Deploy langgenius/dify, an LLM-based chatbot app on Kubernetes with Helm chart.
helm repo add dify https://borispolonsky.github.io/dify-helm helm repo update helm install my-release dify/dify
For customized installation, please refer to the README.md file.
The following diagram illustrates the complete network architecture and service topology of the Dify Helm deployment:
The Nginx proxy handles traffic routing with the following rules:
/console/api → API Service (5001) /api → API Service (5001) /v1 → API Service (5001) /files → API Service (5001) /mcp → API Service (5001) /e/ → Plugin Daemon (5002) /explore → Web Service (3000) /marketplace → External Marketplace API /triggers → API Service (5001) / → Web Service (3000) [Default Route]
| Component | Image | Port | Role |
|---|---|---|---|
| API | langgenius/dify-api:1.10.1-fix.1 | 5001 | RESTful API server, business logic processing |
| Web | langgenius/dify-web:1.10.1-fix.1 | 3000 | Web UI frontend |
| Worker | langgenius/dify-api:1.10.1-fix.1 | - | Background task processing (Celery) |
| Beat | langgenius/dify-api:1.10.1-fix.1 | - | Periodic task scheduler (Celery Beat) |
| Sandbox | langgenius/dify-sandbox:0.2.12 | 8194 | Secure code execution environment |
| Plugin Daemon | langgenius/dify-plugin-daemon:0.4.1-local | 5002, 5003 | Plugin management and execution |
| SSRF Proxy | ubuntu/squid:latest | 3128 | External request security proxy |
| Nginx Proxy | nginx:latest | 80 | Reverse proxy, load balancing |