← Back to Skills Marketplace
Self Hosted Ai
by
Twin Geeks
· GitHub ↗
· v1.0.2
· MIT-0
133
Downloads
2
Stars
2
Active Installs
3
Versions
Install in OpenClaw
/install self-hosted-ai
Description
Self-hosted AI — run your own LLM inference, image generation, speech-to-text, and embeddings. No cloud APIs, no SaaS subscriptions, no data leaving your net...
Usage Guidance
This skill appears to be a legitimate self-hosted router for local LLMs, but several red flags mean you should proceed cautiously. Before installing: (1) review the 'ollama-herd' PyPI package and the linked GitHub repository source to confirm what code will run during 'pip install' and at runtime; (2) confirm what 'herd-node' does — especially its LAN auto-discovery behavior — and whether you are comfortable with it probing your local network; (3) check where it will write files (the SKILL.md references ~/.fleet-manager) and ensure that is acceptable or run in an isolated VM/container; (4) verify the origin and trustworthiness of the 'uv' tool and the model backends it installs (they may download large binaries from third parties); (5) avoid installing on sensitive machines or accounts with high privileges; and (6) if you want to proceed, prefer testing in a sandboxed environment (VM/container) and inspect network traffic during initial runs. If you can provide the PyPI project source or a link to the exact package release, I can re-evaluate and raise/lower confidence.
Capability Assessment
Purpose & Capability
The name/description match the instructions (run a self-hosted router, local LLM/image/speech endpoints). However the SKILL.md expects you to pip install 'ollama-herd', run 'herd'/'herd-node', and use an 'uv' tool to install model backends — yet the registry metadata only declares curl/wget as required binaries and lists no required config paths or env vars. The SKILL.md embeds metadata that refers to '~/.fleet-manager' config files but the registry lists none. These mismatches make it unclear what actual system access and tools the skill expects.
Instruction Scope
Instructions tell the agent/user to pip install a PyPI package, run local services (herd, herd-node) which auto-discover devices on the local network, and curl localhost endpoints. Those actions are coherent with a fleet router but imply network scanning/auto-discovery and writing runtime config/logs under the home directory. The SKILL.md does not explicitly warn that 'herd-node' will probe the LAN or create files under ~/.fleet-manager.
Install Mechanism
There is no formal install spec in the registry; the SKILL.md instructs use of 'pip install ollama-herd' (PyPI). pip installation runs arbitrary code and can modify the system; the skill also instructs running 'uv tool install ...' which may download/execute additional binaries or model weights. No integrity/release URLs or alternative verified install methods are provided.
Credentials
The skill declares no required environment variables or primary credentials (which is appropriate). But it will request network access (local LAN auto-discovery), create config/log files under ~/.fleet-manager (present in SKILL.md metadata), and may download large model artifacts — all of which are broader privileges than a simple local curl/wget requirement would suggest.
Persistence & Privilege
The registry flags are reasonable: always=false and user-invocable=true. The skill will likely create persistent local state (service, config files, logs) when you run the recommended installer/agents, but it does not request forced always-on inclusion in agents.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install self-hosted-ai - After installation, invoke the skill by name or use
/self-hosted-ai - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.2
Cross-platform support: macOS, Linux, and Windows. Updated OS metadata, descriptions, and hardware recommendations.
v1.0.1
- Expanded multilingual description to include Chinese and Spanish.
- Revised all documentation and examples to emphasize "self-hosted" benefits and usage throughout.
- Clarified that all requests, processing, and guardrails operate fully within local, self-hosted infrastructure, with no cloud dependencies.
- Updated cost comparisons and feature tables to specify "self-hosted" alternatives for each cloud service.
- Improved clarity and consistency in instructions, replacing generic terms with "self-hosted" across guides and usage notes.
v1.0.0
- Initial release of self-hosted-ai (v1.0.0).
- Enables self-hosted LLM inference, image generation, speech-to-text, and embeddings on Mac and Linux devices.
- Serves as a local alternative to OpenAI, DALL-E, Whisper API, and cloud embedding services.
- Provides a router for automatic load balancing and device discovery across your local network.
- No cloud APIs, SaaS subscriptions, or data leaving your hardware—data sovereignty and zero per-request costs.
- Easy install with `pip`, no Docker or Kubernetes required.
Metadata
Frequently Asked Questions
What is Self Hosted Ai?
Self-hosted AI — run your own LLM inference, image generation, speech-to-text, and embeddings. No cloud APIs, no SaaS subscriptions, no data leaving your net... It is an AI Agent Skill for Claude Code / OpenClaw, with 133 downloads so far.
How do I install Self Hosted Ai?
Run "/install self-hosted-ai" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Self Hosted Ai free?
Yes, Self Hosted Ai is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Self Hosted Ai support?
Self Hosted Ai is cross-platform and runs anywhere OpenClaw / Claude Code is available (darwin, linux, windows).
Who created Self Hosted Ai?
It is built and maintained by Twin Geeks (@twinsgeeks); the current version is v1.0.2.
More Skills