Setting upOllama.

Ollama is the local engine - the thing that actually runs the AI models on your machine. It ships inside the Typilot installer, so most users never need this page. Read on if you want to connect Typilot to your own external Ollama instance, or manage models manually from the command line.

Install on macOS.

Optional. Typilot already ships with the engine built in - install a standalone Ollama only if you want to run your own external instance.

MACOS

# Option 1 - Homebrew
brew install ollama

# Option 2 - Installer
open https://ollama.com/download

Verify: open a fresh terminal and run

ollama --version

You may need to allow Ollama in System Settings → Privacy & Security if it is blocked on first launch.

Confirm Typilot can reach it.

Point Settings → Connection at your instance if it is not on the default http://localhost:11434.
Open Typilot, head to Monitoring.
The Ollama tile should show Online with a latency reading.
If not, click Start on the tile - Typilot launches it for you.

Download your first model.

From the Typilot Models page or the CLI:

ollama pull llama3.2:3b

Then return to Typilot - the model appears in the Default Model dropdown. See Choosing AI Models for recommendations.

If something breaks.

Ollama not starting.

Check the tray (Windows/macOS) or `systemctl status ollama` (Linux).
Restart the machine.
Reinstall.
Check firewall - Ollama uses port 11434.

Typilot can't connect.

Open http://localhost:11434 in your browser - you should see an API response.
Confirm Typilot Settings → Connection points at http://localhost:11434.
Restart both Ollama and Typilot.
Check if another app is bound to port 11434.

Models won't download.

Stable internet for the first pull (models can be several GB).
Check available disk space.
Try from Typilot's Models page instead of the CLI.
Verify Ollama is running.

NEXT → 10 CHOOSING AI MODELS

Loading chapter

09 - ADVANCED

Setting upOllama.

Install on macOS.

Optional. Typilot already ships with the engine built in - install a standalone Ollama only if you want to run your own external instance.

MACOS

# Option 1 - Homebrew
brew install ollama

# Option 2 - Installer
open https://ollama.com/download

Verify: open a fresh terminal and run

ollama --version

You may need to allow Ollama in System Settings → Privacy & Security if it is blocked on first launch.

Confirm Typilot can reach it.

Point Settings → Connection at your instance if it is not on the default http://localhost:11434.
Open Typilot, head to Monitoring.
The Ollama tile should show Online with a latency reading.
If not, click Start on the tile - Typilot launches it for you.

Download your first model.

From the Typilot Models page or the CLI:

ollama pull llama3.2:3b

Then return to Typilot - the model appears in the Default Model dropdown. See Choosing AI Models for recommendations.

If something breaks.

Ollama not starting.

Check the tray (Windows/macOS) or `systemctl status ollama` (Linux).
Restart the machine.
Reinstall.
Check firewall - Ollama uses port 11434.

Typilot can't connect.

Open http://localhost:11434 in your browser - you should see an API response.
Confirm Typilot Settings → Connection points at http://localhost:11434.
Restart both Ollama and Typilot.
Check if another app is bound to port 11434.

Models won't download.

Stable internet for the first pull (models can be several GB).
Check available disk space.
Try from Typilot's Models page instead of the CLI.
Verify Ollama is running.

NEXT → 10 CHOOSING AI MODELS