mysnipbox

Models & Setup

Ollama setup guide

Install Ollama, allow mysnipbox to connect, download the recommended models and confirm it's running.

Four steps to get Ollama running and connected to mysnipbox. Follow them in order: Step 2 is the one most people miss.

Step 1. Install Ollama

Ollama is the free app that runs AI models on your own computer. Download the installer for your system from the official site.

  • macOS: open the downloaded file, drag Ollama into Applications and launch it. A llama icon appears in the menu bar while it runs.
  • Windows: run OllamaSetup.exe and follow the prompts. Ollama starts in the background and shows a tray icon.
  • Linux: run the one-line install script below, which also sets up the background service.
Linux
curl -fsSL https://ollama.com/install.sh | sh

To check it works, open http://localhost:11434 in a browser. You should see "Ollama is running".

Step 2. Let mysnipbox connect (OLLAMA_ORIGINS)

macOS

  1. 1Quit Ollama from its menu-bar icon.
  2. 2Open Terminal and run the command below. It sets the variable for apps you launch afterwards.
  3. 3Open Ollama again.
Terminal
launchctl setenv OLLAMA_ORIGINS "*"

Prefer the terminal? Quit the Ollama app, then start the server yourself with the command below and keep that window open:

Terminal
OLLAMA_ORIGINS="*" ollama serve

launchctl setenv only lasts until you restart the Mac, so run it again after a reboot (or add it to a login script).

Windows

  1. 1Quit Ollama from the system tray (right-click its icon → Quit).
  2. 2Open the Start menu, search for Edit environment variables for your account and open it.
  3. 3Under User variables, click New…, enter OLLAMA_ORIGINS as the variable name and * as the value, then click OK.
  4. 4Start Ollama again from the Start menu.

Linux (systemd)

Run sudo systemctl edit ollama.service, add the two lines below, save, then run sudo systemctl daemon-reload && sudo systemctl restart ollama.

systemd override
[Service]
Environment="OLLAMA_ORIGINS=*"

Step 3. Download recommended models

mysnipbox uses two kinds of model. Download them from inside the app with one click, or from the terminal. ollama run downloads a model the first time and then opens a chat (type /bye to leave). To only download, use ollama pull <model>.

Tag extraction (vision)

  • Llama 3.2 Vision (11B): high-quality, in-depth analysis. A large download (about 8 GB) that needs plenty of memory.
  • Qwen2.5-VL 7B: a strong all-rounder at a smaller size (about 6 GB).
shell
ollama run llama3.2-vision
ollama run qwen2.5vl:7b

Tag translation (text)

  • Qwen2.5 (3B): light and fast, and it keeps Korean, Chinese and Japanese vocabulary well.
shell
ollama run qwen2.5:3b

Step 4. Check it in mysnipbox

  1. 1Open mysnipbox and go to Settings.
  2. 2In the vision provider section, choose Ollama. Its status should read Running.
  3. 3Your downloaded local models appear in the model list. Pick one for tag extraction (vision) and one for tag translation (text).
  4. 4Run a test image. The first request is slower while the model loads into memory.

Still not detected?

  • Make sure Ollama is running: http://localhost:11434 in a browser should say "Ollama is running".
  • Redo Step 2, then fully quit and reopen Ollama. The variable is only read when Ollama starts.
  • Model not listed: download it first and check the name matches exactly. ollama list shows what's installed.
  • More on hardware, remote servers and troubleshooting: Local models (Ollama).