Run the complete PROMPT stack and CI verification in Docker

.dockerignore

@@ -0,0 +1,12 @@
+.git
+.next
+node_modules
+dist
+coverage
+.env
+.env.*
+!.env.example
+*.tsbuildinfo
+.DS_Store
+.agents
+.codex

.env.example

@@ -1,8 +1,11 @@
-DATABASE_URL=postgresql://prompt:replace-me@127.0.0.1:18132/prompt
-POSTGRES_PASSWORD=replace-me
-PG_PORT=18132
-S3_PORT=18133
-S3_ENDPOINT=http://127.0.0.1:18133
-S3_ACCESS_KEY=replace-me
+# Replace all three placeholders before starting Docker Compose.
+POSTGRES_PASSWORD=replace-with-a-long-random-password
+S3_ACCESS_KEY=replace-with-a-unique-access-key
 S3_SECRET_KEY=replace-with-a-long-random-secret
 S3_BUCKET=prompt
+WEB_BIND_ADDRESS=0.0.0.0
+WEB_PORT=18131
+PG_PORT=18132
+S3_PORT=18133
+PROMPT_IMAGE=prompt:local
+RELEASE_SHA=development

.gitea/workflows/ci.yml

@@ -1,4 +1,4 @@
-name: PROMPT CI and deployment
+name: PROMPT Docker CI and deployment
 
 on:
   push:
@@ -9,42 +9,31 @@ on:
 jobs:
   verify-and-deploy:
     runs-on: ces-control
-    timeout-minutes: 35
+    timeout-minutes: 45
+    env:
+      PROMPT_IMAGE: prompt:${{ gitea.sha }}
+      PROMPT_TEST_IMAGE: prompt-tests:${{ gitea.sha }}
+      RELEASE_SHA: ${{ gitea.sha }}
     steps:
       - name: Checkout
         uses: actions/checkout@v4
-      - name: Set up Node.js
-        uses: actions/setup-node@v4
-        with:
-          node-version: 24
-          cache: npm
-      - name: Install dependencies
-        run: npm ci
-      - name: Type check
-        run: npm run typecheck
-      - name: Unit tests
-        run: npm test
-      - name: Production build
-        run: npm run build
-      - name: Test standalone production server
-        run: npm run test:smoke
-      - name: Prepare Docker and FFmpeg for isolated integration tests
+      - name: Prepare Docker Engine and Compose
         run: |
           set -eu
           if [ "$(id -u)" -eq 0 ]; then SUDO=""; else SUDO="sudo -n"; fi
           if ! command -v docker >/dev/null || ! docker compose version >/dev/null 2>&1; then
             $SUDO apt-get update -qq
             $SUDO apt-get install -y docker.io docker-compose-v2
-            $SUDO systemctl enable --now docker
           fi
-          if ! command -v ffmpeg >/dev/null; then
-            $SUDO apt-get update -qq
-            $SUDO apt-get install -y ffmpeg
-          fi
-      - name: Test real PostgreSQL, S3 and video processing in isolation
+          $SUDO systemctl enable --now docker
+      - name: Validate the Compose configuration
+        run: docker compose --env-file .env.example config --quiet
+      - name: Type check, unit tests, production build and smoke test inside Docker
+        run: docker build --target integration -t "$PROMPT_TEST_IMAGE" .
+      - name: Build non-root production image with bundled FFmpeg
+        run: docker build --target runtime -t "$PROMPT_IMAGE" .
+      - name: Test the complete isolated Docker stack
         run: bash scripts/integration.sh
-      - name: Deploy persistent systemd service
+      - name: Deploy the persistent Docker Compose stack
         if: gitea.ref == 'refs/heads/main' && (gitea.event_name == 'push' || gitea.event_name == 'workflow_dispatch')
-        env:
-          RELEASE_SHA: ${{ gitea.sha }}
-        run: bash scripts/deploy.sh "$RELEASE_SHA" "$(command -v node)"
+        run: bash scripts/deploy.sh "$RELEASE_SHA"

Dockerfile

@@ -0,0 +1,29 @@
+FROM node:24-bookworm-slim AS base
+ENV NEXT_TELEMETRY_DISABLED=1
+WORKDIR /app
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends ffmpeg ca-certificates \
+    && rm -rf /var/lib/apt/lists/*
+
+FROM base AS dependencies
+COPY package.json package-lock.json ./
+RUN npm ci
+
+FROM dependencies AS build
+COPY . .
+RUN npm run typecheck && npm test && npm run build && npm run test:smoke
+
+FROM build AS integration
+USER node
+CMD ["npm", "run", "test:integration"]
+
+FROM base AS runtime
+ENV NODE_ENV=production HOSTNAME=0.0.0.0 PORT=3000 \
+    FFMPEG_PATH=/usr/bin/ffmpeg FFPROBE_PATH=/usr/bin/ffprobe
+COPY --from=build --chown=node:node /app/.next/standalone/ ./
+COPY --from=build --chown=node:node /app/dist/worker.cjs /app/dist/bootstrap.cjs ./
+COPY --from=build --chown=node:node /app/db/ ./db/
+COPY --chown=node:node scripts/container-health.mjs ./
+USER node
+EXPOSE 3000
+CMD ["node", "server.js"]

README.md

@@ -4,22 +4,31 @@
 
 自己的素材与创作工作台：Next.js Pages Router、React、TypeScript、Tailwind CSS、Node.js 24。中文界面包含瀑布流、日期贡献热力图、素材导入、多媒体详情、分类管理、独立素材架与线条图形工具。没有创建作品的按钮；图文和动态作品由代码实现。
 
