Voyager

Overview

Voyager is an open-ended embodied agent that uses large language models to autonomously explore and learn within virtual environments. It leverages LLMs for task decomposition, skill discovery, and plan execution without human intervention, building a growing library of skills through environmental feedback.

Best for

Best for
Researchers and developers building autonomous agents for complex simulated environments like Minecraft

Use cases

• Autonomous exploration and skill acquisition in Minecraft via MineDojo
• Long-horizon task planning with natural language instructions
• Benchmarking open-ended learning algorithms for embodied agents

Notes

Voyager is an open-ended embodied agent that uses large language models to autonomously explore and learn within virtual environments. It leverages LLMs for task decomposition, skill discovery, and plan execution without human intervention, building a growing library of skills through environmental feedback.

6,944 stars on GitHub. Last updated 2024-04-03. Licensed MIT.

Use cases

• Autonomous exploration and skill acquisition in Minecraft via MineDojo
• Long-horizon task planning with natural language instructions
• Benchmarking open-ended learning algorithms for embodied agents

Pros

• Enables continuous, self-directed learning without predefined task curricula
• Leverages large language models for flexible goal setting and adaptation
• Strong community support with nearly 7,000 GitHub stars

Cons

• Requires substantial computational resources for LLM inference
• Primarily tied to the MineDojo environment, limiting out-of-the-box use in other domains
• Open-ended exploration can lead to unpredictable or inefficient behavior without additional constraints

Indexed from awesome-langchain and enriched against its public facts.