Self Hosted Llm Macminim4

Brief View

1. LM Studio
2. Open WebUI
3. CloudFlare tunnel

• [!IMPORTANT]

All setup based on Mac Mini M4

LM Studio

• download from official website
• enable local server mode, default on localhost:1234

1

lms server start

Open WebUI

• [!NOTE]

not support python3.13 at the time of writing

1
2
3
4

python3 -m venv ~/.venv/openwebui
source ~/.venv/openwebui/bin/activate
pip install open-webui
nohup open-webui sevrer > open-webui.log 2>&1 &

• Go to lcocalhost:8080, register admin account
• In Settings > Admin settings > Connections > Manage OpenAI API Connections, set URL to http://127.0.0.1:1234/v1 and key to none.
• Verify connections to see if LM Studio models are available.

CloudFlare Tunnel

• prerequisites: cloudflare account, domain name

• I used namesilo before and switched to porkbun this time.

1
2
3
4
5

- If not installed, install by `brew install cloudflared`

```bash
cloudflared login

• Go to prompted URL to finish login.

1

cloudflared tunnel create <tunnel_name>

• It will generate a json file in ~/.cloudflared/ directory.

Config yaml

• Put in ~/.cloudflared/config.yml

1
2
3
4
5
6
7
8

tunnel: <tunnel_name>
credentials-file: /Users/<user>/.cloudflared/<LONG-UUID>.json

ingress:
  - hostname: <hostname>
    service: http://localhost: <open_webui_port>
  # Default catch-all rule
  - service: http_status:404

• open_webui_port is default to 8080

• run in background

1

nohup cloudflared tunnel run <tunnel_name> > cloudflare_tunnel.log 2>&1 &