-## 启动
+## 全 Docker 启动
+
+本机只需 Docker Engine / Docker Desktop 和 Compose v2。Node、npm、FFmpeg、迁移程序、网站与后台处理都在镜像里。
 
 ```bash
-nvm use
-npm ci
 cp .env.example .env
-# 修改 .env 中的三个密码 / 密钥，不使用示例值。
-docker compose --env-file .env up -d
-set -a; . ./.env; set +a
-npm run storage:prepare
-npm run dev
-# 另一终端加载同一份 .env 后运行后台视频处理：
-npm run worker
+# 修改 .env 中的三个密码 / 密钥，不能使用示例值。
+docker build --target runtime -t prompt:local .
+docker compose --env-file .env up -d --wait postgres objects
+docker compose --env-file .env run --rm migrate
+docker compose --env-file .env up -d --wait web worker
 ```
 
-本机需要 Docker Compose 和 FFmpeg / ffprobe。开发中的 .env、数据库文件和对象不进入 Git；生产凭据由部署脚本随机生成。
+访问 `http://localhost:18131`。图片与其他文件保存在 S3 对象卷，文字、分类、素材包引用和任务保存在 PostgreSQL 卷；停止、重建应用容器不会删除这些数据。
+
+| 容器服务 | 职责 |
+| --- | --- |
+| `web` | Next.js 页面、API、素材上传/读取；宿主机 18131 → 容器 3000 |
+| `worker` | FFmpeg 音画分离、封面与音轨处理；没有 HTTP 端口 |
+| `postgres` | PostgreSQL 元数据和文字；兼容维护用回环端口 18132 |
+| `objects` | RustFS 私有 S3 对象存储；兼容维护用回环端口 18133，控制台关闭 |
+| `migrate` | 一次性数据库迁移与对象桶初始化，执行后退出 |
+| `tests` | CI 的一次性集成测试执行器，执行后退出 |
+
+四个常驻服务使用 `restart: unless-stopped`，Web 和 worker 以非 root 用户运行。worker 通过心跳检查存活；Web 健康检查同时验证当前版本、数据库和 S3。
 
 ## 页面与数据
 
@@ -35,7 +44,7 @@ npm run worker
 
 素材库不插入演示或占位帖子。空分类显示简洁的空状态；空素材包不进入任何瀑布流。分页游标由时间与 UUID 组成，重复时间戳下仍稳定。
 
-详细模型见 `docs/data-model.md`。PostgreSQL 保存引用、分类、元数据、文字和任务；RustFS 通过 S3 协议保存二进制对象。数据库和 S3 端口只监听宿主机 `127.0.0.1`，浏览器通过应用访问媒体，支持 Range 分段读取。
+详细模型见 `docs/data-model.md`。应用容器通过 Docker 网络的 `postgres:5432` 和 `objects:9000` 访问存储，不使用宿主机回环地址。浏览器通过应用访问媒体，支持 Range 分段读取。
 
 视频原件保持不变，后台生成无声 MP4、封面，以及存在声音时的 M4A 音轨。音轨记录在 `asset_variants`，始终属于原视频；不会生成独立声音资产。后台单任务处理，使用数据库锁领取任务，失败重试三次，支持超时任务恢复。FFmpeg 禁止输入网络协议。
 
@@ -53,48 +62,48 @@ npm run worker
 
 `contribution_events` 保存细分的 `kind`、对象类型、对象 ID、数量、发生时间、元数据和可选幂等键。显示时按上海时区汇总所有类型。当前没有将上传、分类或保存动作自动算作贡献；触发规则留待确定。没有随机或伪造热力图数据。
 
-## CI/CD 与运维
+## CI/CD 与生产部署
 
-`.gitea/workflows/ci.yml` 使用 `runs-on: ces-control`。PR 执行验证；main 上的 push 或手动运行，在类型检查、单元测试、生产构建、启动检查及真实 PostgreSQL/S3/视频集成测试通过后才部署。
+`.gitea/workflows/ci.yml` 使用 `runs-on: ces-control`。依赖安装、类型检查、单元测试、生产构建、启动检查都在 Docker 构建阶段运行。集成测试启动隔离的完整容器栈，并由测试容器请求真实的非 root Web / worker 容器，验证 PostgreSQL、RustFS、上传、共享引用和音画分离。
 
-集成测试启动带独立 project 名称和随机本机端口的 Docker 服务，使用临时随机凭据；结束后只删除该测试项目的容器和卷，不触碰生产卷。
+main 上的 push 或手动运行，在全部检查通过后才发布。测试环境使用独立 Compose 项目、随机端口和随机凭据；测试结束只删除它自己的容器和卷。
 
