local-ai-environment

Docker + n8n Plan

Planning document for Phase 2: deploying n8n locally using Docker.

Status: Planning — do not execute yet.
This document records the intended architecture and prerequisites for Phase 2.
Implementation begins only after Phase 1 validation passes.

What is n8n?

n8n (pronounced “n-eight-n”) is an open-source workflow automation platform — self-hosted, like Zapier or Make but running entirely on your machine. It connects APIs, processes data, reacts to webhooks, and can trigger AI inference via HTTP calls.

Key properties:

Why n8n + Local AI?

Benefit Detail
No API costs Calls your local Ollama instead of OpenAI API
Privacy Data never leaves your machine
Automation Trigger AI processing from files, webhooks, schedules
Chaining Chain multiple LLM calls in a single workflow
Integration Connect AI output to files, databases, email, Slack, etc.

Example use cases:

Prerequisites Before Starting Phase 2

Complete this checklist before running 08-Deploy-N8N.ps1:

Verify Docker is Ready

# Check Docker is installed
docker --version

# Check Docker engine is running
docker ps

# Check available disk
docker system df

# Check host.docker.internal resolves (needed for Ollama access from container)
docker run --rm --add-host=host.docker.internal:host-gateway alpine nslookup host.docker.internal

Save this as docker-compose.n8n.yml in the project root. Script 08-Deploy-N8N.ps1 will create this automatically.

version: '3.8'

services:
  n8n:
    image: n8nio/n8n:latest
    container_name: n8n-local
    ports:
      - "5678:5678"
    environment:
      - N8N_HOST=localhost
      - N8N_PORT=5678
      - N8N_PROTOCOL=http
      - WEBHOOK_URL=http://localhost:5678/
      - GENERIC_TIMEZONE=Europe/Warsaw
      - N8N_LOG_LEVEL=info
      # Disable telemetry
      - N8N_DIAGNOSTICS_ENABLED=false
      - N8N_VERSION_NOTIFICATIONS_ENABLED=false
    volumes:
      - n8n_data:/home/node/.n8n
    restart: unless-stopped
    extra_hosts:
      - "host.docker.internal:host-gateway"

volumes:
  n8n_data:
    driver: local

Key Configuration Explained

Setting Value Reason
ports: 5678:5678 localhost:5678 only No external exposure
extra_hosts host-gateway Allows container to call Ollama on the host
n8n_data volume Persistent Workflows survive container restarts
N8N_DIAGNOSTICS_ENABLED=false Disabled Privacy — no telemetry
GENERIC_TIMEZONE Europe/Warsaw Correct cron scheduling

Network Architecture

Windows 11 Host
  │
  ├─ Ollama service
  │    └─ listens on  127.0.0.1:11434
  │
  ├─ Open-WebUI (Python)
  │    └─ listens on  127.0.0.1:3000
  │
  └─ Docker Desktop
       │
       └─ n8n container
            │
            ├─ accessible from host at   localhost:5678
            │
            ├─ calls Ollama via          http://host.docker.internal:11434
            │   (host.docker.internal resolves to Windows host IP from inside container)
            │
            └─ calls Open-WebUI via      http://host.docker.internal:3000

Important: From inside the n8n container, localhost refers to the container itself, not the Windows host. Always use host.docker.internal to reach Ollama and Open-WebUI.

Calling Ollama from n8n

HTTP Request Node — Generate Endpoint

Body (JSON):

{
  "model": "qwen2.5:7b",
  "prompt": "",
  "stream": false
}

Response field: response contains the generated text.

HTTP Request Node — Chat Endpoint (OpenAI-compatible)

Body (JSON):

{
  "model": "qwen2.5:7b",
  "messages": [
    { "role": "user", "content": "" }
  ],
  "stream": false
}

Response field: message.content

Calling Open-WebUI API from n8n

Open-WebUI exposes an OpenAI-compatible API.

Headers:

Content-Type: application/json
Authorization: Bearer <your-open-webui-api-key>

Get an API key: Open-WebUI → Settings → Account → API Keys.

Body (JSON):

{
  "model": "qwen2.5:7b",
  "messages": [
    { "role": "user", "content": "" }
  ]
}

Response field: choices[0].message.content

Data Persistence

n8n stores all state in the Docker volume n8n_data:

Volume location on host (Docker-managed):

docker volume inspect n8n_data
# Shows actual path under Docker's data directory

Backup

# Create backup archive
docker run --rm `
  -v n8n_data:/data `
  -v "${PWD}:/backup" `
  alpine `
  tar czf /backup/n8n-backup-$(Get-Date -Format "yyyyMMdd").tar.gz /data

# Restore from backup
docker run --rm `
  -v n8n_data:/data `
  -v "${PWD}:/backup" `
  alpine `
  tar xzf /backup/n8n-backup-20260906.tar.gz -C /

Security Notes

Future Script: 08-Deploy-N8N.ps1

This script will be created in Phase 2. Planned steps:

  1. Import modules (Logging, Validation, Configuration)
  2. Check Docker is running (Test-DockerRunning)
  3. Check port 5678 is free (Test-PortListening 5678 should return false)
  4. Check host.docker.internal resolves inside Docker
  5. Create docker-compose.n8n.yml from embedded template
  6. Run docker compose -f docker-compose.n8n.yml up -d
  7. Wait up to 60 seconds for n8n health check (/healthz endpoint)
  8. Verify n8n can reach Ollama: 
    docker exec n8n-local wget -qO- http://host.docker.internal:11434
  9. Print access URL and first-run instructions
  10. Save reports/n8n-deploy-report.json and .md

Parameters will include -DryRun, -NonInteractive, -N8nPort (default 5678), -Timezone.

Estimated Resource Usage

Component RAM CPU (idle) Disk
n8n container ~400MB ~0% ~300MB image
n8n volume data — — ~50MB (grows with execution history)
Ollama (model loaded) 4-10GB ~0% 2-10GB per model
Open-WebUI ~200MB ~0% ~500MB
Total (Phase 2) ~5-12GB ~0% ~3-11GB