Part 7 of 8

n8n + Ollama — AI-Powered Automation

Zero API costs, infinite automation possibilities.

40 minutes

4GB+ recommended

Prerequisites

Completed Part 1 (Ollama), Docker, basic automation concepts

Time to Complete

35–45 minutes

Recommended Plan

4GB ($20/mo) for light automation. 8GB ($40/mo) for production workflows

Looking for a quick-start guide? Check out our standalone n8n Deployment Guide for a streamlined setup walkthrough.

Introduction

n8n is a self-hosted workflow automation platform — think Zapier, but private and free. Combined with Ollama, you get AI-powered automation with zero per-execution API costs. Email triage, content generation, data extraction, sentiment analysis — all on your $40/month VPS.

💰 Cost comparison: Zapier AI Workflows: $50+/month with per-task limits. n8n + Ollama on your VPS: $0/execution — unlimited workflows, unlimited AI calls.

Deploying n8n

mkdir -p ~/ai-stack/n8n && cd ~/ai-stack/n8n

docker-compose.yml

version: "3.8"

services:
  n8n:
    image: n8nio/n8n:latest
    container_name: n8n
    restart: unless-stopped
    ports:
      - "5678:5678"
    environment:
      - N8N_BASIC_AUTH_ACTIVE=true
      - N8N_BASIC_AUTH_USER=admin
      - N8N_BASIC_AUTH_PASSWORD=your-secure-password
      - N8N_HOST=0.0.0.0
      - WEBHOOK_URL=http://your-server-ip:5678/
      - GENERIC_TIMEZONE=UTC
    volumes:
      - n8n-data:/home/node/.n8n
    extra_hosts:
      - "host.docker.internal:host-gateway"

volumes:
  n8n-data:

docker compose up -d

Access n8n at http://your-server-ip:5678.

n8n + Ollama Connection

Configure n8n's AI nodes to use your local Ollama:

In n8n, create a new workflow
Add an "Ollama Chat Model" node
Set the Base URL to http://host.docker.internal:11434
Select your model (e.g., mistral)
Configure temperature (0.3 for structured tasks, 0.7 for creative)
Test the connection — you should see a successful response

Workflow 1: Intelligent Email Triage

Build a workflow that classifies and routes incoming emails:

Workflow Logic

IMAP Trigger (new email)
    │
    ▼
Ollama Chat Model
  System: "Classify this email as: URGENT, ACTION_REQUIRED, 
           INFORMATIONAL, or SPAM. Also suggest a brief response."
  User: "Subject: {{subject}}

Body: {{body}}"
    │
    ▼
Switch Node (route by classification)
    ├── URGENT → Slack notification + Draft reply
    ├── ACTION_REQUIRED → Create task in Todoist/Notion
    ├── INFORMATIONAL → Archive + Summary log
    └── SPAM → Move to trash

Each node connects visually in n8n's canvas editor. The Ollama node processes the email content and returns structured classification.

Workflow 2: Content Generation Pipeline

Pipeline Flow

Webhook Trigger (topic input)
    │
    ▼
Ollama #1: Generate blog outline
    │
    ▼
Ollama #2: Write draft from outline
    │
    ▼
Ollama #3: Review and improve draft
    │
    ▼
Save to Google Docs / Markdown file
    │
    ▼
Send notification (email/Slack)

Chain multiple Ollama calls for iterative refinement. Each AI step receives the output of the previous step, progressively improving the content.

Workflow 3: Document Processing

Auto-summarize new documents

Watch Folder (new files)
    │
    ▼
Read File → Extract Text
    │
    ▼
Ollama: "Summarize this document in 3 bullet points.
         Extract any action items or deadlines."
    │
    ▼
Store summary in database
    │
    ▼
(Optional) Feed to RAG pipeline (Part 3)
    │
    ▼
Send notification with summary

Connect to the RAG pipeline from Part 3 for automatic document ingestion into your knowledge base.

Advanced Patterns

Error Handling

Add retry logic on AI nodes (3 retries with exponential backoff). Ollama may timeout on long generations — set generous timeout values (120s+).

Resource Contention

When multiple workflows hit Ollama simultaneously, queue them. Use n8n's concurrency settings to limit parallel AI executions to 1–2 at a time.

Human Approval Gates

Add "Wait for Approval" nodes before destructive actions. Send a preview via email/Slack and wait for human confirmation before proceeding.

Scheduling

Use cron triggers for recurring AI tasks: daily report generation, weekly competitor analysis, monthly performance summaries.

What's Next?

Every workflow you've built runs at zero marginal cost — no per-API-call billing, no usage caps. In Part 8: Production Hardening, we lock everything down with a reverse proxy, centralized auth, automated backups, and resource tuning.

← Part 6: CrewAI Part 8: Production Hardening