penguin Ubuntu Ollama — Ubuntu多机部署Ollama

v1.0.0

在Ubuntu系统上用apt、systemd与NVIDIA CUDA一键安装Ollama,支持fleet routing,把推理请求智能分发到多台Ubuntu机器,实现高可用、负载均衡的本地大模型集群。

0· 69·1 当前·1 累计
twinsgeeks 头像by @twinsgeeks (Twin Geeks)
下载技能包
最后更新
2026/4/4
0
安全扫描
VirusTotal
可疑
查看报告
OpenClaw
安全
medium confidence
The skill's instructions, requirements, and metadata are coherent with its stated purpose (setting up an Ollama fleet on Ubuntu), but it instructs running privileged installs and remote installers which carry operational risk and deserve caution.
评估建议
This skill appears to do what it says (set up an Ollama fleet), but it requires running network installers and creating system-level services — treat it like any other system service install: 1) Verify upstream sources: confirm the ollama.ai installer and the 'ollama-herd' package are the official/reputable releases referenced by the GitHub repo listed in the skill. 2) Inspect the code/package before pip installing (or install in an isolated environment/container/VM). 3) Avoid piping unknown scr...
详细分析 ▾
用途与能力
Name/description (Ubuntu Ollama fleet/router) matches the instructions: installing Ollama, pip-installing a 'herd' package, configuring systemd, opening a port, and using mDNS for discovery. Requested binaries (curl/wget, optional python3/pip, systemctl, apt, nvidia-smi) are appropriate for the described tasks.
指令范围
SKILL.md instructs the agent/operator to run network installs (curl | sh from ollama.ai), pip install a package (ollama-herd), write systemd unit files under /etc/systemd/system, enable services, open firewall ports, and rely on mDNS/HTTP for node discovery. These actions are within the expected scope of deploying a cluster but require root privileges and expose a network service; the doc does not describe authentication for the router endpoint (examples use api_key='not-needed').
安装机制
There is no registry install spec (instruction-only), but the runtime instructions direct execution of remote installers (curl -fsSL https://ollama.ai/install.sh | sh) and pip install of 'ollama-herd'. Fetching and executing remote scripts and installing packages from PyPI are standard but high-impact operations — they should be verified before running. The referenced domains (ollama.ai, github.com repo in homepage) are recognizable, but the skill does not include integrity checks or pinned releases.
凭证需求
The skill requests no environment variables or credentials and the declared config paths (~/.fleet-manager/...) are consistent with a fleet manager. There are no requests for unrelated secrets or system-wide config beyond systemd and network settings.
持久化与权限
Instructions create and enable system-wide systemd services and modify firewall rules — actions that require root privilege and will persist across reboots. This persistence is reasonable for a long-running router/node service, but it increases blast radius if the installed software is malicious or misconfigured (e.g., an unauthenticated HTTP API exposed on port 11435).
安全有层次,运行前请审查代码。

运行时依赖

🖥️ OSLinux

版本

latestv1.0.02026/4/4

New skill: Ubuntu Ollama fleet routing with apt, systemd, and CUDA

可疑

安装命令

点击复制
官方npx clawhub@latest install ubuntu-ollama
镜像加速npx clawhub@latest install ubuntu-ollama --registry https://cn.longxiaskill.com
数据来源ClawHub ↗ · 中文优化:龙虾技能库