We are knocking on the door of 2026, and the Home Assistant integrations landscape has shifted massively. Itโs no longer just about turning lights on; itโs about creating a house that actually thinks.
Iโve spent the last year tearing apart my config, breaking things, and rebuilding them better. This isn’t a textbook definition of “the best” software – this is my personal report on the integrations that have survived the chaos on my server. These are the tools that transformed my setup from a hobby into a sentient machine.
SUPERCHARGE HOME ASSISTANT
Stop playing with basic switches. It’s time to integrate AI vision, enterprise-grade virtualization, and local control. This is the ultimate guide for the power user.
The “Beyond Basics” Stack
Most users stop at adding a few lights. You aren’t most users. To build a truly “smart” home, you need deep integrations that offer local control, AI processing, and hardware-level access. We’ve curated the top-tier integrations that transform Home Assistant from a remote control into a central brain.
01. The Foundation
๐๏ธProxmox VE
Before you install a single integration, you need a rock-solid host. Do not run HA on an SD card. Virtualize it.
02. Computer Vision & AI
From motion detection to intelligent understanding.
Traditional cameras just record. Frigate watches and understands. Pair it with a Coral TPU for efficient object detection, and use LLM Vision to actually “ask” your house what it sees.
-
๐น
Frigate NVR Real-time object detection (Person, Car, Dog)
-
๐ง
LLM Vision Generative AI analysis of camera snapshots
-
๐ผ๏ธ
Advanced Camera Card The ultimate UI for PTZ and WebRTC low latency
Resource vs. Intelligence
03. Total Control
Cloud Free & Low Latency
The golden rule of Home Assistant: Keep it Local. Delays are unacceptable. We compare the latency of cloud-based Tuya against LocalTuya and the ultimate DIY solution, ESPHome.
Featured: Bambu Lab
Monitor your 3D prints directly in HA. Control fans, lights, and speed without opening the app.
*Estimated response times in milliseconds. Lower is better.
The Essential Integrations List
Frigate
NVR / VisionLocal AI object detection with Coral TPU support. Ditch cloud fees.
View on GitHub โAdv. Camera Card
FrontendWebRTC streaming, clip gallery, and interactive PTZ menus.
View on GitHub โLLM Vision
AI / LogicMultimodal image analysis for complex automation logic.
View on GitHub โNFAndroidTV
NotificationOfficial, local, reliable text and image notifications for your TV.
View on GitHub โTvOverlay
External / APKDashboard-like overlays and persistent widgets directly on TV.
View on GitHub โMusic Assistant
MediaUnified audio library, DSP, and seamless cross-platform syncing.
View on GitHub โESPHome
FirmwareNo-code firmware for local logic, Bluetooth Proxy, and Voice.
View Documentation โBambu Lab
3D PrintingDeep telemetry, spaghetti detection, and full LAN mode support.
View on GitHub โTuya Local
IoT ControlLocal control of Tuya devices via database presets. No cloud.
View on GitHub โProxmox (Custom)
Infra / HACSActive control (Start/Stop) of VMs and permission repair.
View on GitHub โSpook
UtilityConfig repair, dangling entity detection, and missing services.
View on GitHub โBubble Card
FrontendMobile-first pop-up interface for slick, modern dashboards.
View on GitHub โ1. Frigate: #1 Home Assistant Integration
Frigate
Gold StandardFrigate represents a fundamental paradigm shift. It is a complete Network Video Recorder (NVR) with real-time object detection powered by local AI inference. It does not merely see movement; it understands context. It distinguishes between a car, a dog, a cat, and a person. Crucially, it does this locally, ensuring that sensitive video feeds never leave the LAN for analysis, addressing the growing privacy concerns associated with cloud-based NVRs.
The Frigate Integration, installed via HACS, serves as the bridge between the NVR and Home Assistant. It connects to the Frigate API and MQTT broker to expose entities. It does not perform the object detection itself; rather, it consumes the data generated by the NVR. This separation of concerns is a hallmark of robust systems design. The integration creates binary sensors (e.g.,ย binary_sensor.front_door_person_occupancy), image entities for snapshots, and switches for controlling detection zones dynamically.ย ย ย
- My Experience: I run this locally using on Intel HD Graphics iGPU (i7-7700) and can process 8 cameras in 1080p on internal GPU without issues.
- Why it stays: It ensures my security recordings stay on my LAN, not in some random cloud server.
2. Advanced Camera Card
Advanced Camera Card
Frontend EssentialIf Frigate is the backend intelligence, the Advanced Camera Card (formerly known as the Frigate Card) is the frontend interface. Standard Home Assistant camera cards are historically limited, often displaying delayed HLS streams or static JPEGs. They lack interactivity. The Advanced Camera Card solves this by creating a "dashboard within a dashboard," offering a rich, interactive user interface for surveillance.
