After building dozens of automation workflows, I’ve settled on a reliable stack. Here’s what works.

Core Stack

n8n (Self-Hosted)

The heart of everything. Self-hosted for:

• Data privacy
• No execution limits
• Custom nodes
• Full control

PostgreSQL

For n8n’s internal database and workflow data storage. Reliable and well-understood.

Redis

Queue management and caching for high-volume workflows.

Caddy

Reverse proxy with automatic HTTPS. Simple config, just works.

Essential Integrations

Communication

• Slack: Team notifications, approvals
• Email (SMTP): Customer communication
• Telegram: Personal alerts

Data Sources

• Airtable: Quick databases and forms
• Google Sheets: Collaborative data entry
• PostgreSQL: Production data

AI/ML

• OpenAI: GPT for text processing
• Anthropic: Claude for analysis tasks
• Local LLMs: Sensitive data processing

Business Tools

• Notion: Documentation triggers
• Linear: Issue management
• Stripe: Payment webhooks

Workflow Patterns I Use

1. Event-Driven Processing

Webhook → Validate → Process → Notify → Log

2. Scheduled Batch Jobs

Cron → Fetch Data → Transform → Sync → Report

3. Human-in-the-Loop

Trigger → AI Draft → Human Review → Execute

4. Error Recovery

Main Flow → Try/Catch → Retry Logic → Alert on Failure

Production Best Practices

Version Control

• Export workflows as JSON
• Store in Git repository
• Tag releases

Monitoring

• Execution logging
• Error alerting
• Performance metrics

Security

• Credential encryption
• Webhook authentication
• Network isolation

Sample Architecture

Internet
    ↓
Caddy (SSL)
    ↓
n8n (Docker)
    ↓
PostgreSQL + Redis

Lessons Learned

1. Start simple: Add complexity only when needed
2. Log everything: You’ll thank yourself later
3. Test in staging: Production bugs are expensive
4. Document workflows: Future you needs context

Running a similar stack? I’d love to compare notes on what’s working for you.