HaynesOps Home Automation
Welcome to the documentation for the HaynesOps smart home — a Home Assistant + AppDaemon automation platform running on Kubernetes.

An AppDaemon-powered dashboard, driven by the code in this repo.
Highlights
- AI-powered camera notifications — motion-triggered detection with LLM-generated summaries and AI illustrations delivered as push notifications
- Occupancy-based lighting — mmWave presence sensors + Inovelli switches for automatic, zone-aware lighting across the house
- Immich photo frame — wall-mounted displays cycling personal photos from a self-hosted Immich library
- System health monitoring — proactive infrastructure checks with auto-repair for fans, printers, and the hot tub
- Custom Lovelace dashboards — purpose-built cards for wall displays, mobile, and desktop
How it works
This project has two distinct halves:
Home Assistant YAML — automations, scripts, dashboard cards, helpers, and blueprints. These are the "frontend" of the smart home: they react to device events, control lights, and render dashboards. The YAML in this repository is a mirror of what runs on the HA server — changes are copy-pasted into the HA UI editor or applied via MCP.
AppDaemon Python — backend automation apps that handle complex workflows, AI integrations, and multi-step sequences. These run in a separate Kubernetes pod and communicate with HA via its API. AppDaemon code in this repo is the source of truth — it's developed here and deployed to production.
┌──────────────────────┐ ┌──────────────────────┐
│ Home Assistant │ API │ AppDaemon │
│ │◀───────▶│ │
│ Automations │ │ door_notify │
│ Scripts │ │ detection_summary │
│ Dashboard Cards │ │ photo_frame_viewer │
│ Helpers │ │ immich_fetcher │
│ │ │ │
│ Devices & Entities │ │ AI Providers │
└──────────────────────┘ └──────────────────────┘
│ │
Physical devices External APIs
(switches, cameras, (OpenAI, Gemini, Ollama,
sensors, locks) Immich, ComfyUI)
The AI journey
This smart home has evolved alongside the rapid advancement of LLMs in programming:
- ~2020–2023 — Everything built by hand. AppDaemon apps, HA automations, and Jinja templates all written manually. Functional but slow to iterate.
- 2023–2024 — Started supplementing complex YAML and Jinja with ChatGPT chat window questions. Copy-paste from chat into the HA editor. Huge productivity boost for one-off automations.
- 2025 — Adopted Cursor with full repo context. Automations that previously took hours were done in minutes. Added AI providers to AppDaemon apps for runtime LLM capabilities (detection summaries, image descriptions).
- 2026 — Integrated ha-mcp for direct HA entity manipulation from the IDE. Now using Claude Code and Codex CLI alongside Cursor for multi-agent development workflows. The AI isn't just helping write code — it's creating and managing HA entities, dashboards, and automations directly.
Explore
| Section | What you'll find |
|---|---|
| GenAI Camera Notifications | AI-powered detection summaries and door open alerts |
| Occupancy-Based Lighting | mmWave presence + Inovelli switch automation |
| Immich Photo Frame | Wall display photo slideshow from self-hosted photos |
| System Health Monitoring | Proactive infrastructure monitoring with auto-repair |
| Architecture | System diagram and data flow |
| Getting Started | Development environment setup |