Skynet-config/guida_ai_server_supermicro.md

Supermicro X9DRW-iF AI Coding Server

Guida Completa Setup Produzione CPU-Only

Hardware

Server

• Supermicro X9DRW-iF
• 2x Intel Xeon E5-2680 v2
• 256GB ECC DDR3

Storage

• SSD SATA 128GB
• HDD WD Red 2TB

Obiettivo

Server AI locale per:

• coding assistant
• repository analysis
• RAG
• AI agent
• inferenza locale CPU-only
• accesso web LAN

---

Architettura Finale

SSD 128GB

Sistema operativo + runtime AI

Contiene:

• Debian
• Ollama
• Open WebUI
• cache runtime
• modello attivo

---

HDD 2TB

Storage AI

Contiene:

• modelli GGUF
• repository Git
• vector database
• dataset
• backup

---

1. Installazione Debian

ISO consigliata

Debian 12 minimal netinstall

Sito: https://www.debian.org/download

---

Durante installazione

Selezionare:

• SSH Server
• Standard system utilities

NON selezionare:

• Desktop Environment
• GNOME
• KDE
• print server
• web server

---

2. Configurazione dischi

SSD 128GB

Partizionamento

Mount	Size
/	100GB
swap	8-16GB

Filesystem:

• ext4

---

HDD 2TB

Filesystem:

• ext4

Mount point:

/mnt/ai-data

---

3. Primo avvio

Aggiornare sistema:

sudo apt update && sudo apt upgrade -y

Installare utility:

sudo apt install -y \
git curl wget htop btop tmux nano \
build-essential cmake python3 python3-pip \
nvme-cli unzip

---

4. Configurazione HDD

Identificare UUID

sudo blkid

Creare mountpoint

sudo mkdir -p /mnt/ai-data

Modificare fstab

sudo nano /etc/fstab

Aggiungere:

UUID=INSERISCI_UUID /mnt/ai-data ext4 defaults,noatime 0 2

Test mount

sudo mount -a

Verifica:

df -h

---

5. Ottimizzazioni CPU

Installare cpupower

sudo apt install linux-cpupower -y

Impostare governor performance

sudo cpupower frequency-set -g performance

Verifica:

cpupower frequency-info

---

6. Installazione Ollama

curl -fsSL https://ollama.com/install.sh | sh

Verifica:

ollama --version

---

7. Spostare storage modelli su HDD

sudo mkdir -p /mnt/ai-data/ollama
sudo systemctl stop ollama
sudo mv /usr/share/ollama/* /mnt/ai-data/ollama/
sudo rm -rf /usr/share/ollama
sudo ln -s /mnt/ai-data/ollama /usr/share/ollama
sudo systemctl start ollama

---

8. Ottimizzazione Ollama

sudo mkdir -p /etc/systemd/system/ollama.service.d
sudo nano /etc/systemd/system/ollama.service.d/override.conf

Contenuto:

[Service]
Environment="OLLAMA_NUM_PARALLEL=1"
Environment="OLLAMA_MAX_LOADED_MODELS=1"
Environment="OLLAMA_KEEP_ALIVE=24h"
Environment="OLLAMA_HOST=0.0.0.0:11434"

Reload:

sudo systemctl daemon-reload
sudo systemctl restart ollama

---

9. Installazione Docker

curl -fsSL https://get.docker.com | sh
sudo usermod -aG docker $USER

Logout/login richiesto.

---

10. Installazione Open WebUI

docker run -d \
-p 3000:8080 \
-v open-webui:/app/backend/data \
--name open-webui \
--restart always \
ghcr.io/open-webui/open-webui:main

---

11. Accesso WebUI

http://IP_SERVER:3000

---

12. Installazione modelli

Qwen2.5 Coder 7B

ollama pull qwen2.5-coder:7b

Qwen2.5 Coder 14B

ollama pull qwen2.5-coder:14b

DeepSeek Coder

ollama pull deepseek-coder:6.7b

---

13. Prestazioni attese

Modello	Velocità
7B Q4	6-15 tok/s
14B	3-7 tok/s
32B	1-3 tok/s

---

14. Integrazione VSCode

VSCode: https://code.visualstudio.com/

Continue: https://continue.dev/

Config esempio:

name: Smart Local Router
version: 1.0.0
schema: v1

models:
  # 🧠 CODING MODEL (DEFAULT)
  - name: DeepSeek Coder
    provider: ollama
    model: deepseek-coder-v2:16b
    apiBase: http://192.168.1.250:11434
    roles:
      - chat
      - edit
      - apply

  # 🤖 AGENT MODEL
  - name: Qwen Agent
    provider: ollama
    model: qwen2.5-coder:14b
    apiBase: http://192.168.1.250:11434
    roles:
      - chat
      - edit
      - apply

context:
  - provider: code
  - provider: diff
  - provider: terminal
  - provider: problems
  - provider: folder
  - provider: repo-map
  - provider: os

---

15. Installazione Aider

pip install aider-chat

Uso:

cd progetto
aider

Esempi:

• Refactor this class
• Add JWT authentication
• Write unit tests

---

16. Struttura consigliata

/mnt/ai-data/
├── models/
├── repos/
├── embeddings/
├── vector-db/
├── backups/
└── datasets/

---

17. Monitoring

htop
btop

Temperature:

sudo apt install lm-sensors
sudo sensors-detect
sensors

---

18. Firewall

sudo apt install ufw
sudo ufw allow ssh
sudo ufw allow 3000/tcp
sudo ufw allow 11434/tcp
sudo ufw enable

---

19. Accesso SSH

ssh utente@IP_SERVER

---

20. Upgrade futuri

1. SSD più grande
2. RTX 3090
3. NVMe PCIe adapter
4. Più storage

---

21. Workflow consigliato

1. VSCode + Continue
2. Ollama locale
3. Aider per refactor
4. Open WebUI per chat
5. Repository indicizzati

---

22. Comandi utili

Lista modelli:

ollama list

Eseguire modello:

ollama run qwen2.5-coder:7b

Eliminare modello:

ollama rm nome-modello

Stato servizio:

systemctl status ollama

---

Fine

Configurazione attuale:

Skynet.lbs.farm

Utente

user lbsadmin
pwd  <solita>

Modello usato

ollama pull deepseek-coder-v2:16b
ollama pull qwen2.5-coder:14b