win10ogod/mc-multimodal-agent

A Mineflayer + OpenAI Responses API agent that acts like a human Minecraft player. It’s intended for developers, researchers, modders and server admins who want an autonomous or assistive agent inside Minecraft that reasons from vision and tools. Key features: - Multimodal LLM loop: first‑person raster frames, structured tool-call outputs, and function tools. - Full Mineflayer toolset: movement, looking, digging/placing, crafting, inventory, pathfinding, chat. - Blueprint support (.litematic) with preflight, continue/repair flows. - Web search via a browser CLI, dynamic item/block catalog, runtime registry for modded servers. - Persistent layered memory (LevelDB), goal trees, scheduled tasks, persona (soul.md), and skill recording (JSON+MD). - Works with OpenAI Responses or Chat Completions (structured outputs), retries/backoff, and configurable/modded server tolerance.

このRepoのdescription / GitHub topics / READMEのAI要約に出現したAI関連キーワードをカテゴリ別に表示。各バッジは該当カテゴリの詳細ランキングへリンクします。