The Entity
The Entity 将你的单人 Minecraft 世界转变为一个具有响应性、由 AI 驱动的沙盒。这个模组不仅仅是一个有用的聊天机器人,它会主动观察你的游戏玩法,并扮演动态的地下城主(Dungeon Master)。
它会读取你的生命值、物品栏、位置以及最近的操作,并实时作出反应。根据其配置的人格,它可能会悄悄引导你前往安全地带,用幻听来对你进行心理操控,或者 actively 试图杀死你。
功能
- 主动监控: 模组会持续监测你的状态。它知道你是否正在失血过多、手持生猪肉、在森林中冲刺,或是正在河中溺水。
- 动态世界控制: AI 可以执行命令来生成生物、改变天气、调整时间、播放声音、掉落物品,并在聊天栏或动作栏中与你对话。
- 无限自定义: The Entity 的大脑完全由简单的 YAML 文件控制。通过编辑或创建
.minecraft/config/the-entity/prompts 中的新配置文件,你可以重写其系统提示词。将其变成一个仁慈的自然之灵、一个神秘的外层空间恐怖存在,或者一个因吐槽你糟糕生存技能而精神失常的管理者,限制只在于你的想象力。
设置指南(本地 LLM)
此模组需要大语言模型(LLM)才能运行。虽然你 可以 使用付费 API,但轮询游戏状态的成本很快会变得高昂。强烈建议运行本地模型或使用如 Groq 这样的免费 API。
以下是使用 llama.cpp 的新手友好型设置步骤:
1. 下载 llama.cpp
从官方仓库获取适用于你操作系统的最新版本:ggml-org/llama.cpp。
2. 获取模型
为了在逻辑能力和硬件性能之间取得最佳平衡,推荐使用 Qwen 3.5 系列模型。你可以在 Hugging Face 上的 Qwen 3.5 unsloth Collection 中找到多个版本。
- 最低要求: 8B 模型。
- 推荐: 27B 或 35B(如果你有足够的显存),以获得最佳的情境感知能力。
3. 启动服务器
打开终端并启动 llama.cpp 服务器。你必须包含特定的标志以确保模组正常工作。例如,使用 27B 模型:
llama-server -hf unsloth/Qwen3.5-27B-GGUF:UD-Q4_K_XL --reasoning off -c 16384
--reasoning off:防止模型陷入内部思考循环。
-c 16384:将上下文窗口设置得足够大,以便 AI 能够记住你最近的游戏历史和物品栏状态。
4. 连接模组
启动 Minecraft,进入你的世界,并从 Mod Menu 打开 The Entity 的配置界面。将端点指向你的本地服务器(通常是 http://127.0.0.1:8080),根据个人喜好调整其余配置值,保存后看看你能存活多久。
注意: 根据你选择的本地模型,此模组对资源消耗非常大。目前仅支持 单人模式。
The Entity
The Entity turns your singleplayer Minecraft world into a reactive, AI-driven sandbox. Instead of being another helpful chatbot, this mod actively watches your gameplay and acts as a dynamic Dungeon Master.
It reads your health, inventory, location, and recent actions, and responds in real-time. Depending on its configured personality, it might quietly guide you to safety, gaslight you with phantom sounds, or actively try to kill you.
Features
- Active Surveillance: The mod constantly monitors your state. It knows if you are bleeding out, holding a raw porkchop, sprinting through a forest, or drowning in a river.
- Dynamic World Control: The AI can execute commands to spawn mobs, change the weather, alter the time of day, play sounds, drop items, and speak to you in chat or via action bars.
- Infinite Customization: The Entity's brain is entirely controlled by simple YAML files. You can rewrite its system prompt by editing or creating new configurations in
.minecraft/config/the-entity/prompts. Turn it into a benevolent nature spirit, a cryptic cosmic horror, or an unhinged manager roasting your terrible survival skills, the limit is only your imagination.
Setup Guide (Local LLM)
This mod requires an LLM to function. While you can use paid APIs, polling the game state gets expensive quickly. Running a local model or using a free API like Groq is highly recommended.
Here is the beginner-friendly setup using llama.cpp:
1. Download llama.cpp
Grab the latest release for your OS from the official repository: ggml-org/llama.cpp.
2. Get a Model
For the best balance of logic and hardware performance, Qwen 3.5 models are recommended. You can find several versions in the Qwen 3.5 unsloth Collection on Hugging Face.
- Minimum requirement: An 8B model.
- Recommended: 27B or 35B (if you have the VRAM for it) for the best situational awareness.
3. Start the Server
Open your terminal and boot up the llama.cpp server. You must include specific flags for the mod to work correctly. For example, using the 27B model:
llama-server -hf unsloth/Qwen3.5-27B-GGUF:UD-Q4_K_XL --reasoning off -c 16384
--reasoning off: Prevents the model from getting stuck in internal thinking loops.
-c 16384: Sets the context window high enough so the AI can remember your recent game history and inventory state.
4. Connect the Mod
Boot up Minecraft, load into your world, and open The Entity's config screen from Mod Menu. Point the endpoint to your local server (usually http://127.0.0.1:8080), tune the rest of the configuration values to your taste, save, and see how long you survive.
Note: This mod is heavily resource-intensive depending on the local model you choose to run. It is currently designed for Singleplayer only.
评论 0
登录后参与评分
还没有评论,来抢沙发吧