-生产约定：
+生产使用固定 Compose project `prompt`，复用原有 `prompt_postgres_data` 和 `prompt_objects` 卷，以及 `/etc/prompt.env` 的凭据。由旧部署升级时，发布脚本会停用 `prompt.service`、`prompt-worker.service`，由容器接管运行；宿主机无需项目专用的 Node/FFmpeg 进程。旧发布目录留作首次迁移失败时的回退依据。
 
 | 项目 | 值 |
 | --- | --- |
-| Web | `prompt.service`，HTTP `18131` |
-| 视频处理 | `prompt-worker.service` |
-| PostgreSQL | Docker Compose `prompt-postgres`，回环端口 `18132` |
-| S3 / RustFS | Docker Compose `prompt-objects`，回环端口 `18133`，控制台关闭 |
-| 发布 | `/opt/prompt/releases/<commit>`，`/opt/prompt/current` 原子软链接 |
-| 部署清单 | `/opt/prompt/infra/compose.yml` |
-| 密钥 | `/etc/prompt.env`，root:prompt / 0640，不进代码或日志 |
-| 运行数据 | `/var/lib/prompt`；数据库与 S3 使用 Docker named volumes |
-| 发布前数据库备份 | `/var/lib/prompt/backups`，保留最近 10 份 |
+| 生产运行 | Docker Compose：web、worker、postgres、objects |
+| 网站端口 | `18131`，映射到 Web 容器 3000 |
+| 应用镜像 | `prompt:<commit>`；回退使用上一版镜像 |
+| 发布清单 | `/opt/prompt/releases/<commit>/compose.yml` 和 `release.env` |
+| 当前发布 | `/opt/prompt/current` 原子软链接 |
+| 密钥 | `/etc/prompt.env`，保留既有凭据，新建时随机生成；不进代码和日志 |
+| 持久数据 | Docker named volumes，不随应用发布删除 |
+| 数据库备份 | `/var/lib/prompt/backups`，每次发布前备份，保留最近 10 份 |
 
-服务由 systemd 管理，支持开机启动、故障重启和失败回退。迁移采用追加方式；发布前创建 PostgreSQL dump，回退应用版本不会破坏已写入数据。S3 对象卷不随发布清理；对象备份需要另外快照该卷，数据库 dump 不能代替对象备份。
+发布使用宿主机文件锁串行执行，自动跳过已落后的提交。先备份、迁移，再切换 Web / worker；新版本失败时恢复上一版容器，首次迁移失败时可恢复原先的宿主机服务。生产发布从不执行 `down -v`。对象卷需要独立快照，数据库 dump 不能代替对象备份。
+
+在服务器上维护当前版本：
 
 ```bash
-curl -fsS http://127.0.0.1:18131/api/health # 当前应用版本
-curl -fsS http://127.0.0.1:18131/api/ready  # 数据库和 S3 连接
-sudo systemctl status prompt.service prompt-worker.service
-sudo journalctl -u prompt-worker.service -n 100 --no-pager
-sudo docker compose -p prompt --env-file /etc/prompt.env -f /opt/prompt/infra/compose.yml ps
+cd /opt/prompt/current
+sudo docker compose -p prompt --env-file /etc/prompt.env --env-file release.env ps
+sudo docker compose -p prompt --env-file /etc/prompt.env --env-file release.env logs --tail 100 web worker
+sudo docker compose -p prompt --env-file /etc/prompt.env --env-file release.env exec web node container-health.mjs
 ```
 
-首次部署可能安装 Docker Compose 和 FFmpeg；runner 需能访问宿主机 systemd、Docker，并具备 root 或免密 sudo。服务不配置公网域名或防火墙，按现有内网访问方式使用。
+`GET /api/health` 返回应用版本；`GET /api/ready` 检查数据库与对象存储。
 
-## 验证命令
+## 内网与公网访问
+
+`172.31.212.211` 是服务器内网 IP，不能作为公网入口。Web 端口默认映射为 `0.0.0.0:18131:3000`；公网访问仍需要实例绑定公网 IP / EIP 或反向代理，并配置相应安全组和防火墙。Docker 化不会使内网 IP 自动变成公网 IP。本项目不会擅自修改云安全组或开放数据库/S3 的公网端口。
+
+## 独立运行 CI 验证
 
 ```bash
-npm run typecheck
-npm test
-npm run build
-npm run test:smoke
-# 在已配置的隔离 PostgreSQL / S3 环境：
-npm run storage:prepare
-PROMPT_INTEGRATION_TEST=1 npm run test:integration
-# 或在 Docker 主机上一键创建并清理隔离环境：
-bash scripts/integration.sh
+docker build --target integration -t prompt:tests .
+docker build --target runtime -t prompt:local .
+PROMPT_IMAGE=prompt:local PROMPT_TEST_IMAGE=prompt:tests bash scripts/integration.sh
 ```
+
+整个验证过程无需在宿主机安装 Node.js、PostgreSQL、S3 或 FFmpeg。runner 仅需 Docker Engine、Compose、Git、Bash、flock 和管理 Docker/旧服务所需的权限。

