Why

Dumping MQTT logs to a Telegram bot does not allow me to derive valuable insights. I want to have more information about my device usage patterns, monitor overall health metrics and control some of my devices remotely.

My previous setup had too many layers of complexity and provided separate services for various use cases. I needed a simpler, centralized solution. Moreover, Apple being Apple™, Apple Home does not work without a Apple HomePod. I am locked out of my devices outside my house, even if my phone is connected to an overlay network. That means I cannot interact with any of my devices without paying for a HomePod.

Ingredients

• Server: Zigbee2MQTT is unparalleled when it comes to eliminating the need for proprietary Zigbee bridges. Grab a Raspberry Pi or HP t530 Thin Client for €30, and install a MQTT broker
• Zigbee adapter: An adapter compatible Zigbee2MQTT is required. I use SONOFF Zigbee 3.0 USB Dongle Plus at home and I find it very reliable
• Camera with RTSP: Ideally grab a camera without internet connection. Stay away from unknown and unreliable brands for security reasons. I configured an unused E1 Pro camera I have lying around
• RTSP to HLS converter: Video stream must be consumable in browser. Hence, conversion to m3u8 is needed. Previously I hosted ffmpeg instances but currently I use a native converter called RTSPtoWeb
• Sensors: Any compatible with Zigbee2MQTT. Here is a list of supported and affordable contact and occupancy sensors I previously used
• Lights (or plugs connected to lights): For this project, I will use a plug to control the light located in my studio, as I don't want to screw and unscrew my lights. Here is a list of supported and affordable plugs and lights I previously used

You need to bring your ingredients and deployment strategies of your choice. This project only contains the React front-end that allows you to monitor health, control devices and watch the stream, and a back-end to process and serve the data end-points.

Caution

It is not advised to expose this dashboard to internet. Use peer-to-peer overlay networks such as Nebula to access the local computer where the dashboard is running.

Preparation

You will need

• rustup
• node
• pnpm

You may choose to use asdf and use pinned node and pnpm versions for straight-forward application runtime management. However, installing rust via rustup is recommended.

asdf plugin add nodejs
asdf plugin-add pnpm
asdf install

Installing dependencies and building the project

Logging MQTT messages

Check root and workspace package.json files to see available scripts. Make sure mqtt-sqlite is running with a complete device configuration. Database package reads the configuration file and database created by that repository.

Running the GraphQL server

Once mqtt-sqlite is running and logging MQTT messages, clone this repository where mqtt-sqlite resides. Head over to the server package and build a docker image, then run the image by mounting the shared directory where configuration.json and database.sqlite are located. These files will be created when mqtt-sqlite is setup and shared with the server package.

docker build -t mqtt-sqlite-server .
docker run -dti -v /home/koraybey/mqtt-sqlite/shared:/data/shared -p 4000:4000  -e DATABASE_URL=shared/database.sqlite -e CONFIG_URL=shared/configuration.json mqtt-sqlite-server

Setting up the video stream

Obtain the .m3u8 URL for your stream and change the existing url in @/components/Video.