先看效果：
使用huggingface和英伟达白嫖OpenClaw
1.打开 https://build.nvidia.com/ 申请API密钥，复制API密钥，调用地址，模型ID到txt文稿。
2.打开huggingface，新建 Docker space。
3.添加文件，文件名为Dockerfile。
# 核心镜像选择
FROM node:22-slim
# 1. 基础依赖补全
RUN apt-get update && apt-get install -y --no-install-recommends \
git openssh-client build-essential python3 python3-pip \
g++ make ca-certificates \
&& rm -rf /var/lib/apt/lists/*
# 2. 安装 HF 数据交互工具
RUN pip3 install --no-cache-dir huggingface_hub --break-system-packages
# 3. 构建环境与 Git 协议优化
RUN update-ca-certificates && \
git config --global http.sslVerify false && \
git config --global url."https://github.com/".insteadOf ssh://[email protected]/
# 4. OpenClaw 核心安装
RUN npm install -g openclaw@latest --unsafe-perm
# 5. 环境变量预设
ENV PORT=7860 \
OPENCLAW_GATEWAY_MODE=local \
HOME=/root
# 6. Python 同步引擎 (sync.py)
RUN echo 'import os, sys, tarfile\n\
from huggingface_hub import HfApi, hf_hub_download\n\
from datetime import datetime, timedelta\n\
api = HfApi()\n\
repo_id = os.getenv("HF_DATASET")\n\
token = os.getenv("HF_TOKEN")\n\
def restore():\n\
try:\n\
files = api.list_repo_files(repo_id=repo_id, repo_type="dataset", token=token)\n\
now = datetime.now()\n\
for i in range(5):\n\
day = (now - timedelta(days=i)).strftime("%Y-%m-%d")\n\
name = f"backup_{day}.tar.gz"\n\
if name in files:\n\
path = hf_hub_download(repo_id=repo_id, filename=name, repo_type="dataset", token=token)\n\
with tarfile.open(path, "r:gz") as tar: tar.extractall(path="/root/.openclaw/")\n\
print(f"Success: Restored from {name}")\n\
return True\n\
except Exception as e: print(f"Restore Error: {e}")\n\
def backup():\n\
try:\n\
day = datetime.now().strftime("%Y-%m-%d")\n\
name = f"backup_{day}.tar.gz"\n\
with tarfile.open(name, "w:gz") as tar:\n\
if os.path.exists("/root/.openclaw/sessions"): tar.add("/root/.openclaw/sessions", arcname="sessions")\n\
tar.add("/root/.openclaw/openclaw.json", arcname="openclaw.json")\n\
api.upload_file(path_or_fileobj=name, path_in_repo=name, repo_id=repo_id, repo_type="dataset", token=token)\n\
print(f"Backup {name} Success.")\n\
except Exception as e: print(f"Backup Error: {e}")\n\
if __name__ == "__main__":\n\
if len(sys.argv) > 1 and sys.argv[1] == "backup": backup()\n\
else: restore()' > /usr/local/bin/sync.py
# 7. 启动控制逻辑（NVIDIA 配置版）
RUN echo "#!/bin/bash\n\
set -e\n\
mkdir -p /root/.openclaw/sessions\n\
\n\
# 阶段 3: 执行启动前恢复\n\
python3 /usr/local/bin/sync.py restore\n\
\n\
# 处理地址逻辑\n\
CLEAN_BASE=\$(echo \"\$OPENAI_API_BASE\" | sed \"s|/chat/completions||g\" | sed \"s|/v1/|/v1|g\" | sed \"s|/v1\$|/v1|g\")\n\
\n\
# 阶段 2: 生成网关与模型配置（NVIDIA 版）\n\
cat > /root/.openclaw/openclaw.json <<EOF\n\
{\n\
\"models\": {\n\
\"providers\": {\n\
\"nvidia\": {\n\
\"baseUrl\": \"\$CLEAN_BASE\", \n\
\"apiKey\": \"\$OPENAI_API_KEY\", \n\
\"api\": \"openai-completions\",\n\
\"models\": [{ \n\
\"id\": \"\$MODEL\", \n\
\"name\": \"Kimi K2.5\", \n\
\"contextWindow\": 256000 \n\
}]\n\
}\n\
}\n\
},\n\
\"agents\": { \n\
\"defaults\": { \n\
\"model\": { \n\
\"primary\": \"nvidia/\$MODEL\" \n\
} \n\
} \n\
},\n\
\"gateway\": {\n\
\"mode\": \"local\", \n\
\"bind\": \"lan\", \n\
\"port\": \$PORT,\n\
\"trustedProxies\": [\"0.0.0.0/0\", \"10.0.0.0/8\", \"172.16.0.0/12\", \"192.168.0.0/16\"],\n\
\"auth\": { \n\
\"mode\": \"token\", \n\
\"token\": \"\$OPENCLAW_GATEWAY_TOKEN\" \n\
},\n\
\"remote\": { \n\
\"token\": \"\$OPENCLAW_GATEWAY_TOKEN\" \n\
},\n\
\"controlUi\": { \n\
\"allowInsecureAuth\": true,\n\
\"dangerouslyAllowHostHeaderOriginFallback\": true,\n\
\"dangerouslyDisableDeviceAuth\": true \n\
}\n\
}\n\
}\n\
EOF\n\
\n\
# 增量备份循环 (每 6 小时)\n\
(while true; do sleep 21600; python3 /usr/local/bin/sync.py backup; done) &\n\
\n\
openclaw doctor --fix\n\
exec openclaw gateway run --port \$PORT\n\
" > /usr/local/bin/start-openclaw && chmod +x /usr/local/bin/start-openclaw
EXPOSE 7860
CMD ["/usr/local/bin/start-openclaw"]
添加环境变量：OPENAI_API_BASE（调用地址）（ https://integrate.api.nvidia.com/v1/chat/completions ）
MODEL（模型ID）（我的是moonshotai/kimi-k2.5）
HF_TOKEN（你的hf token，没有就创建，权限为写 一定是私密变量）
OPENAI_API_KEY（你的英伟达token 一定是私密变量）
OPENCLAW_GATEWAY_TOKEN（你的OpenClaw访问密码）
创建datasets，名字是openclaw-data。
再添加一个环境变量： HF_DATASET（你的用户名/openclaw-data）
部署完访问链接：你的用户名-你的space名.hf.space