In the AI lab n8n is locally deployed on Kubernetes and consists of several core components that together provide workflow execution, scaling, persistence, queue processing, and integrations with AI services such as Milvus and PostgreSQL. Here is a detailed overview of the n8n components deployed on Kubernetes:
N8N Main Application Pods
These are the primary n8n application containers.
Responsibilities:
- Web UI
- REST API
- Workflow Editing
- Authentication
- Webhook Management
- Workflow Orchestration
- AI Agent Execution
- Integration with External APIs and LLMs
N8N Worker Pods (Queue Mode)
In this deployment, n8n worker pods are configured in Queue Mode. This separates:
- Workflow execution from
- Frontend/UI/API handling
Worker pods execute:
- AI Workflows
- Long Running Jobs
- Document Ingestion
- Embedding Generation
- Vector DB Operations
- External API Calls
Redis Queue Layer
Redis is deployed as the message broker for n8n queue mode.
Responsibilities:
- Job Queues
- Workflow Scheduling
- Distributed Execution Coordination
- Retry Management
PostgreSQL Database
PostgreSQL is used by n8n for persistent storage.
Stores:
- Workflows
- Credentials
- Execution History
- Chat Memory
- AI Agent Session State
- Users & Permissions
- Metadata
PostgreSQL also serves as:
- Lightweight conversational memory.
- Structured metadata store
Milvus Vector Database
Milvus handles:
- Embedding Storage
- Semantic Search
- Similarity Retrieval
- Vector Indexing
N8N workflows interact with Milvus for:
- RAG Retrieval
- Knowledge Base Search
- AI Memory Retrieval
- Semantic Document Lookup
Object Storage Layer
Many AI workflows require object storage for:
- PDFs
- Embeddings
- OCR Files
- Uploaded Documents
- Workflow Artifacts
Traefik API Gateway/Ingress Controller
Kubernetes ingress exposes:
- N8N UI
- Webhook Endpoints
- APIs
Responsibilities:
- TLS Termination
- Routing
- Rate Limiting
- Authentication Integration
Observability Stack
Monitoring:
- Prometheus
- Grafana
Logging:
- Loki
- ELK Stack
Used For:
- Workflow Monitoring
- GPU Utilization
- Vector DB Latency
- AI Inference Metrics