; then
+        log_info "✅ 回滚命令执行完成"
         return 0
     else
-        log_error "部分服务回滚失败"
+        log_error "❌ 回滚命令执行失败"
         return 1
     fi
 }
 
 # ===========================================
-# 5. 发送通知
+# 4. 发送通知
 # ===========================================
 send_notification() {
     local status="$1"   # success / failure / rollback
@@ -250,12 +394,12 @@ send_notification() {
         return 0
     fi
 
-    # 调用通知脚本
     if [ -f "$SCRIPT_DIR/deploy_notify.py" ]; then
         python3 "$SCRIPT_DIR/deploy_notify.py" \
             --status "$status" \
             --detail "$detail" \
             --webhook "$webhook" \
+            --env staging \
             2>/dev/null || log_warn "通知发送失败（非致命）"
     else
         log_warn "找不到 deploy_notify.py，跳过通知"
@@ -267,41 +411,42 @@ send_notification() {
 # ===========================================
 main() {
     echo ""
-    echo "=========================================="
-    echo "  CI Staging 健康检查 + 自动回滚"
-    echo "=========================================="
+    echo "==========================================="
+    echo "  CI Staging 健康检查 + 自动回滚（SSH模式）"
+    echo "==========================================="
     echo ""
 
     local deploy_status="success"
     local deploy_detail=""
 
-    # 1. 记录部署前状态（回滚目标）
-    save_rollback_target
+    # 1. 设置 SSH
+    if ! setup_ssh; then
+        log_error "SSH 配置失败，无法执行回滚"
+    fi
 
-    # 2. 等待 Watchtower 更新
-    wait_for_watchtower
+    # 2. 记录部署前状态（回滚目标）
+    save_rollback_target || true
 
-    # 3. 健康检查
+    # 3. 健康检查（公网视角）
     if ! health_check; then
         log_error "健康检查失败"
         deploy_status="failure"
-        deploy_detail="健康检查超时，部署后服务未正常启动"
+        deploy_detail="公网健康检查超时，部署后服务未正常响应"
 
         # 自动回滚
         if [ "${SKIP_ROLLBACK:-false}" != "true" ]; then
             log_warn "开始自动回滚..."
             if do_rollback; then
                 deploy_status="rollback"
-                deploy_detail="健康检查失败，已自动回滚到上一版本"
+                deploy_detail="健康检查失败，已自动回滚到上一版本 (${ROLLBACK_TAG:-未知})"
 
-                # 回滚后再检查一下回滚后的状态
-                log_info "等待 Watchtower 应用回滚（90s）..."
-                sleep 90
+                # 回滚后再检查一下公网状态
+                log_info "回滚完成，重新检查公网健康状态..."
                 if health_check; then
                     log_info "✅ 回滚后服务已恢复"
                     deploy_detail="${deploy_detail}，回滚后服务已恢复"
                 else
-                    log_error "⚠️ 回滚后健康检查仍未通过，请手动排查"
+                    log_error "⚠️  回滚后健康检查仍未通过，请手动排查"
                     deploy_detail="${deploy_detail}，但回滚后仍未恢复，请紧急排查"
                 fi
             else
@@ -321,7 +466,7 @@ main() {
     log_info "  ✅ Staging 部署成功！"
     log_info "=================================="
 
-    deploy_detail="部署成功，所有健康检查通过"
+    deploy_detail="部署成功，所有健康检查通过 (${GITHUB_SHA:-未知版本})"
     send_notification "success" "$deploy_detail"
 }