Voyager
Voyager is an open-ended embodied agent that uses large language models (LLMs) to enable lifelong learning in Minecraft. It automatically explores, acquires diverse skills, and makes discoveries without human intervention, solving the problem of building agents that can continuously learn and adapt. It is intended for researchers and developers interested in embodied AI and LLM-powered agents.
✨ Key features
- Automatic curriculum that maximizes exploration
- Ever-growing skill library of executable code
- Iterative prompting mechanism with environment feedback
- Uses GPT-4 via blackbox queries, no fine-tuning
- Skills are temporally extended, interpretable, and compositional
- Outperforms prior SOTA in Minecraft tasks
🎯 Use cases
- Lifelong learning in Minecraft environments
- Acquiring diverse skills without human intervention
- Solving novel tasks from scratch using learned skills
- Research on embodied agents and LLM integration
📦 Installation
🧰 Requirements: Python ≥ 3.9, Node.js ≥ 16.13.0, Minecraft game, OpenAI API key, and optionally Azure login.
git clone https://github.com/MineDojo/Voyager
cd Voyager
pip install -e .
cd voyager/env/mineflayer
npm install -g npx
npm install
cd mineflayer-collectblock
npx tsc
cd ..
npm install
Additionally, follow the instructions in installation/minecraft_instance_install.md and installation/fabric_mods_install.md to set up Minecraft and Fabric mods.
🚀 Usage
from voyager import Voyager
azure_login = {
"client_id": "YOUR_CLIENT_ID",
"redirect_url": "https://127.0.0.1/auth-response",
"secret_value": "[OPTIONAL] YOUR_SECRET_VALUE",
"version": "fabric-loader-0.14.18-1.19",
}
openai_api_key = "YOUR_API_KEY"
voyager = Voyager(
azure_login=azure_login,
openai_api_key=openai_api_key,
)
# start lifelong learning
voyager.learn()
⚠️ Good to know
The project is strictly for research purposes and not an official product from NVIDIA.
❓ FAQ
What are the system requirements for Voyager?
Voyager requires Python ≥ 3.9 and Node.js ≥ 16.13.0. It has been tested on Ubuntu 20.04, Windows 11, and macOS.
How do I get an OpenAI API key?
You can obtain an OpenAI API key from https://platform.openai.com/account/api-keys.
Can I resume learning from a checkpoint?
Yes, you can resume by instantiating Voyager with ckpt_dir and resume=True.
How do I run Voyager for a specific task with a learned skill library?
Instantiate Voyager with skill_library_dir and resume=False, then use decompose_task and inference methods.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.