Remote AI Access (Mac mini at home)

Goal: use your Mac mini from laptop/phone while traveling.

Recommended Architecture

• Mac mini runs:
• ollama service on localhost:11434
• Open WebUI in Docker on localhost:3000 (optional)
• Secure remote access via Tailscale + SSH (no public open ports)

1) On Mac mini

# Start local LLM server
brew services start ollama

# Optional chat UI
scripts/ai/start-open-webui.sh

2) Set up secure network

1. Install Tailscale on Mac mini, laptop, and phone.
2. Sign into same tailnet account.
3. Enable macOS SSH on mini: System Settings -> General -> Sharing -> Remote Login.

3) From laptop

CLI coding on remote mini

ssh youruser@<mac-mini-tailnet-ip>

Use tmux for persistent sessions:

tmux new -s dev

Tunnel Ollama API to local laptop apps

ssh -N -L 11434:localhost:11434 youruser@<mac-mini-tailnet-ip>

Now on laptop, apps can call http://localhost:11434.

Tunnel Open WebUI

ssh -N -L 3000:localhost:3000 youruser@<mac-mini-tailnet-ip>

Then open http://localhost:3000 on laptop browser.

4) From phone

Two options:

1. Use Tailscale app on phone and open browser to http://<mac-mini-tailnet-ip>:3000
2. Use a zero-trust reverse proxy setup (more advanced)

Security Rules

• Keep Ollama bound to localhost.
• Do not expose 11434 directly to the public internet.
• Prefer Tailscale + SSH tunnels.
• Use long passwords + 2FA for all accounts.

Troubleshooting

• Ollama not responding: brew services restart ollama
• Open WebUI not starting: docker compose -f ai/docker-compose.open-webui.yml logs -f
• SSH issues: verify Remote Login enabled + correct tailnet IP