compose.yml

@@ -1,4 +1,68 @@
+x-app-environment: &app-environment
+  HOSTNAME: 0.0.0.0
+  PORT: "3000"
+  DATABASE_URL: postgresql://prompt:${POSTGRES_PASSWORD:?Set a unique PostgreSQL password}@postgres:5432/prompt
+  S3_ENDPOINT: http://objects:9000
+  S3_ACCESS_KEY: ${S3_ACCESS_KEY:?Set a unique S3 access key}
+  S3_SECRET_KEY: ${S3_SECRET_KEY:?Set a unique S3 secret key}
+  S3_BUCKET: ${S3_BUCKET:-prompt}
+  RELEASE_SHA: ${RELEASE_SHA:-development}
+  FFMPEG_PATH: /usr/bin/ffmpeg
+  FFPROBE_PATH: /usr/bin/ffprobe
+
+x-app: &app
+  image: ${PROMPT_IMAGE:-prompt:local}
+  restart: unless-stopped
+  environment: *app-environment
+  init: true
+  security_opt:
+    - no-new-privileges:true
+  cap_drop: [ALL]
+  depends_on:
+    postgres:
+      condition: service_healthy
+    objects:
+      condition: service_started
+
 services:
+  web:
+    <<: *app
+    ports:
+      - "${WEB_BIND_ADDRESS:-0.0.0.0}:${WEB_PORT:-18131}:3000"
+    healthcheck:
+      test: ["CMD", "node", "container-health.mjs"]
+      interval: 5s
+      timeout: 8s
+      retries: 24
+      start_period: 10s
+  worker:
+    <<: *app
+    command: ["node", "worker.cjs"]
+    stop_grace_period: 11m
+    healthcheck:
+      test: ["CMD", "node", "-e", "const f=require('fs');try{if(Date.now()-f.statSync('/tmp/prompt-worker-health').mtimeMs>60000)process.exit(1)}catch{process.exit(1)}"]
+      interval: 15s
+      timeout: 5s
+      retries: 4
+      start_period: 10s
+  migrate:
+    <<: *app
+    profiles: [tools]
+    restart: "no"
+    command: ["node", "bootstrap.cjs"]
+  tests:
+    image: ${PROMPT_TEST_IMAGE:-prompt:tests}
+    profiles: [tools]
+    init: true
+    environment:
+      <<: *app-environment
+      PROMPT_INTEGRATION_TEST: "1"
+      PROMPT_TEST_BASE_URL: http://web:3000
+    depends_on:
+      web:
+        condition: service_healthy
+      worker:
+        condition: service_healthy
   postgres:
     image: postgres:17-alpine
     restart: unless-stopped

package.json

@@ -7,7 +7,7 @@
   },
   "scripts": {
     "dev": "next dev",
-    "build": "next build && esbuild src/worker.ts --bundle --platform=node --format=cjs --external:pg-native --outfile=dist/worker.cjs && node scripts/prepare-standalone.mjs",
+    "build": "next build && esbuild src/worker.ts --bundle --platform=node --format=cjs --external:pg-native --outfile=dist/worker.cjs && esbuild scripts/bootstrap.ts --bundle --platform=node --format=cjs --external:pg-native --outfile=dist/bootstrap.cjs && node scripts/prepare-standalone.mjs",
     "start": "next start",
     "typecheck": "next typegen && tsc --noEmit",
     "test": "tsx --test tests/health.test.ts tests/model.test.ts",

scripts/container-health.mjs

@@ -0,0 +1,9 @@
+try {
+  const port = process.env.PORT || '3000';
+  const [health, ready] = await Promise.all([
+    fetch(`http://127.0.0.1:${port}/api/health`, { signal: AbortSignal.timeout(5000) }),
+    fetch(`http://127.0.0.1:${port}/api/ready`, { signal: AbortSignal.timeout(5000) }),
+  ]);
+  const data = await health.json();
+  if (!health.ok || !ready.ok || data.service !== 'prompt' || data.release !== (process.env.RELEASE_SHA || 'development')) process.exit(1);
+} catch { process.exit(1); }

