← 返回 Skills 市场
Ollama Load Balancer
作者
Twin Geeks
· GitHub ↗
· v1.0.4
· MIT-0
256
总下载
0
收藏
3
当前安装
6
版本数
在 OpenClaw 中安装
/install ollama-load-balancer
功能描述
Ollama load balancer for Llama, Qwen, DeepSeek, and Mistral inference across multiple machines. Load balancing with auto-discovery via mDNS, health checks, q...
安全使用建议
This skill appears internally consistent for running a local Ollama load balancer, but it relies on you pip installing a third-party package and running local servers that manage model downloads and expose admin HTTP endpoints. Before installing: (1) review the ollama-herd PyPI package and GitHub repo to ensure the code matches expectations, (2) prefer installing in a sandbox/VM or isolated network, (3) disable or review auto-pull behavior to avoid unexpected large downloads, and (4) restrict access to the daemon's HTTP port (localhost-only or firewall) to prevent unauthorized remote control.
能力评估
Purpose & Capability
Name/description (Ollama load balancer) matches the runtime instructions: auto-discovery, health checks, routing, and admin HTTP endpoints. Required binaries (curl/wget) and optional python/pip/sqlite3 are appropriate for a Python-based local service.
Instruction Scope
SKILL.md instructs the agent to pip install ollama-herd and run local commands (herd / herd-node) and to call local HTTP endpoints on localhost:11435. It does not instruct reading unrelated system files or exfiltrating secrets. It does include administrative endpoints (pull/delete models) — expected for a load-balancer but potentially powerful if misused.
Install Mechanism
There is no registry install spec; the README tells the user to pip install ollama-herd from PyPI. pip installs are common but will execute third-party code on the host — moderate risk. The SKILL.md points to a PyPI project and GitHub repo (traceable), which is better than an arbitrary download, but users should review the package/source before installing.
Credentials
The skill declares no required environment variables or credentials. The use of FLEET_MAX_RETRIES and runtime settings is plausible for a load balancer; no unexplained secret access is requested.
Persistence & Privilege
always:false (normal). The instructions create and use local config paths (~/.fleet-manager/...), run long-lived processes, and expose an admin HTTP interface. This is consistent with the stated functionality but means the package will persist on disk and run services — run with appropriate isolation and review.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install ollama-load-balancer - 安装完成后，直接呼叫该 Skill 的名称或使用
/ollama-load-balancer触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.4
Cross-platform support: macOS, Linux, and Windows. Updated OS metadata, descriptions, and hardware recommendations.
v1.0.3
Version 1.0.3
- Expanded internationalization in the description (added Chinese and Spanish).
- Reworded across the documentation to consistently refer to "load balancer" for clarity.
- Enhanced deployment and API examples for more explicit load balancer usage.
- Updated feature and endpoint explanations to clarify their relation to the load balancer.
- Incremented version metadata to 1.0.3.
v1.1.0
- Updated the description for improved clarity and to highlight support for Llama, Qwen, DeepSeek, and Mistral.
- Reduced the zombie reaper’s request cleanup threshold from 15 minutes to 10 minutes.
- No changes to functionality or API; documentation only.
v1.0.2
- Updated version to 1.0.2 in SKILL.md.
- Clarified that auto-pull of missing models is now optional and disabled by default; enable via settings API.
- Updated metadata for configPaths location.
- No code or feature changes; documentation and metadata only.
v1.0.1
- Added "optionalBins" and "configPaths" fields to the metadata section in SKILL.md.
- Now explicitly lists optional dependencies: python3, sqlite3, and pip.
- Declares expected config and log file locations for fleet manager operation.
- No changes to code or user-facing features.
v1.0.0
Initial release.
- Load balances Ollama inference across multiple machines with automatic discovery, health checks, and real-time monitoring dashboard.
- Built-in queue management, zero configuration setup, and automatic failover with retry logic.
- Includes zombie request cleanup, VRAM-aware model routing, and operational analytics via SQL queries.
- Web dashboard provides fleet status, trends, insights, and runtime toggles.
- Exposes multiple REST API endpoints for fleet health, request traces, usage, and model management.
- Designed for high availability and operational visibility in Ollama deployments.
元数据
常见问题
Ollama Load Balancer 是什么？
Ollama load balancer for Llama, Qwen, DeepSeek, and Mistral inference across multiple machines. Load balancing with auto-discovery via mDNS, health checks, q... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 256 次。
如何安装 Ollama Load Balancer？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install ollama-load-balancer」即可一键安装，无需额外配置。
Ollama Load Balancer 是免费的吗？
是的，Ollama Load Balancer 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Ollama Load Balancer 支持哪些平台？
Ollama Load Balancer 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（darwin, linux, windows）。
谁开发了 Ollama Load Balancer？
由 Twin Geeks（@twinsgeeks）开发并维护，当前版本 v1.0.4。
推荐 Skills