Skip to content

Using Ollama ​

Run local language models in 120 AI Chat or 120 HN through Ollama.

What you need ​

  • A computer that can run both your 120.dev app and Ollama.
  • 120 AI Chat or 120 HN.
  • An internet connection to download Ollama and models.

Check Ollama's macOS requirements or Windows requirements before installing.

Install Ollama ​

  1. Download the installer for your operating system from Ollama.
  2. On macOS, open the .dmg and drag Ollama into Applications. On Windows, run the installer and follow its prompts.
  3. Open Ollama and keep it running so your 120.dev app can connect to its local server.

Download a local model ​

  1. Choose a model from the Ollama library that fits your computer's memory and storage.
  2. Open Terminal on macOS or PowerShell on Windows.
  3. Run ollama pull followed by the model's tag. For example:
sh
ollama pull mistral
  1. Wait for the download to finish. Run ollama list to check which models are available locally.

The model stays on your computer after downloading. Download size and memory use depend on the model variant, quantisation, and context length.

Connect your 120.dev app ​

  1. Open 120 AI Chat or 120 HN, then go to Settings > AI Providers.
  2. Select Configure beside Ollama.
  3. Check that API URL is http://localhost:11434 for a standard local installation.
  4. Select Connect to load the available models. If you changed the URL, select Save to keep it.
  5. Choose a downloaded model in the app.

A standard local Ollama server does not require an API key. See Ollama's quickstart for server setup help.

Use account features ​

Basic AI Chat conversations work without signing in. Sign in through Settings > Account to use attachments, web search, and parallel threads, or to use AI summaries in 120 HN. Use your usual 120.dev account.

Local processing and internet access ​

When you use a local model through a server on your computer, model processing takes place on your device. Downloads, account sign-in, updates, and online features such as web search still require internet access. You may need to reconnect periodically to refresh account access.

Example local models ​

The mistral command above is an example, not a requirement. See the Mistral model page for its variants and download details, or browse the Ollama library for alternatives.

Choose a model that fits your available memory and storage. Download size and memory requirements depend on the model variant, quantisation, and context length. Check the selected variant's details instead of relying on a fixed size or RAM estimate for an entire model family.