scripts/deploy.sh

@@ -1,171 +1,97 @@
 #!/usr/bin/env bash
 set -Eeuo pipefail
-
-# Run on the actual ces-control host, with root or passwordless sudo.
-if [ "$(id -u)" -ne 0 ]; then
-  exec sudo -n bash "$0" "$@"
-fi
-
+if [ "$(id -u)" -ne 0 ]; then exec sudo -n bash "$0" "$@"; fi
 RELEASE_SHA="${1:?Release SHA is required}"
-NODE_SOURCE="${2:?Node executable path is required}"
 [[ "$RELEASE_SHA" =~ ^[0-9a-f]{40}$ ]] || { echo 'Invalid release SHA'; exit 1; }
-test -x "$NODE_SOURCE"
-test -d /run/systemd/system || { echo 'Deployment requires the host systemd environment'; exit 1; }
-systemctl show-environment >/dev/null
-test -f .next/standalone/server.js
-
+IMAGE="prompt:$RELEASE_SHA"
 APP_ROOT=/opt/prompt
-DATA_ROOT=/var/lib/prompt
-SERVICE=prompt.service
-SERVICE_USER=prompt
-PORT=18131
 RELEASE_DIR="$APP_ROOT/releases/$RELEASE_SHA"
-
+PORT=18131
+command -v docker >/dev/null
+docker compose version >/dev/null
+docker image inspect "$IMAGE" >/dev/null
 mkdir -p "$APP_ROOT/releases"
 exec 9>"$APP_ROOT/.deploy.lock"
 flock -w 300 9 || { echo 'Another PROMPT deployment is running'; exit 1; }
-
-# Do not replace the release of a newer main commit after a slower old run.
-if [ "$(git -c safe.directory="$PWD" rev-parse HEAD)" != "$RELEASE_SHA" ]; then
-  echo 'Checkout does not match release SHA'; exit 1
-fi
+test "$(git -c safe.directory="$PWD" rev-parse HEAD)" = "$RELEASE_SHA"
 LATEST_MAIN=$(git -c safe.directory="$PWD" ls-remote origin refs/heads/main | cut -f1)
-if [ "$LATEST_MAIN" != "$RELEASE_SHA" ]; then
-  echo 'Skipping deployment: main has a newer commit'
-  exit 0
-fi
-
+if [ "$LATEST_MAIN" != "$RELEASE_SHA" ]; then echo 'Skipping deployment: main has a newer commit'; exit 0; fi
 PREVIOUS=""
-if [ -L "$APP_ROOT/current" ] && [ -d "$APP_ROOT/current" ]; then
-  PREVIOUS=$(readlink -f "$APP_ROOT/current")
+if [ -L "$APP_ROOT/current" ] && [ -d "$APP_ROOT/current" ]; then PREVIOUS=$(readlink -f "$APP_ROOT/current"); fi
+bash scripts/infra.sh "$IMAGE"
+mkdir -p "$RELEASE_DIR"
+install -m 0644 compose.yml "$RELEASE_DIR/compose.yml"
+printf 'PROMPT_IMAGE=%s\nRELEASE_SHA=%s\nWEB_PORT=%s\n' "$IMAGE" "$RELEASE_SHA" "$PORT" > "$RELEASE_DIR/release.env"
+compose_at(){ local release="$1"; shift; env -u PROMPT_IMAGE -u PROMPT_TEST_IMAGE -u RELEASE_SHA -u WEB_PORT -u WEB_BIND_ADDRESS docker compose -p prompt --env-file /etc/prompt.env --env-file "$release/release.env" -f "$release/compose.yml" "$@"; }
+compose(){ compose_at "$RELEASE_DIR" "$@"; }
+LEGACY_WEB=0
+LEGACY_WORKER=0
+LEGACY_WEB_ENABLED=0
+LEGACY_WORKER_ENABLED=0
+if command -v systemctl >/dev/null; then
+  if systemctl is-active --quiet prompt.service; then LEGACY_WEB=1; fi
+  if systemctl is-active --quiet prompt-worker.service; then LEGACY_WORKER=1; fi
+  if systemctl is-enabled --quiet prompt.service 2>/dev/null; then LEGACY_WEB_ENABLED=1; fi
+  if systemctl is-enabled --quiet prompt-worker.service 2>/dev/null; then LEGACY_WORKER_ENABLED=1; fi
 fi
-if [ -z "$PREVIOUS" ]; then
-  "$NODE_SOURCE" -e "const s=require('net').createServer();s.once('error',()=>process.exit(1));s.listen($PORT,'0.0.0.0',()=>s.close());" || {
-    echo "Port $PORT is already occupied"; exit 1;
-  }
-fi
-
-if ! id "$SERVICE_USER" >/dev/null 2>&1; then
-  useradd --system --home "$DATA_ROOT" --shell /usr/sbin/nologin "$SERVICE_USER"
-fi
-install -d -o "$SERVICE_USER" -g "$SERVICE_USER" -m 0750 "$DATA_ROOT"
-bash scripts/infra.sh "$NODE_SOURCE" "$RELEASE_SHA"
-set -a
-. /etc/prompt.env
-set +a
-"$NODE_SOURCE" node_modules/tsx/dist/cli.mjs scripts/bootstrap.ts
-# Keep runtime data outside the checkout and releases; future modules can use it.
-if [ ! -d "$RELEASE_DIR" ]; then
-  STAGING=$(mktemp -d "$APP_ROOT/releases/.staging.XXXXXX")
-  cp -a .next/standalone/. "$STAGING/"
-  mkdir -p "$STAGING/.next" "$STAGING/bin"
-  cp -a .next/static "$STAGING/.next/static"
-  if [ -d public ]; then cp -a public "$STAGING/public"; fi
-  install -m 0755 "$NODE_SOURCE" "$STAGING/bin/node"
-  install -m 0644 dist/worker.cjs "$STAGING/worker.cjs"
-  printf 'RELEASE_SHA=%s\n' "$RELEASE_SHA" > "$STAGING/release.env"
-  chmod -R a+rX "$STAGING"
-  chmod 0755 "$STAGING"
-  mv "$STAGING" "$RELEASE_DIR"
-fi
-
-cat > /etc/systemd/system/prompt.service <<EOF
-[Unit]
-Description=PROMPT
-After=network-online.target
-Wants=network-online.target
-
-[Service]
-Type=simple
-User=$SERVICE_USER
-Group=$SERVICE_USER
-WorkingDirectory=$APP_ROOT/current
-Environment=NODE_ENV=production
-Environment=HOSTNAME=0.0.0.0
-Environment=PORT=$PORT
-Environment=PROMPT_DATA_DIR=$DATA_ROOT
-EnvironmentFile=$APP_ROOT/current/release.env
-EnvironmentFile=/etc/prompt.env
-ExecStart=$APP_ROOT/current/bin/node server.js
-Restart=always
-RestartSec=5
-NoNewPrivileges=true
-UMask=0027
-
-[Install]
-WantedBy=multi-user.target
-EOF
-
-cat > /etc/systemd/system/prompt-worker.service <<EOF
-[Unit]
-Description=PROMPT media processing
-After=network-online.target docker.service
-
-[Service]
-Type=simple
-User=$SERVICE_USER