- The Killer Feature: Itโs the "dashboard within a dashboard." I can scrub through recorded clips of the delivery driver directly on my wall-mounted tablet without opening a separate media browser.
- Performance: It uses substream switching, so my tablet doesn't melt trying to stream 4K 24/7. It only goes HD when I click it.
3. NFAndroidTV vs. TvOverlay
NFAndroidTV
Official IntegrationTvOverlay
ExperimentalIn a fully integrated smart home, the television is a primary display surface. When the doorbell rings, the content playing on the TV should pause, and a video feed of the visitor should appear. Achieving this on Android TV (Google TV, Nvidia Shield, Fire TV) requires specific software bridges.
When the doorbell rings, I want my TV to tell me. Iโve experimented with everything, but the ecosystem has split into two camps:
- NFAndroidTV (The Reliable One):
- TvOverlay (The Fancy One): If you want a full dashboard overlay on your TV with persistent clocks and widgets, this is the play. Itโs cooler, but requires a bit more tinkering.
| Feature | NFAndroidTV | TvOverlay | PiPup (Legacy) |
| Control Method | HTTP / HA Service | HTTP / MQTT | HTTP / ADB |
| Video Latency | N/A (Image focused) | Moderate/High (RTSP) | Low (WebRTC) |
| Customization | Basic (Position, Color) | High (Layouts, Fixed Widgets) | Moderate |
| Maintenance | Active (Official) | Active (Community) | Archived |
| Local Requirement | 100% Local | Local + API Options | 100% Local |
My Take: I stick with NFAndroidTV for the stability factor.
4. ESPHome: The Hardware Backbone
ESPHome converts the complex task of embedded programming into a configuration-based workflow. Instead of writing C++ in the Arduino IDE to program an ESP32 microcontroller, users define the desired behaviour in a YAML file. ESPHome compiles this into a binary and uploads it to the device OTA (Over-The-Air).ย I can't imagine my smart home without ESPHome. It turns $5 ESP32 chips into enterprise-grade sensors.
Repository: GitHub - esphome/esphome
In 2025, ESPHome has expanded beyond simple sensors into complex subsystems:
- Bluetooth Proxy: With the BTHome standard, any ESP32 running ESPHome can act as a Bluetooth range extender for Home Assistant. This allows users to place "passive" Bluetooth sensors (like Xiaomi thermometers or SwitchBot devices) anywhere in the house, with the ESP32 relaying the data back to HA over WiFi. It effectively creates a mesh network for Bluetooth devices without proprietary hubs.
- Voice Satellites: ESPHome is the hardware backbone of Home Assistant's "Year of Voice." An ESP32-S3 with a microphone and speaker can function as a local voice assistant satellite, running wake-word detection (using openWakeWord) on-chip and streaming audio to HA for processing. This enables a privacy-focused alternative to Alexa or Google Home.
- mmWave Presence: The integration of 24GHz and 60GHz mmWave sensors (like the LD2410) into ESPHome allows for "True Presence" detection. Unlike PIR sensors which require movement, mmWave detects the micro-movements of breathing, ensuring lights stay on even when a user is sitting still reading.
5. Bambu Lab (HA Integration)
Bambu Lab
Maker EssentialBambu Lab printers (X1C, P1S, A1) have dominated the market, but their reliance on cloud services poses a challenge for local control enthusiasts. The ha-bambulab integration by greghesp bridges this gap, exposing granular data and control to Home Assistant. As someone who lives in the overlap between 3D printing and automation, the ha-bambulab integration is vital.
Connectivity Modes: Cloud vs. LAN
Architecturally, the integration supports two connection modes, each with distinct trade-offs:
- Cloud Mode: The integration connects to Bambu Lab's MQTT broker in the cloud. It is easy to set up but introduces latency and external dependency.
- LAN Mode: The integration connects directly to the printer's local MQTT broker. This is the preferred method for power users, ensuring privacy and resilience. However, Bambu Lab's firmware updates in late 2024 and 2025 have occasionally altered the authentication mechanism for LAN mode, requiring the integration to adapt rapidly. Users must generate an Access Code from the printer's screen to authenticate.
Intelligent Failure Detection (Spaghetti Detection)
A critical "Day 2" operation for 3D printing is failure mitigation. While the flagship X1C has built-in AI detection, the P1 and A1 series do not. The Home Assistant ecosystem solves this via the ha-bambu-lab-p1-spaghetti-detection project.
Mechanism: This integration utilizes the camera feed exposed by the main Bambu integration. It pipes frames to an external ML server (often a self-hosted instance of Obico or a local Docker container).
Automation: When the ML model detects a print failure ("spaghetti"), it triggers a Home Assistant automation to Pause the printer immediately and send a notification to mobiles & TV.
