Selora AI Local

What is Selora AI Local?

Selora AI Local is Selora’s own model, fine-tuned specifically for Home Assistant. It understands entities, areas, automations, and the way Home Assistant is configured, so it produces relevant automation suggestions and natural-language responses without relying on a general-purpose cloud model.

Because it runs entirely on your own hardware, it is the best option for optimal privacy: your device names, entity states, and automation text never leave your home network, and Selora AI keeps working even without an internet connection.

The models are published on Hugging Face under the selorahomes account :

• Selora-AI-LLM-1.7B : the default, runs on every Selora Hub and comfortably on modest hardware.
• Selora-AI-LLM-14B : the larger, more accurate line. Plan for about 16 GB of memory for the model, its cache and Home Assistant together. On Selora Hub, it runs on Hubs with 32 GB of RAM, which come with larger plans.
• Fused builds, 1.7B and 14B : a single model for runtimes that cannot hot-swap LoRA adapters, Ollama being the common case. They score lower than the adapters, so use the adapters when your runtime supports them.

Why run locally?

• Optimal privacy: nothing is sent to Selora Homes or any third party. All inference happens on your machine.
• Works offline: no internet connection required once the model is downloaded.
• No per-request cost: no API keys, no credits, no usage billing.
• Tailored to Home Assistant: fine-tuned on Home Assistant configuration and automation patterns rather than general web text.

Setup

Selora Hub users

Selora Hub runs Selora AI Local for you. Enable it from your Selora Homes account , with no separate download or server setup required. Every Hub runs the 1.7B, and the 14B is rolling out to Hubs with 32 GB of RAM.

Self-hosted

Requirements:

• A machine on your local network capable of running the model (a modern CPU works; a GPU significantly improves response speed).
• llama.cpp running as a local server, reachable from your Home Assistant host.
• Home Assistant with the Selora AI integration installed (see Installation ).

Steps:

1. Build or install llama.cpp on a machine reachable from Home Assistant.
2. Download Selora-AI-LLM-1.7B , or Selora-AI-LLM-14B on a larger machine, and serve it with the llama.cpp server.
3. In the Selora AI integration, choose the local backend and point it at your llama.cpp server URL and the Selora AI model name. See the local backend settings under Configuration .

For full privacy details across every backend, see Privacy & Support .