Local OpenAI LLM for Home Assistant

Local OpenAI LLM for Home Assistant connects conversation agents and AI tasks to models served through an OpenAI-compatible API. It's for Home Assistant users who want to choose their own model server for voice conversations and generation tasks.

• Streamed conversations: Conversation agents stream model responses and support streamed text-to-speech, with history trimming to help stay within the model's context window.
• Image and text tasks: AI Task entities accept image inputs and support image generation when the selected model and server provide those capabilities.
• Tool use: Generate Data actions can use tools, and the integration supports parallel tool calls where the model and server support them.
• Loaded model reuse: With llama.cpp, the integration can prefer a compatible model already in memory to avoid waiting for another model to load.

The integration runs inside Home Assistant; a separate server runs the model. Supported local backends include Ollama, llama.cpp, LM Studio, LocalAI and vLLM. API keys are optional for local servers that don't require authentication.

It also connects to cloud providers such as OpenRouter, DeepSeek and Google Gemini. Those connections send requests to the selected provider rather than processing them on your hardware. Image inputs, image generation and tool use depend on the chosen backend's capabilities.

The project is a fork of Home Assistant's OpenRouter integration and is open source under Apache 2.0.