-Group=$SERVICE_USER
-WorkingDirectory=$APP_ROOT/current
-Environment=NODE_ENV=production
-EnvironmentFile=/etc/prompt.env
-ExecStart=$APP_ROOT/current/bin/node worker.cjs
-Restart=always
-RestartSec=5
-TimeoutStopSec=650
-NoNewPrivileges=true
-UMask=0027
-
-[Install]
-WantedBy=multi-user.target
-EOF
-
-SWITCHED=0
-rollback() {
+CUTOVER=0
+rollback(){
   local code=$?
   trap - ERR
   set +e
-  if [ "$SWITCHED" -eq 1 ]; then
+  if [ "$CUTOVER" -eq 1 ]; then
+    if [ -n "$PREVIOUS" ] && [ -f "$PREVIOUS/compose.yml" ]; then
+      compose_at "$PREVIOUS" up -d --no-build web worker
+      echo 'Restored the previous Docker application image'
+    else
+      compose stop web worker
+      compose rm -f web worker
+      if [ "$LEGACY_WORKER_ENABLED" -eq 1 ]; then systemctl enable prompt-worker.service; fi
+      if [ "$LEGACY_WEB_ENABLED" -eq 1 ]; then systemctl enable prompt.service; fi
+      if [ "$LEGACY_WEB" -eq 1 ]; then systemctl start prompt.service; fi
+      if [ "$LEGACY_WORKER" -eq 1 ]; then systemctl start prompt-worker.service; fi
+      echo 'Restored legacy application services if previously running'
+    fi
     if [ -n "$PREVIOUS" ]; then
       ln -sfn "$PREVIOUS" "$APP_ROOT/.current.rollback"
       mv -Tf "$APP_ROOT/.current.rollback" "$APP_ROOT/current"
-      systemctl restart "$SERVICE"
-      if [ -f "$PREVIOUS/worker.cjs" ]; then systemctl restart prompt-worker.service; else systemctl disable --now prompt-worker.service; fi
-      echo 'Restored previous release'
-    else
-      systemctl disable --now "$SERVICE"
-      systemctl disable --now prompt-worker.service
-      rm -f "$APP_ROOT/current"
-      echo 'Stopped failed first deployment'
     fi
   fi
-  echo "Deployment failed (exit $code)"
+  echo "Docker deployment failed (exit $code); persistent data volumes were retained"
   exit "$code"
 }
 trap rollback ERR
+# Keep the same prompt_postgres_data and prompt_objects volumes. No production down -v.
+compose up -d --no-build --wait --wait-timeout 120 postgres objects
+install -d -o root -g root -m 0700 /var/lib/prompt/backups
+compose exec -T postgres pg_dump -U prompt -d prompt -Fc > "/var/lib/prompt/backups/$(date -u +%Y%m%dT%H%M%S)-$RELEASE_SHA.dump"
+find /var/lib/prompt/backups -name '*.dump' -printf '%T@ %p\n' | sort -rn | tail -n +11 | cut -d' ' -f2- | xargs -r rm --
+compose run --rm --no-deps migrate
+CUTOVER=1
+# One-time migration: finish a current host worker job before switching the web port.
+if [ "$LEGACY_WORKER" -eq 1 ] || [ "$LEGACY_WORKER_ENABLED" -eq 1 ]; then systemctl disable --now prompt-worker.service; fi
+if [ "$LEGACY_WEB" -eq 1 ] || [ "$LEGACY_WEB_ENABLED" -eq 1 ]; then systemctl disable --now prompt.service; fi
+compose up -d --no-build --wait --wait-timeout 180 web worker
+compose exec -T web node container-health.mjs
+compose exec -T web node -e "fetch('http://127.0.0.1:3000/').then(r=>{if(!r.ok)process.exit(1)}).catch(()=>process.exit(1))"
+compose exec -T worker sh -c 'test "$(id -u)" -ne 0 && ffmpeg -version >/dev/null'
 ln -sfn "$RELEASE_DIR" "$APP_ROOT/.current.next"
 mv -Tf "$APP_ROOT/.current.next" "$APP_ROOT/current"
-SWITCHED=1
-systemctl daemon-reload
-systemctl enable "$SERVICE"
-systemctl restart "$SERVICE"
-systemctl enable --now prompt-worker.service
-systemctl restart prompt-worker.service
-
-HEALTHY=0
-for attempt in $(seq 1 30); do
-  if "$NODE_SOURCE" -e "Promise.all([fetch('http://127.0.0.1:$PORT/api/health'),fetch('http://127.0.0.1:$PORT/api/ready')]).then(async ([r,s])=>{const d=await r.json();if(!r.ok||!s.ok||d.service!=='prompt'||d.release!=='$RELEASE_SHA')process.exit(1)}).catch(()=>process.exit(1))"; then
-    HEALTHY=1
-    break
-  fi
-  sleep 2
+compose ps
+for service in web worker postgres objects; do
+  id=$(compose ps -q "$service")
+  docker inspect --format '{{.Name}} running={{.State.Running}} restart={{.HostConfig.RestartPolicy.Name}}' "$id"
 done
-if [ "$HEALTHY" -ne 1 ]; then
-  echo 'Health check failed; rolling back'
-  false
+if command -v systemctl >/dev/null; then
+  if systemctl is-active --quiet prompt.service || systemctl is-active --quiet prompt-worker.service; then echo 'Legacy host service is unexpectedly running'; false; fi
 fi
-systemctl is-enabled "$SERVICE"
-systemctl is-active "$SERVICE"
-systemctl is-active prompt-worker.service
-echo "Persistent deployment verified: service=$SERVICE port=$PORT release=$RELEASE_SHA"
+echo "Full Docker deployment verified: web+worker+postgres+objects port=$PORT release=$RELEASE_SHA"
+# Read-only network diagnostics: a private address alone is not a public entrypoint.
 echo "Host: $(hostname)"
 hostname -I || true
+if command -v curl >/dev/null; then
+  for field in eipv4 public-ipv4; do
+    address=$(curl --noproxy '*' -fsS --connect-timeout 2 --max-time 4 "http://100.100.100.200/latest/meta-data/$field" 2>/dev/null || true)
+    if [[ "$address" =~ ^[0-9]+\.[0-9]+\.[0-9]+\.[0-9]+$ ]]; then echo "Instance $field: $address"; fi
+  done
+fi

