当前位置：首页 > news >正文

DeepInsight部署实战：Docker与Kubernetes容器化方案终极指南

news 2026/6/30 17:38:11

DeepInsight部署实战：Docker与Kubernetes容器化方案终极指南

【免费下载链接】deepInsightThe deep-research enables efficient RAG retrieval and multi-source data analysis, supporting intelligent reasoning for automated complex research tasks.项目地址: https://gitcode.com/openeuler/deepInsight

前往项目官网免费下载：https://ar.openeuler.org/ar/

DeepInsight是面向企业的深度研究智能体，通过多Agent协同、上下文工程和异构知识检索技术，构建高效的研究能力。本文将为您提供完整的DeepInsight容器化部署指南，涵盖Docker单机部署与Kubernetes集群部署两种方案，帮助您快速搭建开箱即用的深度研究平台。💡

为什么选择容器化部署？

容器化部署为DeepInsight带来了多重优势：

环境一致性：消除"在我机器上能运行"的问题
快速部署：一键启动完整的DeepInsight服务
资源隔离：确保研究任务互不干扰
弹性伸缩：轻松应对不同规模的研究需求

准备工作与环境配置

系统要求

Docker 20.10+ 或 Kubernetes 1.24+
至少8GB内存（建议16GB+）
50GB可用磁盘空间
支持CUDA的GPU（可选，用于加速）

获取项目源码

git clone https://gitcode.com/openeuler/deepInsight cd deepInsight

Docker单机部署方案

1. 基础Docker部署

DeepInsight项目已提供完整的Dockerfile，支持一键构建：

# 构建DeepInsight镜像 docker build -t deepinsight:latest . # 运行容器 docker run -d \ --name deepinsight \ -p 8888:8888 \ -v $(pwd)/data:/deepinsight/data \ -e DEEPSEEK_API_KEY=your_api_key \ deepinsight:latest

2. Docker Compose完整部署

创建docker-compose.yml文件：

version: '3.8' services: deepinsight: build: . container_name: deepinsight ports: - "8888:8888" volumes: - ./data:/deepinsight/data - ./config.yaml:/deepinsight/config.yaml environment: - DEEPSEEK_API_KEY=${DEEPSEEK_API_KEY} - HF_ENDPOINT=https://hf-mirror.com restart: unless-stopped healthcheck: test: ["CMD", "curl", "-f", "http://localhost:8888/api/v1/health"] interval: 30s timeout: 10s retries: 3 # 可选：MinerU解析服务（离线版本） mineru: image: mineru-api:latest container_name: mineru-api ports: - "8000:8000" volumes: - ./mineru_data:/data environment: - MINERU_API_ENABLE_FASTAPI_DOCS=0 restart: unless-stopped

启动服务：

# 复制环境变量模板 cp .env.example .env # 编辑.env文件，配置API密钥 nano .env # 启动所有服务 docker-compose up -d

3. 配置优化

编辑配置文件config.yaml，优化容器化部署：

app: name: deepinsight host: 0.0.0.0 # 容器内监听所有接口 port: 8888 api_prefix: /api/v1 reload: false # 生产环境关闭热重载 database: url: sqlite:///data/deepinsight.db # 使用持久化存储 workspace: work_root: /deepinsight/data # 容器内路径 image_path_mode: base_url image_base_url: http://${HOST_IP}:8888/api/v1/deepinsight/charts/image rag: work_root: /deepinsight/data engine: type: lightrag

Kubernetes集群部署方案

1. 创建命名空间和配置

# namespace.yaml apiVersion: v1 kind: Namespace metadata: name: deepinsight

# configmap.yaml apiVersion: v1 kind: ConfigMap metadata: name: deepinsight-config namespace: deepinsight data: config.yaml: | app: name: deepinsight host: 0.0.0.0 port: 8888 database: url: sqlite:///data/deepinsight.db workspace: work_root: /deepinsight/data

2. 部署DeepInsight服务

# deployment.yaml apiVersion: apps/v1 kind: Deployment metadata: name: deepinsight namespace: deepinsight spec: replicas: 2 selector: matchLabels: app: deepinsight template: metadata: labels: app: deepinsight spec: containers: - name: deepinsight image: deepinsight:latest ports: - containerPort: 8888 env: - name: DEEPSEEK_API_KEY valueFrom: secretKeyRef: name: api-secrets key: deepseek-key volumeMounts: - name:># service.yaml apiVersion: v1 kind: Service metadata: name: deepinsight-service namespace: deepinsight spec: selector: app: deepinsight ports: - port: 8888 targetPort: 8888 type: LoadBalancer # ingress.yaml（如果需要域名访问） apiVersion: networking.k8s.io/v1 kind: Ingress metadata: name: deepinsight-ingress namespace: deepinsight spec: rules: - host: deepinsight.your-domain.com http: paths: - path: / pathType: Prefix backend: service: name: deepinsight-service port: number: 8888

高级部署配置

1. GPU加速支持

如果您的集群支持GPU，可以添加GPU资源：

# 在deployment.yaml的container部分添加 resources: limits: nvidia.com/gpu: 1 # 申请1个GPU

2. 水平自动扩缩容（HPA）

apiVersion: autoscaling/v2 kind: HorizontalPodAutoscaler metadata: name: deepinsight-hpa namespace: deepinsight spec: scaleTargetRef: apiVersion: apps/v1 kind: Deployment name: deepinsight minReplicas: 2 maxReplicas: 10 metrics: - type: Resource resource: name: cpu target: type: Utilization averageUtilization: 70 - type: Resource resource: name: memory target: type: Utilization averageUtilization: 80

3. 监控与日志

集成Prometheus监控：

# 添加annotations到deployment metadata: annotations: prometheus.io/scrape: "true" prometheus.io/port: "8888" prometheus.io/path: "/metrics"

部署验证与测试

1. 服务健康检查

# 检查服务状态 curl http://localhost:8888/api/v1/health # 测试API接口 curl -X POST http://localhost:8888/api/v1/deepinsight/research \ -H "Content-Type: application/json" \ -d '{"topic": "人工智能发展趋势"}'

2. 功能验证

使用DeepInsight CLI进行功能测试：

# 进入容器 docker exec -it deepinsight bash # 测试会议洞察功能 di conf gen --name "HOTOS 2025" --files-src ./example/papers/hotos-2025/ # 测试深度研究功能 di resch gen --topic "大语言模型技术演进"