# Stage 1: Install dependencies (dev deps needed for esbuild) FROM node:20-alpine AS deps RUN corepack enable && corepack prepare pnpm@10.33.2 --activate WORKDIR /app COPY package.json pnpm-lock.yaml ./ RUN pnpm install --frozen-lockfile --prod=false # Stage 2: Build the worker bundle FROM node:20-alpine AS builder RUN corepack enable && corepack prepare pnpm@10.33.2 --activate WORKDIR /app COPY --from=deps /app/node_modules ./node_modules COPY . . ENV SKIP_ENV_VALIDATION=1 RUN pnpm build:worker # Stage 3: Production runner (prod deps only). # # Critical ordering: create the worker user FIRST and chown the workdir # BEFORE pnpm install, so node_modules + lazy-cache directories # (tesseract.js, sharp) are owned by the worker user. Without this, the # previous layout had pnpm install run as root → node_modules root-owned # → tesseract.js / sharp wrote to node_modules/.cache and EACCES'd at # first PDF parse in prod (auditor-K §39). FROM node:20-alpine AS runner RUN corepack enable && corepack prepare pnpm@10.33.2 --activate RUN addgroup --system --gid 1001 nodejs && adduser --system --uid 1001 worker WORKDIR /app RUN chown -R worker:nodejs /app USER worker COPY --chown=worker:nodejs package.json pnpm-lock.yaml ./ RUN pnpm install --frozen-lockfile --prod COPY --from=builder --chown=worker:nodejs /app/dist/worker.js ./worker.js # Healthcheck — pings Redis from inside the worker container. Without # this, a worker whose Redis connection has silently dropped (BullMQ # rejects new jobs but the Node process is alive) is invisible to # compose / swarm and jobs queue indefinitely (auditor-K §40). HEALTHCHECK --interval=30s --timeout=5s --start-period=20s --retries=3 \ CMD node -e "const Redis=require('ioredis');const r=new Redis(process.env.REDIS_URL,{maxRetriesPerRequest:1,connectTimeout:3000,lazyConnect:true});r.connect().then(()=>r.ping()).then(()=>{r.disconnect();process.exit(0)}).catch(()=>process.exit(1))" || exit 1 CMD ["node", "worker.js"]