scripts/infra.sh

@@ -1,34 +1,12 @@
 #!/usr/bin/env bash
 set -Eeuo pipefail
-NODE_SOURCE="${1:?Node executable is required}"
-if ! command -v docker >/dev/null || ! docker compose version >/dev/null 2>&1; then
-  apt-get update -qq
-  DEBIAN_FRONTEND=noninteractive apt-get install -y docker.io docker-compose-v2
-  systemctl enable --now docker
-fi
-if ! command -v ffmpeg >/dev/null; then
-  apt-get update -qq
-  DEBIAN_FRONTEND=noninteractive apt-get install -y ffmpeg
-fi
-umask 027
+IMAGE="${1:?Application image is required}"
+# Preserve existing credentials and Docker project/volume names during migration.
 if [ ! -f /etc/prompt.env ]; then
-  "$NODE_SOURCE" <<'NODE' > /etc/prompt.env
-const {randomBytes}=require('crypto');
-const password=randomBytes(32).toString('hex');
-console.log(`POSTGRES_PASSWORD=${password}\nPG_PORT=18132\nS3_PORT=18133\nDATABASE_URL=postgresql://prompt:${password}@127.0.0.1:18132/prompt\nS3_ENDPOINT=http://127.0.0.1:18133\nS3_ACCESS_KEY=${randomBytes(16).toString('hex')}\nS3_SECRET_KEY=${randomBytes(32).toString('hex')}\nS3_BUCKET=prompt\nFFMPEG_PATH=/usr/bin/ffmpeg\nFFPROBE_PATH=/usr/bin/ffprobe`);
-NODE
+  umask 077
+  TEMP_ENV=$(mktemp /etc/prompt.env.XXXXXX)
+  trap 'rm -f "$TEMP_ENV"' EXIT
+  docker run --rm --entrypoint node "$IMAGE" -e "const {randomBytes}=require('crypto');const hex=()=>randomBytes(32).toString('hex');console.log('POSTGRES_PASSWORD='+hex()+'\nS3_ACCESS_KEY='+hex()+'\nS3_SECRET_KEY='+hex()+'\nS3_BUCKET=prompt')" > "$TEMP_ENV"
+  mv "$TEMP_ENV" /etc/prompt.env
+  trap - EXIT
 fi
-chown root:prompt /etc/prompt.env
-chmod 0640 /etc/prompt.env
-install -d -m 0755 /opt/prompt/infra
-install -m 0644 compose.yml /opt/prompt/infra/compose.yml
-docker compose -p prompt --env-file /etc/prompt.env -f /opt/prompt/infra/compose.yml up -d
-for attempt in $(seq 1 60); do
-  if docker compose -p prompt --env-file /etc/prompt.env -f /opt/prompt/infra/compose.yml exec -T postgres pg_isready -U prompt -d prompt >/dev/null 2>&1; then break; fi
-  if [ "$attempt" -eq 60 ]; then echo 'PostgreSQL startup timed out'; exit 1; fi
-  sleep 2
-done
-# Additive migrations follow this backup. Keep the most recent 10 database dumps.
-install -d -o root -g root -m 0700 /var/lib/prompt/backups
-docker compose -p prompt --env-file /etc/prompt.env -f /opt/prompt/infra/compose.yml exec -T postgres pg_dump -U prompt -d prompt -Fc > "/var/lib/prompt/backups/$(date -u +%Y%m%dT%H%M%S)-${2:-deploy}.dump"
-find /var/lib/prompt/backups -name '*.dump' -printf '%T@ %p\n' | sort -rn | tail -n +11 | cut -d' ' -f2- | xargs -r rm --

scripts/integration.sh

@@ -1,18 +1,20 @@
 #!/usr/bin/env bash
 set -Eeuo pipefail
+: "${PROMPT_IMAGE:?Build the runtime image first}"
+: "${PROMPT_TEST_IMAGE:?Build the integration image first}"
 export COMPOSE_PROJECT_NAME="prompt-ci-${GITEA_RUN_ID:-$$}"
-export POSTGRES_PASSWORD S3_ACCESS_KEY S3_SECRET_KEY
-POSTGRES_PASSWORD=$(node -e "console.log(require('crypto').randomBytes(24).toString('hex'))")
-S3_ACCESS_KEY=$(node -e "console.log(require('crypto').randomBytes(12).toString('hex'))")
-S3_SECRET_KEY=$(node -e "console.log(require('crypto').randomBytes(24).toString('hex'))")
-export PG_PORT=0 S3_PORT=0
-cleanup(){ docker compose -f compose.yml down -v --remove-orphans; }
+export PG_PORT=0 S3_PORT=0 WEB_PORT=0 WEB_BIND_ADDRESS=127.0.0.1
+TEST_ENV=$(mktemp)
+chmod 0600 "$TEST_ENV"
+compose(){ docker compose -p "$COMPOSE_PROJECT_NAME" --env-file "$TEST_ENV" -f compose.yml "$@"; }
+cleanup(){ compose down -v --remove-orphans; rm -f "$TEST_ENV"; }
 trap cleanup EXIT
-docker compose -f compose.yml up -d
-PG_PORT=$(docker compose -f compose.yml port postgres 5432 | tail -1 | sed 's/.*://')
-S3_PORT=$(docker compose -f compose.yml port objects 9000 | tail -1 | sed 's/.*://')
-export DATABASE_URL="postgresql://prompt:$POSTGRES_PASSWORD@127.0.0.1:$PG_PORT/prompt"
-export S3_ENDPOINT="http://127.0.0.1:$S3_PORT" S3_BUCKET=prompt
-export PROMPT_INTEGRATION_TEST=1
-npm run storage:prepare
-npm run test:integration
+# Random credentials exist only for this disposable stack; no host Node required.
+docker run --rm --entrypoint node "$PROMPT_IMAGE" -e "const c=require('crypto');const hex=()=>c.randomBytes(24).toString('hex');console.log('POSTGRES_PASSWORD='+hex()+'\nS3_ACCESS_KEY='+hex()+'\nS3_SECRET_KEY='+hex())" > "$TEST_ENV"
+compose up -d --no-build --wait --wait-timeout 120 postgres objects
+compose run --rm --no-deps migrate
+compose up -d --no-build --wait --wait-timeout 150 web worker
+# The test runner calls the actual non-root production web/worker containers.
+compose run --rm --no-deps tests
+compose exec -T web node container-health.mjs
+compose exec -T worker sh -c 'test "$(id -u)" -ne 0 && ffmpeg -version >/dev/null'

