← Back to Skills Marketplace
Ollama Ollama Herd
by
Twin Geeks
· GitHub ↗
· v1.2.1
· MIT-0
159
Downloads
2
Stars
2
Active Installs
5
Versions
Install in OpenClaw
/install ollama-ollama-herd
Description
Ollama Ollama Herd — multimodal Ollama model router that herds your Ollama LLMs into one smart Ollama endpoint. Route Ollama Llama, Qwen, DeepSeek, Phi, Mist...
Usage Guidance
This skill appears to do what it says (a local Ollama router) but take ordinary precautions before installing and running it: 1) Inspect the PyPI package and its GitHub repository (code, recent commits, maintainer reputation) before running 'pip install'. 2) Install in a contained virtualenv or isolated host (not your primary machine) because pip packages can run code on install. 3) Be aware that 'Auto-pull' will download large model binaries and write to disk and may increase network and storage usage; confirm model sources are trusted. 4) Run herd-node only on machines you control and protect network access (firewall, ACLs) so the router cannot reach unintended hosts. 5) Review files under ~/.fleet-manager for sensitive logging. If you need higher assurance, ask the publisher for the package tarball or source to audit, or request a registry install spec rather than following freeform SKILL.md install steps.
Capability Assessment
Purpose & Capability
Name and description describe a local Ollama router and the SKILL.md instructs installing a Python package (ollama-herd), running 'herd' and 'herd-node', and using local HTTP endpoints — these requirements (curl/wget, optional python/pip/sqlite) are appropriate and expected.
Instruction Scope
Instructions remain within the stated purpose (route requests, list models, embeddings, image/STT endpoints). They do mention an 'Auto-pull' feature that will download missing models automatically — this is consistent with a fleet router but expands scope to outbound network activity and large disk writes. No instructions ask for unrelated credentials or system-wide config access.
Install Mechanism
The SKILL.md tells users to 'pip install ollama-herd' from PyPI. Installing a community PyPI package is a normal mechanism for this functionality but carries the usual risk that an install/run-time package can execute arbitrary code. The registry provides no packaged code to inspect, so the actual package contents and behavior are not validated here.
Credentials
The skill declares no required environment variables or sensitive credentials and only references config paths under ~/.fleet-manager, which are proportional to a fleet manager. The potential for model downloads and writing logs/DB files is expected but may expose disk/network usage and local data — no unexpected credentials are requested.
Persistence & Privilege
always is false and the skill is user-invocable. It runs services (herd/herd-node) and may persist state under ~/.fleet-manager and download models to nodes. If the agent is allowed to invoke skills autonomously, that could trigger downloads/CPU/GPU activity; this is a usage risk rather than an incoherence with the stated purpose.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install ollama-ollama-herd - After installation, invoke the skill by name or use
/ollama-ollama-herd - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.2.1
Cross-platform support: macOS, Linux, and Windows. Updated OS metadata, descriptions, and hardware recommendations.
v1.2.0
- Intensively rebranded the documentation to amplify "Ollama" in all feature descriptions and usage instructions.
- Updated all code snippets, explanations, and examples to highlight Ollama-specific concepts and naming.
- Refined the feature overviews and API usage documentation for clarity and consistent Ollama branding.
- Added multilingual/brief non-English tagline to the description.
- No changes to runtime code, user interface, or configuring features.
v1.1.0
- Updated description to clarify Mac compatibility and multimodal features.
- Minor textual and formatting improvements throughout documentation.
- No functional or behavioral changes to the skill itself.
v1.0.0
- Initial release of Ollama Herd: a smart endpoint for routing Ollama LLM requests across multiple devices.
- Features include 7-signal node scoring, auto-retry, VRAM-aware fallback, and context protection.
- Supports image generation, speech-to-text, and embeddings routing.
- Drop-in compatible with OpenAI SDK and standard Ollama API.
- Includes a web dashboard for monitoring and management.
v1.0.1
- Improved documentation with detailed setup instructions and usage examples for Ollama fleet router.
- Clarified how requests are routed across multiple devices using a 7-signal scoring system.
- Added illustrated features: VRAM-aware fallback, auto-retry, context protection, and automatic model downloading.
- Highlighted additional capabilities, including image generation, speech-to-text, embeddings, and dashboard overview.
- Provided clear guidance on safe operations and guardrails for users.
Metadata
Frequently Asked Questions
What is Ollama Ollama Herd?
Ollama Ollama Herd — multimodal Ollama model router that herds your Ollama LLMs into one smart Ollama endpoint. Route Ollama Llama, Qwen, DeepSeek, Phi, Mist... It is an AI Agent Skill for Claude Code / OpenClaw, with 159 downloads so far.
How do I install Ollama Ollama Herd?
Run "/install ollama-ollama-herd" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Ollama Ollama Herd free?
Yes, Ollama Ollama Herd is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Ollama Ollama Herd support?
Ollama Ollama Herd is cross-platform and runs anywhere OpenClaw / Claude Code is available (darwin, linux, windows).
Who created Ollama Ollama Herd?
It is built and maintained by Twin Geeks (@twinsgeeks); the current version is v1.2.1.
More Skills