AI Baritone Commander v2

Control Minecraft with natural language using local AI

Type /ai do mine 10 diamonds and watch your character automatically navigate and mine!

---

✨ Features

• Natural Language Control - Just type what you want in plain English
• 100% Local & Free - Uses Ollama, runs entirely on your computer
• Powered by Baritone - Reliable pathfinding and automation
• Two-Stage Planning - Fast router + smart planner for optimal speed
• Plan Caching - Avoids redundant LLM calls for repeated commands
• Model Presets - Choose between fast, balanced, and thinking modes
• Safe Mode - Prevents dangerous actions when low on health

---

🚀 Quick Installation

Option 1: Download Release (Recommended)

1. Download the latest .jar file from Releases
2. Install prerequisites:
   • Fabric API (for Minecraft 1.20.4)
   • Baritone (baritone-api-fabric-1.10.4.jar)
   • Ollama
3. Place all .jar files in your .minecraft/mods folder
4. Start Ollama: ollama serve
5. Download recommended models: /ai download
6. Launch Minecraft and try: /ai do mine 10 stone

Option 2: Build from Source

git clone https://github.com/yourusername/ai-baritone-commander-v2.git
cd ai-baritone-commander-v2
./gradlew build

The compiled mod will be in build/libs/aicommander-*.jar

---

📋 Requirements

• Minecraft 1.20.4
• Java 17-21
• Fabric Loader 0.15+
• Fabric API
• Baritone
• Ollama

---

💬 Example Commands

/ai do mine 10 iron ore
/ai do go to 100 64 -200
/ai do explore the area
/ai do farm wheat
/ai do follow Steve
/ai stop

---

🎮 Commands

Core Commands

Command	Description
/ai do <instruction>	Execute AI command
/ai stop	Stop all actions and clear queue
/ai pause	Pause execution
/ai resume	Resume execution
/ai status	Show system status
/ai help	Show help

Model Management

Command	Description
/ai download	Show recommended models
/ai download qwen	Download qwen2.5:7b
/ai download phi	Download phi3.5:mini
/ai download llama	Download llama3.2:3b
/ai download mistral	Download mistral-nemo:12b
/ai models	List installed models
/ai set <preset>	Set preset (fast/balanced/thinking)
/ai modelA <name>	Override Stage A model
/ai modelB <name>	Override Stage B model

Utility Commands

Command	Description
/ai cache clear	Clear plan cache
/ai cache stats	Show cache statistics
/ai debug	Toggle debug mode
/ai dryrun	Toggle dry run (preview only)
/ai safe	Toggle safe mode
/ai reload	Reload config

---

🤖 Recommended AI Models

Rank	Model	Use Case	Command
🥇 #1	qwen2.5:7b	Best overall planner	/ai download qwen
🥈 #2	phi3.5:mini	Best fast router	/ai download phi
🥉 #3	llama3.2:3b	Best simple all-round	/ai download llama
🏅 #4	mistral-nemo:12b	Best thinking mode	/ai download mistral

Model Presets

• Fast - phi3.5:mini → qwen2.5:7b (fastest responses)
• Balanced - llama3.2:3b → qwen2.5:7b (default, best overall)
• Thinking - llama3.2:3b → mistral-nemo:12b (smartest, slower)

Use /ai set fast, /ai set balanced, or /ai set thinking to switch.

---

🔧 Configuration

Config file location: .minecraft/config/aicommander.json

Key settings:

• activePreset - Model preset (fast/balanced/thinking)
• safeMode - Enable safety checks (default: true)
• maxActionsPerPlan - Max actions per plan (default: 8)
• planCacheTTLSeconds - Cache time-to-live (default: 45)

---

🔧 Troubleshooting

"Ollama not running" → Start Ollama: ollama serve

"Baritone not detected" → Install baritone-api-fabric-1.10.4.jar

"Model not installed" → Use /ai download to see and install recommended models

Slow responses → Use fast preset: /ai set fast

Messages not showing in chat → Messages are queued - wait a moment for them to appear

---

🏗️ Architecture

• CommandRouter - Fast-path routing for common commands (no LLM)
• PlanCache - LRU cache with 45s TTL (max 20 entries)
• TwoStagePlanner - Stage A (fast intent parsing) → Stage B (full planning)
• MinimalWorldStateCollector - Compact world state (~500 tokens)
• ActionExecutor - Queue-based execution with Baritone integration

---

📄 License

MIT License - See LICENSE

---

Made with ❤️ for Minecraft automation