src/worker.ts

@@ -1,4 +1,4 @@
-import { mkdtemp,rm } from "node:fs/promises";
+import { mkdtemp,rm,writeFile } from "node:fs/promises";
 import { createWriteStream } from "node:fs";
 import { tmpdir } from "node:os";
 import { join } from "node:path";
@@ -40,5 +40,10 @@ async function processOne(){
  }finally{await rm(dir,{recursive:true,force:true});}
  return true;
 }
-async function main(){while(!stopping){try{if(!await processOne())await setTimeout(2000);}catch{console.error('Worker connection error; retrying');await setTimeout(5000);}}await db().end();}
+async function main(){
+ const heartbeat=()=>writeFile('/tmp/prompt-worker-health',String(Date.now())).catch(()=>console.error('Worker heartbeat failed'));
+ await heartbeat();const timer=setInterval(()=>{void heartbeat();},15000);
+ try{while(!stopping){try{if(!await processOne())await setTimeout(2000);}catch{console.error('Worker connection error; retrying');await setTimeout(5000);}}}
+ finally{clearInterval(timer);await db().end();}
+}
 main().catch(()=>process.exit(1));

tests/integration.test.ts

@@ -15,8 +15,10 @@ import {probe,run} from '../src/lib/probe';
 test('real PostgreSQL + S3 library, sharing, taxonomy, uploads, range and video separation', {timeout:180000}, async()=>{
  assert.equal(process.env.PROMPT_INTEGRATION_TEST,'1','Run only against an explicitly isolated test database');
  const socket=createServer();socket.listen(0,'127.0.0.1');await once(socket,'listening');const address=socket.address();assert.ok(address&&typeof address!=='string');const port=address.port;await new Promise<void>(r=>socket.close(()=>r()));
- const base=`http://127.0.0.1:${port}`;const server=spawn(process.execPath,['.next/standalone/server.js'],{env:{...process.env,PORT:String(port),HOSTNAME:'127.0.0.1'},stdio:'inherit'});
- const worker=spawn(process.execPath,['dist/worker.cjs'],{env:process.env,stdio:'inherit'});const dir=await mkdtemp(join(tmpdir(),'prompt-test-'));
+ const external=process.env.PROMPT_TEST_BASE_URL;
+ const base=external||`http://127.0.0.1:${port}`;
+ const server=external?null:spawn(process.execPath,['.next/standalone/server.js'],{env:{...process.env,PORT:String(port),HOSTNAME:'127.0.0.1'},stdio:'inherit'});
+ const worker=external?null:spawn(process.execPath,['dist/worker.cjs'],{env:process.env,stdio:'inherit'});const dir=await mkdtemp(join(tmpdir(),'prompt-test-'));
  const request=async(path:string,body?:unknown)=>{const res=await fetch(base+path,body?{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify(body)}:undefined);assert.ok(res.ok,`${path}: ${res.status} ${res.ok?'':await res.text()}`);return res.json();};
  try{
   let ready=false;for(let i=0;i<60;i++){try{if((await fetch(base+'/api/health')).ok){ready=true;break;}}catch{}await setTimeout(250);}assert.ok(ready);
@@ -45,7 +47,7 @@ test('real PostgreSQL + S3 library, sharing, taxonomy, uploads, range and video
   assert.equal((await db().query("SELECT count(*)::integer AS n FROM assets WHERE kind='audio'")).rows[0].n,0,'Bound soundtracks are variants, not independent audio assets');
   const audio=join(dir,'audio.wav');await run('ffmpeg',['-y','-v','error','-f','lavfi','-i','sine=frequency=220','-t','0.2',audio]);await upload('audio',[{data:await readFile(audio),name:'audio.wav',mime:'audio/wav'}]);assert.equal((await request('/api/feed?section=audio')).posts.length,1);
  }finally{
-  for(const child of [worker,server])if(child.exitCode===null&&child.signalCode===null){const stopped=once(child,'exit');child.kill('SIGTERM');await stopped;}
+  for(const child of [worker,server])if(child&&child.exitCode===null&&child.signalCode===null){const stopped=once(child,'exit');child.kill('SIGTERM');await stopped;}
   await rm(dir,{recursive:true,force:true});await db().end();
  }
 });