Local AI Stack

A self-hosted AI stack running locally with Docker Compose.

This repository contains everything needed to run:

• Ollama (local LLM inference)
• Open WebUI (chat interface)
• SearXNG (web search)
• ComfyUI (image generation workflows)
• n8n (automation and agents)

---

Architecture

┌──────────────┐
│ Open WebUI   │
└──────┬───────┘
       │
       ├──► Ollama
       │
       ├──► SearXNG
       │
       ├──► ComfyUI
       │
       └──► n8n

---

Services

Service	URL	Purpose
Open WebUI	http://localhost:3000	Chat interface
Ollama	http://localhost:11434	Local LLM backend
SearXNG	http://localhost:8080	Private web search
ComfyUI	http://localhost:8188	Image generation
n8n	http://localhost:5678	Automations & agents

---

Requirements

Hardware

Recommended:

• NVIDIA GPU
• 16 GB+ VRAM
• 64 GB+ RAM

Current development machine:

• AMD Ryzen 7 8700G
• NVIDIA RTX 5070 eGPU
• 96 GB RAM

---

Installation

Clone repository:

git clone https://github.com/davcastroruiz/local-ai-stack.git
cd local-ai-stack

Start stack:

docker compose up -d

Verify:

docker ps

---

Updating

Update all containers:

docker compose pull
docker compose down
docker compose up -d

Update a single service:

docker compose pull open-webui
docker compose up -d open-webui

Examples:

docker compose pull ollama
docker compose up -d ollama

docker compose pull searxng
docker compose up -d searxng

docker compose pull comfyui
docker compose up -d comfyui

docker compose pull n8n
docker compose up -d n8n

---

Watchtower (Optional)

Automatic container updates.

Example service:

watchtower:
  image: containrrr/watchtower
  container_name: watchtower
  restart: unless-stopped
  volumes:
    - /var/run/docker.sock:/var/run/docker.sock
  command: --cleanup --interval 86400

Start:

docker compose up -d watchtower

Logs:

docker logs -f watchtower

Force update check:

docker exec watchtower /watchtower --run-once

---

SearXNG Configuration

Open WebUI search integration requires JSON output.

Enter container:

docker exec -it searxng sh

Edit:

/etc/searxng/settings.yml

Required configuration:

search:
  formats:
    - html
    - json

Restart:

docker restart searxng

Validation:

curl "http://localhost:8080/search?q=test&format=json"

Expected result:

• JSON response
• Not HTML

---

Logs

docker logs -f ollama
docker logs -f open-webui
docker logs -f searxng
docker logs -f comfyui
docker logs -f n8n

---

Restart Services

docker restart ollama
docker restart open-webui
docker restart searxng
docker restart comfyui
docker restart n8n

---

Backups

Important data:

Open WebUI

webui volume

Ollama

ollama volume

n8n

./n8n_data

Compose Configuration

docker compose config > compose-resolved.yml

---

Migration to a New Machine

1. Install Docker Desktop
2. Copy repository
3. Restore volumes
4. Restore n8n_data
5. Run:

docker compose up -d

6. Verify services

---

Troubleshooting

Open WebUI cannot search the web

Verify:

curl "http://localhost:8080/search?q=test&format=json"

If HTML is returned:

search:
  formats:
    - html
    - json

Restart SearXNG.

---

Verify running containers

docker ps

Expected:

ollama
open-webui
searxng
comfyui
n8n

---

Future Improvements

• HTTPS reverse proxy
• Authentication hardening
• Automatic backups
• Monitoring
• GPU metrics
• Multi-user support
• Agent workflows through n8n
• ComfyUI API integration with Open WebUI

---

License

MIT License