feat(docling): persist model cache via Docker volume

- Add docling_models volume mounted at /models in container
- Set HF_HOME=/models/huggingface at runtime (via docker-compose env)
- Models download once → persist in volume → survive container rebuilds
- Build-time preload uses || to not block build if network unavailable

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

docker-compose.yml

@@ -256,8 +256,12 @@ services:
       dockerfile: Dockerfile
     container_name: iconsulting-docling
     restart: unless-stopped
+    environment:
+      HF_HOME: /models/huggingface
     ports:
       - "3007:3007"
+    volumes:
+      - docling_models:/models
     healthcheck:
       test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:3007/health')"]
       interval: 30s
@@ -478,3 +482,5 @@ volumes:
     driver: local
   minio_data:
     driver: local
+  docling_models:
+    driver: local

packages/services/docling-service/Dockerfile

@@ -13,8 +13,10 @@ RUN pip install --no-cache-dir -r requirements.txt
 COPY app/ ./app/
 COPY scripts/ ./scripts/
 
-# 构建时预下载模型（实际执行一次 PDF 转换触发 HuggingFace 模型下载）
+# 模型缓存目录（运行时通过 Docker volume 持久化到 /models）
-RUN python scripts/preload_models.py
+# 构建时预下载到镜像默认路径；运行时 HF_HOME=/models/huggingface 由 volume 接管
+ENV HF_HOME=/root/.cache/huggingface
+RUN python scripts/preload_models.py || echo "Model preload skipped (will download on first request)"
 
 EXPOSE 3007