Selora Homes

Selora Homes makes Selora AI, the AI layer for Home Assistant: an assistant that responds to plain-English commands, answers questions about your home, and drafts automations for you to review. Selora AI can use a cloud model, or run entirely on your own hardware with the models published here, in which case your home's data never leaves your network.

Selora AI LLM

The language models behind Selora AI's local mode, fine-tuned from Qwen3 for Home Assistant. They come in two sizes, and each size is published in two forms:

• Adapter bundle: one base model plus five small task-specific adapters (LoRA), and the Selora AI integration's default setup. The integration picks the adapter for each request:
  • command controls devices;
  • automation builds automations;
  • answer answers questions about the home;
  • clarification asks a follow-up when a request is ambiguous;
  • utilities gives help drawn from the Home Assistant documentation, with citations.
• Fused build: the five adapters merged into a single model, for runtimes that cannot switch adapters, such as Ollama.

Each model card lists its files and how to run it; the adapter bundles' cards also report benchmark results. The models are released under Apache-2.0, the license of the Qwen3 models they are derived from.

To try a fused build with Ollama:

ollama run hf.co/selorahomes/Selora-AI-LLM-1.7B-ollama:selora-ollama.Q6_K.gguf

Using the models with Home Assistant

Install the Selora AI integration and choose its local backend; the Selora AI Local guide walks through it. On a Selora Hub, the local model is set up for you.

selorahomes.com · GitHub · X · LinkedIn