Ha Llmvision

LLM Vision is a Home Assistant integration that analyzes images, videos and live camera streams with multimodal large language models, so a smart home can describe what it sees and act on it. A typical use is asking whether a package has been delivered and getting an answer based on what the cameras captured. It is aimed at Home Assistant users who want smarter notifications and automations. The integration exposes four actions, for images, videos, streams and data, and works with any camera that can be viewed in Home Assistant. It handles local image and video files as references for people or actions, analyzes camera feeds and Frigate events, and builds a timeline of events around the home that can be queried later. An optimized image pipeline picks the most relevant frames and downscales them to keep token usage and latency low, and it can use hardware acceleration. Users can connect any AI provider they already use or host their own model for maximum privacy. Configuration is written as simple Home Assistant actions in YAML, for example an image analyzer action that takes a message, camera entities and token settings. The project site links to its GitHub repository and release notes, with version 1.7.0 current at the time of review. No pricing page exists and no prices are published, so the integration itself appears to carry no listed cost, though any AI provider you connect may charge for usage.