[App][Pro][CLI] LLMVision for Homey Pro v1.7.0 - AI-Powered Camera Analysis

Apps
1

LLMvision for Homey Pro v1.7.0 — AI-Powered Camera Analysis

I’m pleased to share LLMVision for Homey Pro, an independently developed application for AI-powered camera analysis and Homey Advanced Flows.

Why I built it

I wanted to move beyond complicated line-crossing rules and individual camera events.

Instead of simply knowing that a person or vehicle was detected, I wanted my automations to understand what actually happened.

For example, my system can analyze a delivery vehicle arriving, a driver approaching the front door, leaving a package, and departing—and produce one coherent description of the activity.

What it supports

  • OpenAI, Google Gemini, and Anthropic Claude integrations

  • Multi-frame and multi-camera analysis

  • Batch processing and Best Image Selection

  • Person, vehicle, package, and delivery detection

  • Movement direction and security classifications

  • Eight Homey Flow action cards

  • Existing MQTT integration with Home Assistant

Inspiration and independence

This project was inspired by LLM Vision for Home Assistant.

However, LLMVision for Homey Pro is independently developed. It is not a fork, port, or direct replacement, and the projects share no source code or affiliation.

Download and installation

GitHub: GitHub - mckinneypl/homey-llmvision: LLMVision for Homey Pro — AI-powered multi-camera analysis and Advanced Flow integration. · GitHub

Release v1.7.0: Release LLMVision for Homey Pro v1.7.0 · mckinneypl/homey-llmvision · GitHub

The repository includes installation instructions, application source, and documentation.

This is a source-only release, not a Homey App Store application.

Community maintenance

I’m sharing this project under the MIT License in the hope that others will find it useful.

I’m not taking on ongoing development, support, or bug-fixing responsibilities.

If an experienced Homey developer is interested in becoming a volunteer maintainer, I’d welcome that.

— Percy McKinney

2

How I use LLMVision with Home Assistant

Homey Pro is my primary automation platform, while Home Assistant provides the dashboard and event-history interface.

I’ve integrated LLMVision for Homey Pro with Home Assistant using MQTT. When LLMVision analyzes security camera activity, the results are sent to Home Assistant, including:

  • Camera images: The selected Best Image representing the event.

  • Written summaries: An AI-generated description of what happened.

  • Activity details: People, vehicles, deliveries, and movement direction.

  • Security classifications: The assessed security level and confidence.

I use these results to create a visual security dashboard and event timeline in Home Assistant.

For example, when a delivery driver arrives, walks to the front door, leaves a package, and departs, LLMVision can analyze multiple camera frames and generate one written summary of the entire event, along with the most representative image.

This gives me a chronological record of security activity that I can review without searching through multiple camera recordings.

The important distinction: Homey Pro handles the automation and AI analysis, while Home Assistant provides the visualization and event history.

This is how I use the two platforms together. Home Assistant is not required for LLMVision’s core camera analysis functionality.

image
image1605×866 192 KB

3

Sounds quite cool. Think you can add an option to use a local AI (lm studio, llama.cpp, ollama, …)?

4

Thanks! Local AI support is an interesting possibility, especially for privacy and avoiding cloud API costs.

The current version supports OpenAI, Gemini, and Claude. Adding support for Ollama, LM Studio, or other local AI endpoints could be a worthwhile enhancement.

I don’t have any immediate plans to add new functionality. However, I continue to use LLMVision in my own home automation system, and if I develop additional enhancements, I’ll be happy to share them with the community.

I also welcome contributions from anyone interested in extending the application or volunteering to maintain it.

My goal in releasing LLMVision was to share something useful and give back to the community.