This guide covers setting up Ollama, a local AI runtime, to assist with modding tasks like code generation, debugging, and documentation.

Note: Using AI for code generation is subject to the mod’s AI Policy.

---

Quick Setup

# 1. Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# 2. Pull a model
ollama pull codellama

# 3. Run the model
ollama run codellama

---

What is Ollama?

Ollama is a tool for running large language models (LLMs) locally on your machine. Unlike cloud-based AI (ChatGPT, Claude), everything runs offline on your hardware.

Benefits for Modding:

• No internet required after initial setup
• Your code stays private
• Free to use (no API costs)
• Can run continuously for assistance

---

Installation

macOS / Linux

# Install via curl
curl -fsSL https://ollama.com/install.sh | sh

# Or via Homebrew
brew install ollama

Windows

Download from ollama.com and run the installer.

Verify Installation

ollama --version

---

Model Selection

Recommended Models for Modding

Pulling Models

# Pull a model (downloads ~4-13GB)
ollama pull codellama
ollama pull llama3

# Check installed models
ollama list

Hardware Requirements

---

Integration with Development

Interactive Mode

# Start an interactive session
ollama run codellama

# Ask questions directly
> How do I add a new focus tree in Millennium Dawn?

Using with Claude Code

You can create a script to pipe context to Ollama:

#!/bin/bash
# ollama-ask.sh

# Get the last commit's diff
DIFF=$(git diff HEAD~1 --stat)

# Send to Ollama with mod context
ollama run codellama <<EOF
You are helping with Hearts of Iron IV mod development.
The recent changes are:
$DIFF

Provide a code review focusing on:
1. Style consistency with Millennium Dawn standards
2. Potential bugs
3. Performance considerations
EOF

VS Code Integration

Using Continue Extension

1. Install Continue VS Code extension
2. Configure in ~/.continue/config.json:

{
  "models": [
    {
      "model": "codellama",
      "provider": "ollama"
    }
  ]
}

Using Cline Extension

1. Install Cline VS Code extension
2. Configure in VS Code settings (Ctrl/Cmd + , → Settings → Extensions → Cline):

{
  "cline.models": [
    {
      "name": "codellama",
      "provider": "ollama",
      "baseUrl": "http://localhost:11434",
      "model": "codellama"
    }
  ],
  "cline.defaultModel": "codellama"
}

3. Alternative: Create .cline/config.json in your project root:

{
  "models": [
    {
      "name": "codellama",
      "provider": "ollama",
      "baseUrl": "http://localhost:11434",
      "model": "codellama"
    }
  ],
  "defaultModel": "codellama"
}

Cline vs Continue Differences:

• Cline: More lightweight, direct Ollama integration, simpler configuration
• Continue: More features, multi-model support, advanced context management
• Performance: Cline typically faster for local models
• Setup: Cline requires less configuration for basic Ollama usage

Cline Usage:

• Use Ctrl/Cmd + Shift + P → “Cline: Ask” for quick queries
• Select text and use Ctrl/Cmd + Shift + P → “Cline: Ask about selection”
• Configure custom prompts in .cline/prompts.json

Millennium Dawn Project Setup with Cline:

Create .cline/prompts.json in your Millennium Dawn project root:

{
  "prompts": [
    {
      "name": "MD Code Review",
      "prompt": "Review this Millennium Dawn code for:\n1. Style compliance (1 tab indent, logging)\n2. Performance issues (avoid every_country)\n3. MD-specific scripted effects usage\n4. Common HOI4 modding errors\n\nCode:\n{selection}"
    },
    {
      "name": "MD Focus Tree Generator",
      "prompt": "Generate a Millennium Dawn focus tree for {country}.\nRules:\n- File: 05_{country}_focus.txt\n- Use relative_position_id\n- Include ai_will_do with game rules\n- Add completion_reward with logging\n- Use search_filters\n\nContext: {context}"
    },
    {
      "name": "MD Debug Script",
      "prompt": "Debug this HOI4 script error:\nError: {error}\n\nCode:\n{selection}\n\nCheck for:\n- Invalid triggers/effects\n- Missing semicolons\n- Incorrect scope usage\n- MD-specific requirements"
    }
  ]
}

Project-Specific Cline Commands:

• Cline: Ask about selection → Use “MD Code Review” prompt
• Cline: Ask → Use “MD Focus Tree Generator” for new focuses
• Cline: Ask → Use “MD Debug Script” for troubleshooting

Using Claude Code with Ollama

Create an alias in your shell:

# Add to ~/.zshrc
alias md-ai='ollama run codellama'

---

Prompts for Modding Tasks

Code Review

You are reviewing Hearts of Iron IV mod code for Millennium Dawn.
Check for:
1. Correct use of MD-specific scripted effects
2. Performance issues (avoid every_country, MTTH events)
3. Style compliance (1 tab indent, logging in completion_reward)
4. Common errors in triggers and effects

Review this code:
[PASTE CODE]

Generating Focus Trees

Create a Millennium Dawn focus tree for [COUNTRY].
Follow these rules:
- File: 05_[country]_focus.txt
- Use relative_position_id for positioning
- Include ai_will_do with game rule checks
- Add completion_reward with logging
- Use search_filters

Country context: [DESCRIBE POLITICAL SITUATION]

Debugging

Help debug this HOI4 script error:
[ERROR MESSAGE]

Code context:
[RELEVANT CODE]

Converting Ideas to Effects

Convert this idea definition to use MD's economic effects:
[IDEA CODE]

Use the scripted effects from common/scripted_effects/00_money_system.txt

---

Advanced Configuration

GPU Acceleration

Ollama uses GPU automatically if available. To verify:

# Check if GPU is being used
ollama list
# Look for GPU indicator in model info

Custom System Prompts

Create a custom model with MD-specific instructions:

# Create a Modfile
cat > Modelfile << 'EOF'
FROM codellama

PARAMETER temperature 0.3

SYSTEM """
You are an expert Hearts of Iron IV modder specializing in Millennium Dawn.
- Follow MD coding standards: 1 tab indent, logging in effects
- Use MD-specific scripted effects from common/scripted_effects/
- Reference docs in docs/dev-resources/ for modifiers and effects
- Always include ai_will_do in focus trees
"""
EOF

# Create custom model
ollama create md-expert -f Modelfile

# Use it
ollama run md-expert

API Server Mode

Run Ollama as an API server for integration:

# Start server on port 11434
ollama serve

# In another terminal, query via API
curl -X POST http://localhost:11434/api/generate -d '{
  "model": "codellama",
  "prompt": "How do I add a new idea in MD?",
  "stream": false
}'

---

Troubleshooting

Slow Performance

• Use a smaller model (4GB instead of 13GB)
• Close other applications to free RAM
• Check GPU utilization

Out of Memory

# Check available memory
free -h  # Linux
Activity Monitor  # macOS

# Kill other Ollama instances
pkill -f ollama

Model Not Found

# Re-pull the model
ollama pull codellama

# Check model list
ollama list

---

Best Practices

1. Use as Assistant, Not Replacement: AI helps with repetitive tasks but review all code
2. Keep Context Local: Don’t paste sensitive information into cloud AIs
3. Verify Before Commit: Always test generated code in-game
4. Use Appropriate Models: codellama for code, llama3 for documentation
5. Iterate: Start with simple prompts, refine based on results

---

Related Resources

• Ollama Documentation
• HOI4 Modding Wiki
• Code Stylization Guide
• Code Resources