6. Tuya, LocalTuya & Tuya Local
The Tuya Protocol War
3-Way Shootout- LocalTuya: GitHub - rospogrigio/localtuya
- Tuya Local: GitHub - make-all/tuya-local
Tuya is the underlying platform for thousands of "white label" smart devices (Smart Life, Gosund, Teckin). The official Tuya integration relies on the Tuya IoT Cloud API. This introduces three major issues:
- Latency: Commands must round-trip to servers (often in China or the US) before executing.
- Reliability: If the internet is down, the light switch does not work.
- Privacy: Usage data is externalized.
Power users universally reject this model in favor of local control. This has led to the development of two competing custom integrations: LocalTuya and Tuya Local.
Iโve switched to Tuya Local (maintained by make-all). It has a massive community database. You add a device, select it from a list, and it just works. Itโs much more stable and reconnects faster after a power outage.
| Feature | Official Tuya | LocalTuya (rospogrigio) | Tuya Local (make-all) |
| Control | Cloud API | Local LAN | Local LAN |
| Setup Difficulty | Low | High (Manual DPS mapping) | Moderate (Device Presets) |
| Reliability | Internet dependent | Variable (Connection drops) | High |
| Device Support | Universal | Manual Config required | Database Driven |
7. LLM Vision
LLM Vision
AI BrainThe "Brain": This is the most futuristic part of my setup. I use LLM Vision to let Home Assistant "see."
Use Cases for the Power User
This integration opens up entirely new categories of automation that were previously impossible with simple motion sensors or object detectors.
- Contextual Security:
- Trigger: Frigate detects a person at the door.
- Action: Send the snapshot to LLM Vision with the prompt: "Describe the person. Are they wearing a delivery uniform? Are they holding a package? Is there a weapon visible?"
- Result: Home Assistant sends a notification: "FedEx Driver detected with a package" rather than a generic "Person Detected".
- Analog to Digital Conversion:
- Scenario: You have an old boiler with an analog pressure gauge.
- Action: A camera takes a photo. LLM Vision is prompted: "Read the value on the pressure gauge. Return only the number."
- Result: You now have a digital sensor entity tracking the boiler pressure without invasive plumbing work.
- Visual Verification:
- Prompt: "Is the garage door open or closed?" (Analyzed from a camera feed, functioning as a redundant sensor if the reed switch fails).
8. Proxmox (with Custom Integration)
Proxmox (Custom)
InfrastructureOfficial vs. Custom Integration:
- Official Integration: Provides read-only sensors: CPU usage, RAM usage, and binary sensors for VM state (Running/Stopped). It is useful for basic monitoring dashboards.
- Custom Integration (
dougiteixeira): This HACS integration is designed for active management. It replaces the core integration and adds control capabilities. It exposes switches to Start, Stop, Restart, Suspend, and Hibernate VMs and LXC containers. This allows for automations like "Restart the Frigate LXC container if the camera feed goes unavailable for 5 minutes."
- Official Integration: Home Assistant Core - Proxmox VE
- Custom Integration: GitHub - dougiteixeira/proxmoxve
9. Music Assistant (MASS 2.0)
Music Assistant
Audio OSThe Unifier: Audio in a smart home is usually a mess of different apps. Music Assistant acts as my "Music OS." The killer feature for audiophiles is the advanced audio pipeline. MASS supports DSP (Digital Signal Processing) profiles for specific players. It can normalize volume across tracks from different sources. Crucially, it enables Cross-Platform Grouping. You can group a Sonos speaker, a Google Nest Hub, and a DLNA receiver into a single sync group. MASS handles the stream synchronization and transcoding (e.g., converting FLAC to MP3 for legacy devices) on the fly.
The Fix: It lets me group my Sonos speakers, my Google Nest Hubs, and my DLNA receivers into one sync group. It handles all the syncing and transcoding. It basically turns my mixed-bag of hardware into a unified whole-home audio system.
10. Spook
Spook
System UtilityThe Janitor: Created by Frenck, this isn't a normal integrationโitโs a toolbox. Spook runs in the background and yells at me when I mess up.
- Dangling Entities: It finds automations that reference devices I deleted months ago.
- Repairs: It creates "Repair" issues in the dashboard, pointing me to the exact line of YAML that is broken. It keeps my system clean and efficient.
11. Bubble Card
Bubble Card
Mobile FirstFinally, the frontend. Standard dashboards look terrible on phones. Bubble Card changes the UI to feel like a native app. The defining feature of Bubble Card is the Pop-up. In a traditional dashboard, detailed controls (like a thermostat dial or color wheel) take up massive screen real estate. Bubble Card hides these details behind a button. When tapped, a smooth, animated overlay (pop-up) slides up from the bottom of the screenโmimicking the behavior of Apple HomeKit or Android Device Controls.
- Pop-Ups: Instead of cluttering my screen with thermostat dials, I have a clean button. Tapping it slides up a smooth pop-up menu (just like Apple HomeKit).
- WAF (Wife Acceptance Factor): High. Itโs intuitive, fast, and looks professional.
