Skip to content

Ollama Preset Examples

This directory contains examples demonstrating the Ollama preset for managing Ollama installations, service configuration, and LLM model management.

Quick Start

New to Ollama? Start here:

ollama-quick-start.yml

Simple 5-minute demo that installs Ollama and runs test queries: 

mooncake run -c examples/ollama/ollama-quick-start.yml --ask-become-pass
- Installs Ollama - Pulls tinyllama (smallest model, ~637MB) - Starts server - Runs test queries (math, geography)

Examples

ollama-example.yml (Comprehensive)

Complete example demonstrating all Ollama preset capabilities:

  • Installation variations (basic, with service, via specific method)
  • Model management (single, multiple, force re-pull)
  • Service configuration (custom host, models directory, environment variables)
  • Complete deployment workflow
  • Uninstallation scenarios
  • Platform-specific examples (Linux/macOS)
  • Integration with other actions
# Dry-run mode (shows what would happen)
mooncake run -c examples/ollama/ollama-example.yml --dry-run

# Actual execution (requires sudo)
mooncake run -c examples/ollama/ollama-example.yml --ask-become-pass

ollama-quick-start.yml (Beginner-Friendly)

Fast introduction to Ollama preset with minimal configuration:

  • Quick installation
  • Single model download
  • Simple test queries
  • Good for first-time users

Basic Usage

1. Basic Installation

- name: Install Ollama
  preset: ollama
  with:
    state: present
  become: true

2. Install with Service

- name: Install Ollama with service
  preset: ollama
  with:
    state: present
    service: true
  become: true

3. Install and Pull Models

- name: Install Ollama and pull models
  preset: ollama
  with:
    state: present
    service: true
    pull:
      - "llama3.1:8b"
      - "mistral:latest"
  become: true

4. Complete Configuration

- name: Full Ollama deployment
  preset: ollama
  with:
    state: present
    service: true
    method: auto
    host: "0.0.0.0:11434"
    models_dir: "/data/ollama"
    pull:
      - "llama3.1:8b"
    env:
      OLLAMA_DEBUG: "1"
  become: true

Features Demonstrated

  • Installation management (auto, script, package methods)
  • Service configuration (systemd on Linux, launchd on macOS)
  • Model pulling (single, multiple, with force flag)
  • Custom configuration (host, models directory, environment variables)
  • Uninstallation (with optional model removal)
  • Facts integration (automatic detection)
  • Idempotency (won't reinstall if present)
  • Platform support (Linux, macOS)

Supported Platforms

  • Linux (Ubuntu, Debian, Fedora, Arch, etc.)
  • systemd service management

  • Package managers: apt, dnf, yum, pacman, zypper, apk

  • macOS

  • launchd service management
  • Homebrew integration

Tips

  1. Start with dry-run: Use --dry-run to see what will happen
  2. Use facts: Check {{ ollama_version }} before installation
  3. Idempotency: The preset won't reinstall if Ollama is already present
  4. Model size: Consider starting with tinyllama (~637MB) for testing
  5. Service management: Use service: true for production deployments
  6. Sudo required: Most operations need become: true or --ask-become-pass

Documentation

For complete documentation, see:

For questions or issues, see the main Mooncake documentation.